This is a repository copy of The impact of AI, machine learning, automation and robotics

on the information professions: a report for CILIP.

White Rose Research Online URL for this paper:

https://eprints.whiterose.ac.uk/174522/

Version: Published Version

Monograph:
Cox, A. orcid.org/0000-0002-2587-245X (2021) The impact of AI, machine learning,
automation and robotics on the information professions: a report for CILIP. Report. CILIP

© CILIP 2021. Reproduced with permission from the copyright holder.

Reuse
Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless
indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by
national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of
the full text version. This is indicated by the licence information on the White Rose Research Online record
for the item.

Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing [email protected] including the URL of the record and the reason for the withdrawal request.

[email protected]
https://eprints.whiterose.ac.uk/
The impact of AI, machine learning,
automation and robotics
on the information professions
A report for CILIP
Research report: The impact of AI, machine learning,
automation and robotics on the information professions

This report is published by CILIP: the library and information

association with the support of Health Education England.
It was written by Andrew M. Cox (Information School, University
of Sheffield).
With grateful acknowledgement to Sue Lacey Bryant, National
Lead NHS Library and Knowledge Services for her support and
guidance in the development of this report.

About CILIP
CILIP is the UK library and information association. We are
the only chartered body in the world dedicated to uniting,
supporting and advocating for information professionals
and librarians – the people who help the world make better
decisions. Our membership is open to everyone working in
libraries, information or knowledge management, data science
and analytics or a related professional role. We work with
employers, learning providers and suppliers across the library
and information sector in the UK and internationally to develop
talent, promote innovation, encourage workforce diversity
and ultimately to secure the long-term future of our profession.

Reference group
The author would like to extend this thanks to the following
colleagues who generously took time to share their insights with
the project:
Stephen Ayre (William Harvey Library, George Eliot Hospital),
Ken Chad (Ken Chad Consulting), Alex Fenlon (Library Services,
The University of Birmingham), Julie Glanville (Independent
consultant in Information Retrieval), Felix Greaves (Director,
Science, Evidence and Analytics at National Institute for
Health and Care Excellence (NICE)), Jake Hearn, Timothy Jacobs
(The Christie NHS Foundation Trust), Indra Joshi (Director AI, NHS),
Niel Kempson, Tony Lewis (TmL Consultancy), Stephen Phillips
(vice chair CILIP K&IM committee), Alexandra Pooley (Chair SLSIG),
Mia Ridge (Digital Curator, Western Heritage Collections, Digital
Research team, British Library), Hélène Russell (The Knowledge
Business), Tony Russell-Rose (Goldsmiths, University of London),
Edward Saperia (Newspeak House), Philip Scott (Reader in Health
Informatics, School of Computing, University of Portsmouth),
Matthew Soare (Makerspace Development Manager, Hull Central
Library), Alex Smith (Global RAVN Product Lead), Marion Spring
(Associate Director, Information Services, National Institute for
Health and Care Excellence (NICE)), Martin White (Managing
Director, Intranet Focus Ltd).

Published in May 2021.

© CILIP 2021
RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Contents

03 Executive summary

05 The context and the brief

The context
The brief

07 CHAPTER 1 Automation in the 4th Industrial Revolution: grasping the complexity

12 CHAPTER 2 The promise, risks and threat of AI and robots

2.1 The promise of AI and robots

2.2 The ethics of AI
2.3 The threats to the profession from AI and robots

17 CHAPTER 3 Applications of AI and robots in information work

3.1 Intelligent web and mobile search: the need for algorithmic literacy
3.2 AI interfaces to existing knowledge discovery systems
3.3 AI in knowledge discovery: the need for socio-technical infrastructures
CASE STUDY Living systematic reviews
CASE STUDY Library research collections and the creation of a community of data scientists

3.4 AI interacting with users: new ways to connect

3.4.1 Conversational agents aka chatbots
3.4.2 Voice assistants
3.5 AI managing people: questions about ethical use
3.5.1 User management
3.5.2 Human resource management
3.5.3 Marketing

3.6 Robotic process automation (RPA)

CASE STUDY The ad hoc application of AI to a specific information challenge

3.7 Smart spaces and robotics

3.7.1 Smart spaces
3.7.2 Robots and libraries

3.8 4th Industrial Revolution automation: the need for AI, robotics and data literacy
CASE STUDY Create Crates

A REPORT FOR CILIP 01

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

27 CHAPTER 4 The challenges and opportunities

4.1 The hype challenge: the need for understanding and a vision
4.2 The ownership challenge: the need for collaboration
4.3 The procurement challenge: the need for expertise in licensing, copyright and IPR
4.4 The data challenges: the need for data stewardship
4.5 The technical infrastructure and workflow challenges: the need for information
architecting
4.6 The post-implementation challenges: the need for support, training and promotion
4.7 The ethical use challenge: the need for responsible use rooted in professional
values and ethics

33 CHAPTER 5 Competencies needed to take the opportunities presented by AI and robots

5.1 The relevance of existing information professional competencies

5.2 Computational thinking, fusion skills and soft skills
5.3 Data science and data stewardship

39 CHAPTER 6 Strengths and vulnerabilities of the profession

42 Recommendations

For CILIP
For information services and libraries
For individual information professionals
For educational institutions at all levels
For other training providers
For research

44 References

48 Appendixes

52 Resources and case study sources

02 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Executive summary

The purpose of this research is to help CILIP and our professional

community to understand how AI, machine learning, process
automation and robotics are either already impacting the daily work
of information professionals or likely to do so in the near future.

Specifically, it seeks to answer the following questions:

How do we ensure that today’s workforce has the skills and understanding they
need in order to enable them to support their users in participating safely and
successfully in a modern world that is increasingly powered by artificial intelligence
(AI), machine learning, process automation and robotics?
What are the ethical implications of our approach to these technologies –
how can we deploy the existing ethical framework for librarians and ensure
that it aligns to emerging work on Data Ethics and responsible technology?
What should the skillset of the future workforce look like and what is the
curriculum by which we will ensure that the next generation of information
professionals have the skills to keep pace with future developments in technology?

The primary function of the outputs of link between AI and the wider digital
the research will be to drive the refresh transformation, and its interrelations with
of CILIP’s Professional Knowledge and other social changes. The complexity of
Skills Base, and from this the “ecosystem” the issues is summarised in chapter 1.
of teaching, learning and professional
Chapter 2 provides an overview of how
development for current and future
AI and robots are relevant to the work of
information, knowledge management
the profession. This arises particularly from
and library professionals.
how it might change how text and other
The data for the project was an extensive forms of content are described, searched
literature review and conversations with and used. This could constitute a shift from
twenty-one experts from across the UK the paradigm of searching for items to
profession conducted by the author in read manually to the model of extracting
November and December 2020. knowledge from collections of text, images
and other data through algorithms. But
In the interests of direct expression the
there are changes happening across
phrase “AI and robots” is used in the
the whole information value chain: in its
report to define the scope of technologies
production, organisation and consumption.
under consideration, although there was
The technologies open up exciting
little consensus about terminology in the
opportunities for a more information rich
literature or among interviewees.
world. However, they also pose a large
For many reasons this is a complex topic: number of ethical concerns. They will also
partly because of the lack of agreed have a potential impact on jobs, including
terminology, but also because of the on the work of information professionals.
emotive responses it evokes, the sectoral
patchiness of its impact, the complex

A REPORT FOR CILIP 03

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

The most obviously relevant change Many of the competencies needed to take
AI has effected is in web and mobile these opportunities already exist at some
search, so chapter 3, which identifies level in the profession and are reflected
the applications of AI and robots in in the CILIP Professional Knowledge and
information, knowledge management and Skills Base (PKSB). Chapter 5 presents For many
library work, starts with this. Changes to these. A longer discussion is offered about reasons this
web and mobile search imply a need for the repositioning of the profession in
what has been called algorithmic literacy, relation to computational thinking, data
is a complex
as an aspect of information literacy. AI analysis and data science, the fusion skills topic: partly
will also increasingly appear in interfaces to manage the relations between humans because
to knowledge discovery systems licensed and AI, and the soft skills that AI are unlikely of the lack
or run by information services. However, to develop in the near future. A concept of
it is in its impact on knowledge discovery data stewardship is also defined.
of agreed
that AI will have the greatest impact. What terminology,
The information, knowledge management
types of socio-technical infrastructure
and library professional is in a strong but also
are needed to support this and how because of
position to help organisations move into
information professionals might be
involved varies a lot between contexts.
the 4th Industrial Revolution deploying AI the emotive
and robots, which is reiterated at the start responses
Conversational agents or chatbots offer
of chapter 6. Yet some vulnerabilities
a new means to communicate with
remain, particularly in relation to adjacent it evokes.
users. AI might also be cautiously used
professions. These vulnerabilities can be
in managing people, but the ethical
partly addressed through the thirteen
issues here are significant. Automation
recommendations the report makes
of routine administrative tasks including
to six different stakeholder groups.
within AI pipelines is facilitated through
Of particular importance are:
robotic process automation. Sensor data
could be used to make buildings such as o For CILIP to articulate the relevance
libraries smarter to improve design and of the profession’s skills, values and
for wayfinding. Four potential uses of ethical principles, to identify pathfinder
robots in libraries are in handling books, organisations and people who encapsulate
acting as an information desk, for learning, these possibilities, and to promote
and for generic tasks such as cleaning or knowledge sharing within the profession.
through robotic furniture. Finally, there is
a potential library role in promoting public
o For individual information professionals
and information organisations to explore
understanding of AI and robots, through
the use of AI tools and share what they
AI and data literacy. Case studies provide
learn with others across the profession.
more in-depth narratives around some of
the key applications and how information o For the profession to use the
professionals can be involved. technologies in their own work and to
support information users in engaging
Chapter 4 suggests how opportunities
productively and safely with them for
for information professionals are
social good.
created through the need to overcome
organisational challenges to reaping the
benefits of AI and robots, especially in
knowledge discovery. These include a
vision; collaboration; expertise in licensing,
copyright and IPR; data stewardship;
information architecture; support, training
and promotion; and responsible use rooted
in professional values and ethics.

04 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

The context & the brief

by Nick Poole, CILIP CEO (November 2020)

The context: Internet and the digitisation of print books

that characterised the late 20th century for
We are in the middle of a 4th our profession.
Industrial Revolution, every bit as
As is so often the case with these
transformative as the great industrial
large-scale industrial changes, the
and economic shifts that preceded it.
most visible part is technological. We
can ‘see’ the mainstreaming of mobile
Building on the great advances in
Internet-connected devices, the use
information technology and connectivity
of smart technology in the home, the
in the postwar era, this new revolution
emergence of AI-powered assistants in
is described by HM Government in their
the interfaces we use in our daily lives,
White Paper Regulation for the Fourth
the near-driverless cars, etc.
Industrial Revolution:
Harder to discern but arguably much more
“The Fourth Industrial Revolution is of a scale,
significant are the huge changes in human
speed and complexity that is unprecedented.
and physical infrastructure that enable
It is characterised by a fusion of technologies –
these innovations – the data centres built in
such as artificial intelligence, gene editing and
the desert powered by hydro-electric dams,
advanced robotics – that is blurring the lines
the expansion of network cables and the
between the physical, digital and biological
emergence of a workforce with the skills
worlds. It will disrupt nearly every industry in
needed to make it all work.
every country, creating new opportunities and
challenges for people, places and businesses The figures go some way to illuminating
to which we must respond.” the scale of social, economic and industrial
re-engineering that has been going on for
The White Paper envisages four ‘grand
much of the past decade (source: KPMG
challenges’ for the UK economy and
research data for CILIP/KPMG Report
industry if we are to position ourselves
Information as an Asset):
to take advantage of these changes:
To put the UK at the forefront of the o Chief Data Officers in large enterprises
had an average budget in 2017 of £6m,
artificial intelligence and data revolution
a 23% increase year-on-year since 2014;
Maximise the advantages for UK
industry of the shift to ‘clean’ growth o 10% of major corporations in 2017
predicted they would have a ‘Chief Robotics
Become a world leader in shaping the Officer’ in their senior executive team
future of mobility by 2021;
Harness the power of innovation to help o The total volume of data being created,
meet the needs of an ageing society managed and shared globally is predicted
to increase tenfold between 2016 and 2030
These challenges have a profound
(from 16Zb in 2016 to 162Zb in 2030);
implication for the future development of
the information, knowledge management o In 2020, 1.7Mb of data are being created
and library workforce. The pace of per person, per hour globally;
adoption, the ubiquity of the associated
technologies, the fundamental changes o The global market for AI and machine
learning technologies is predicted to grow
in user behaviour and the sheer scale of
from £16bn in 2018 to £143bn in 2025;
machine-readable information dwarf the
challenges of structured search on the

A REPORT FOR CILIP 05

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

o The global market for industrial and non- With these opportunities come
industrial robotics is predicted to grow from considerable risks– for example, a
£22bn in 2018 to £178bn in 2025; specialist UCL research unit has already
found examples of Internet-connected
o The number of ‘Internet of Things’ smart smart home devices being used for The scale
connected devices is predicted to grow
coercive control in the home by abusive and pace
from 27bn in 2017 to more than 125bn
partners. The literature and news are
in 2030.
full of many examples of unaccountable
of the 4th
Every day, the technologies associated with algorithms and AI-driven decisions which Industrial
the 4th Industrial Revolution – AI, machine exacerbate human failings such as sexism Revolution
learning, automation and robotics – are and prejudice. will
finding new applications. Examples include: The role of information professionals inevitably
o Fraud/risk detection and management has always been to enable our users and disrupt
in financial services communities to profit from new advances
in technology and to make better use of the library,
o Regulatory compliance at scale (for information for their own advancement. information
example with Data Protection legislation) This was as true of the cuneiform & knowledge
o Decision support collections of antiquity as it was of the management
great book and print collections of the past
o Automation of inventory across century as it will be of the vast collections workforce.
supply-chain, logistics and stock
of data and personal information being
management
generated and shared online.
o Use of chatbots to automate customer The opportunity and the challenge for
service enquiries
CILIP as a professional association is to
o Use of self-help apps, for example for ensure that we are able to use our ‘pipeline
self-management of long-term conditions of talent’ to prepare the information
professional workforce to play this role
once again in the 4th Industrial Revolution.
The brief:
Specifically, there are three key questions
The role of the information, which we will need to be able to answer in
knowledge management and order to ensure that our profession is ready
library professions. and able to fulfil this role:
The tremendous scale and pace of the o How do we ensure that today’s
4th Industrial Revolution and its associated workforce has the skills and understanding
technologies will inevitably disrupt the they need in order to enable them to
library, information and knowledge support their users in participating safely
management workforce. In 2017, McKinsey and successfully in a modern world that
estimated that 50% of current work is increasingly powered by AI, machine
activities were technically automatable, learning, process automation and robotics?
and that 60% of jobs have at least 30%
of their responsibilities that are capable o What are the ethical implications of our
of being automated. approach to these technologies – how can
we deploy the existing ethical framework
The opportunities are potentially for librarians and ensure that it aligns
immense – people will lead increasingly to emerging work on data ethics and
information-rich lives in which they can responsible technology?
use data and analytics to improve their
health, be more productive and release o What should the skillset of the future
more time for family, friends and leisure workforce look like and what is the
activities. In so doing, they will need expert, curriculum by which we will ensure
ethical and accountable information that the next generation of information
professionals to help them select and use professionals has the skills to keep pace
the right technologies, to maximise their with future developments in technology?
information-handling skills and to keep
them and their families safe from harm.

06 A REPORT FOR CILIP

Automation in
the 4th industrial
revolution: grasping
the complexity
 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Automation in the 4th

industrial revolution:
grasping the complexity
There are many reasons why the pattern and implications
of the current wave of automation are hard to fully grasp.

the first automation initiatives in the 1930s

AI and robots are our past, right through to automation becoming
present and future. “business as usual” in the 1990s.
When will automation impact us? All the While it is true that AI, robots, even machine
technologies behind AI and robots have a learning have a long history, and some
long history. There seems to be a tendency promises of technology never materialise,
to treat AI applications that have already at the same time it is also true that things
been implemented as not counting as AI. are changing rather fast and things that
For example, many of the familiar features seemed impossible a few years ago are
of Google and Amazon are powered by being achieved. Some of these changes
AI. Digitisation techniques such as OCR could constitute a disruptive change. They
are a form of machine learning. RFID is could vastly improve information access;
an instance of the Internet of Things. The they might change information professional
introduction of book returns and sorting jobs and indeed in some sectors such as
machines in physical libraries is effectively health, they already are.
an application of robotics. This report is
not about these technologies, but it is
relevant to remember how the process
AI and robots are ideas rather
of implementation happened and ask how than specific technologies.
the implementation of the latest round of
If they were specific technologies their
automation might be the same or different.
precise impact might arguably be relatively
While the roots of current changes may lie easy to delimit. But in fact they are both
in the past, it is also important to remember ideas: ideas about machines emulating
that the diffusion of innovation takes time. or even replacing humans. AI imagines
Proof that a concept is potentially valuable machines that can reason, make decisions,
through special projects does not always learn and interact like humans. Robots
lead quickly to organisation or sector wide mimic or substitute for human bodies.
change or necessarily to change at all. The idea of a machine that can think or has
For example, chatbots are widely used in a body has been around a long time and
some industries and have been promoted there is a strong public imaginary of what
for a while as relevant to information they might mean, which we see expressed
services, such as libraries. There does not through the media and popular science
seem strong evidence that the uptake has fiction movies. The notion of automation
been what might have been predicted. also has a long history. The figure of the
This could be because of the nature of handloom weaver swept away by the
reference enquiries being too complex first industrial revolution echoes in our
or the continuing value of human contact. minds. The long cultural history attached
Self-issue and returns systems spread to ideas around automation, along with
more quickly, but it did take years for it the spectre of job losses, complicates
to become business as usual, starting with our understanding of what is happening
right now.

08 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

AI and robots produce emotive information services. This carries a danger

of making the specialist tools we support
responses. looking outdated.
Are AI and robots a “good thing”? The On the whole, we can guess that most AI
immediate arrival of the technologies is will be developed for application in high Most AI
surrounded by excitement and some hype, value tasks and then at a later date applied will likely
secrecy and fear. AI promises to make in information contexts, such as libraries. be developed
information more accessible, through This means that what is being adapted to
indexing material more quickly and in new
for high value
our context may often make assumptions
ways, such as by permitting searching by that do not fit our usual way of doing applications,
voice command. Excitement is well justified, things or the language we use to describe then later
particularly for information professionals things. This pattern of development also applied in
whose core value is that everyone has potentially limits our ability to shape
access to the information they need.
information
the technology.
contexts –
Yet in some very competitive sectors Reflecting the patchiness in change,
AI is being used as marketing, to present potentially
because the information sector is so
organisations as innovative and on broad and the roles we carry through limiting our
the cutting edge. For the same reason quite diverse, when and how AI and robots ability to
precisely how it is being used is kept impact on the information profession will shape the
secret. Sometimes this may mask that also be patchy. Specifically, the differing
there is little that is being offered that technology.
levels of digital maturity across sectors will
is new. At the same time much media shape their impact. How AI might impact
coverage focuses on the fear that AI and a research library will be quite different
robots will steal people’s jobs and make from how it impacts a legal information
bad, biased decisions. High profile failures professional, a knowledge and information
such as of the algorithm used to calibrate service in government or the health service,
A level results in the summer of 2020 cast or an information professional working in
suspicion on the technology. The emotion a bank. Large organisations are more likely
and secrecy is not very helpful for getting to have the resources and data to apply the
a clear picture of how the technologies latest automation technologies, so even
will impact our profession’s work. For within sectors the impact will be patchy.
information professionals working in
such highly competitive sectors this could How information professionals are affected
be a particular challenge. In other, less is connected to how AI is impacting their
competitive, more open sectors the new wider sector. The “smart library” could
technologies can be appraised in a cooler sit within a “smart campus” and beyond
light. Yet in all cases AI does carry with it that a “smart city”, so information related
a host of ethical issues. applications are part of a wider nexus of
changes, making the nature of the future
more unclear. The organisations within
AI and robot adoption is patchy which information services operate may
geographically and by sector. be transformed with many effects on
information professionals. For example,
Where are AI and robots being applied? one thinks of the wide range of applications
Another complication making it hard to of AI identified by NHSX (2019) both in
understand patterns of change is that clinical decision making (e.g. in radiology
the impact of technologies arrives in and pathology) but also in operational
different sectors of work and different aspects (e.g. managing waiting lists) (see
parts of the world at a differential rate. also Stanfill & Marc, 2019). Faggella (2020)
What is futuristic in one industry or in one sees six main ways that law firms are
country has already happened in another. applying AI to their operations: in due
Many of the key aspects of AI are already diligence, predicting litigation outcomes,
quite familiar in technologies such as in legal analytics, analysis of IP portfolios,
web and mobile search (auto-suggest, automation of document completion, and
relevance ranking, recommendation, billing for partners’ time (see also Kroski,
personalisation) or word processing (auto- 2020; Neary & Chen, 2017; Smith, 2016).
complete, spell checking, translation, voice Some of these directly impact on traditional
recognition). However, they have yet to areas of information work; others may have
arrive in most search systems used in ripple effects on them. In the university

A REPORT FOR CILIP 09

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

context, Cox (2021) identifies a wide range because of its connotations or lack of
of applications in teaching (Intelligent agreed definition. Yet other terms, such
Tutoring Systems, conversational agents as machine learning, although an important
in MOOCs, plagiarism detection), research reference point, are too narrow. For the
(text and data mining, robot scientists, purposes of the report we will prefer the
automated peer review) and university term AI because it usefully captures the
administration and estates management range of technologies supporting 4th
(the smart campus). This means that Industrial Revolution automation. The
information professionals embedded in term AI is often used in the literature to
their sector may see very different patterns encompass robots, but sometimes it will
of change happening around them. be useful to differentiate AI and robots.
Information professionals work in every
There is also a blurred boundary between
sector so a comprehensive perspective
the bundle of technologies addressed
is impossible.
in the report and others, such as virtual
reality (VR) and augmented reality (AR) that
AI is an umbrella term, are also making their own way through

not a single technology.

the famous Gartner hype cycle. Perhaps
their application will be interlinked with AI;
What is AI? Terms such as machine several authors suggest so. But this does
learning, natural language processing, make it very hard to delimit the report.
voice recognition and computer vision We have decided to exclude VR, AR, and
refer to some of the computational blockchain from direct consideration
techniques associated with the idea of AI. here, but this may mean that some trends
The next section explores in more depth are missed.
the range of technologies involved, but
Even within the technologies often
it is important to say that each has its
associated with AI there is significant
own history and impacts. It might have
difference. For example, robotic process
been easier to delimit the topic of this
automation (RPA) automates routine tasks.
report if we had focussed just on machine
It increases speed and reliability of carrying
learning, say. But the wider scope set is
through routine tasks. The argument for
productive in reflecting that there are a
using RPA relate to cost saving. In contrast
range of technologies and interrelated
many of the claims behind artificial
developments involved. At the same time
intelligence promise to do familiar things
AI is premised not just on clever algorithms,
in new types of way. Their allure lies in the
but also on modern computing power
promise a true digital transformation and
and on data. Many of the algorithms were
disruptive change.
written some time ago; it is only now that
the computing power exists to execute
them fast enough to create usable services AI and robots will create, change
from them. And, it is only now that there
is enough data to allow them to be useful. and destroy jobs.
What will the impact of AI and robots be on
There is not yet agreed terminology.
jobs and skills? AI will replace human work,
Each of the technologies behind AI has
There is not its own history. None are wholly new. But
particularly repetitive forms of knowledge
and service work. Frey and Osborne (2017)
yet agreed sometimes this is partly masked by shifts
predict the disappearance of many jobs,
terminology. of terminology. A few years ago there was
including “clerical library assistants” and
Each of the a lot of talk in parts of the profession about
“library technicians” (though these roles are
text and data mining (TDM), linked to new
technologies not clearly defined). Robots will also replace
exception to copyright creating a right to
human labour, again particularly in highly
behind AI mine texts. Now we would probably talk
repetitive, predictable tasks. Similarly,
has its own much more about AI or machine learning.
McKinsey (2018) suggest that jobs based on
history, but This makes tracing the growing impact of
both basic cognitive skills and physical and
AI a little more difficult.
none are manual skills will shrink. This change could
As is typical of an arena of change, be in the form of automating the most
wholly new.
terminology has not settled down. Some repetitive tasks, allowing the employee to
interviewees for this report, for example, focus on the more interesting aspects of
felt very comfortable with the term artificial jobs, such as those requiring imagination,
intelligence; others thought it unhelpful, problem solving and collaboration.

10 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

It could lead to role expansion. Work

could be rehumanized (Wendehorst,
AI and robots are part of wider
2020). But it could lead to de-skilling too, social changes.
a more polarized workforce or perhaps to We also need to factor in other changes
employee disempowerment (Coombs et al., to our society that will interact with the AI will
2020; Wendehorst, 2020). new technologies. This includes both replace
AI and robots will also create many jobs: dramatic events such as Brexit and the repetitive
people developing tools, analysis, and for COVID-19 pandemic, and more long-term
social trends such as ageing population
forms of
domain experts. Indeed, for the foreseeable
future most AI will keep a “human in the structures or a growing concern with knowledge
loop” to ensure that human judgement environmental sustainability. These wider and service
can shape critical decisions. Many experts changes do interact with the current wave work, and
on AI differentiate weak and strong AI, of technology. For example, one of our
robots will
and see weak AI as a near reality in a way interviewees thought that the urgency
strong AI is not. Weak AI does a specific task created by COVID-19 has pushed ahead replace
very well and probably more accurately innovation in the automation of systematic highly
and faster than any human. Strong AI reviewing faster than might have been repetitive,
that applies a general understanding expected. The Black Lives Matter movement
predictable
to different contexts is much harder to seems to have helped change the moral
achieve, perhaps even impossible. It is climate within which bias in AI systems labour.
certainly many years off. The same may is being challenged. And although often
be true of robotics, where robots that presented as entirely immaterial, as virtual,
perform very narrow specific tasks, such AI does have significant environmental
as sweeping the floor, are far more likely impacts through its demands on the
than ones that can perform a range of tasks power system. It also creates potentially
as in our idea of a human-looking domestic exploitative relations, e.g. for clickworkers.
robot. So McKinsey (2018) anticipate the Untangling such complex nexuses of
greatest growth in technological skills. change is beyond the scope of this report
They also predict a growth in the need for but is relevant to how the reader needs
higher cognitive skills, though they see a to use its findings.
decline in the value of “advanced literacy
and writing” which may be significant for The future with AI and robots
some information professional work. But
they also see a shift in increased emphasis is not predetermined.
on social and emotional skills, areas where Do we have power to influence how AI and
automation is much harder. Thus, the robots are used? A final point to emphasise
balance in our skillsets may be shifting. is about the nature of technological change.
Sometimes technologies are presented
AI and robots are part of as external forces that cause inexorable
changes to our lives as if from outside.
the digital transformation. This form of technological determinism
Where do AI and robots fit into wider does reflect how we sometimes experience
changes? A further complexity is that we change, particularly when it is badly
need to reflect on how the current wave managed. But in truth technologies are
of technologies fit into the wider digital social creations. Ultimately it is societies
transformation. Is it a step change in the that make choices about technologies
digital shift or just one more step on the they develop and use. By extension the
move to digital? If the latter we might information, knowledge management and
need to be thinking how to adapt our library workforce has opportunities and
existing models rather than creating wholly an important role in choosing about the
new ones. collective future through how it responds
to the current wave of technologies.

A REPORT FOR CILIP 11

The promise,
risks and threat
of AI and robots
 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

The promise, risks

and threat of AI and robots

In both these definitions AI is seen

2.1 The promise of AI and robots as encompassing a wide range of
AI applications are already with us, most computational techniques and includes
obviously in web and mobile search, in the different forms of robotics.
form of features such as ranking of results,
AI has been defined as: “computer
auto-suggestion and personalised results.
programs that perform tasks requiring
For everyday users, these applications are
intelligence when done by humans.”
experienced as helpful, informative and
(cited Holmes et al., 2019)
fun. AI is used to offer recommendations
in ecommerce sites. In daily information A contrasting definition is:
tasks such as writing a document we benefit “An AI system is a machine-based system that
from auto-correction. Automated Writing is capable of influencing the Environment
Evaluation tools (such as Grammarly) are by making recommendations, predictions or
increasingly being advertised to students decisions for a given set of Objectives. It does
and others. A little more novel are so by utilizing machine and/or human-based
auto-summarisation of text, auto captioning inputs/data to: i) perceive real and/or virtual
of videos (through speech recognition) environments; ii) abstract such perceptions
and translation of documents. While not into models manually or automatically; and iii)
working perfectly, these new tools make use Model Interpretations to formulate options
information tasks easier and quicker. for outcomes”. (OECD, 2019)
From these examples we can understand The computational techniques that make up
Gartner’s (Eliot et al., 2020: 5) description AI differ in importance in different settings.
of AI as a “general-purpose technology” For example, in the high tech sector,
and their claim that “AI will be in everything.” according to McKinsey (2019), machine
AI is often used as an umbrella term for learning is the most important technology.
a number of technologies. Thus Gartner In financial services robotic process
(Lowendahl & Calhoun Williams, 2018) refer automation is particularly important; in
to “six core interconnected AI technologies”: healthcare services, computer vision and
business analytics and data science; natural natural language processing. Given that
language processing, speech recognition information professionals work in every
The changes and text to speech; machine learning, deep sector it is not possible to say which of
learning and neural networks; machine these might be more important for us.
in how reasoning, decision making and algorithms; However, it is most likely the changes
text can be computer vision; and robots and sensors. in how text can be processed that will
processed Similarly, McKinsey (2019) identify a number impact information professional work
are most of inter-related strands of “AI capability” most strongly, because historically much
likely to in: robotic process automation; computer of our work revolves around text in various
vision; machine learning; natural language forms. Much AI is what we might have
impact
text understanding; virtual agents or called text mining five years ago (Anderson
information conversational interfaces; physical robotics; & Craiglow, 2017; Glanville, 2019; Glanville
professional natural language speech understanding; & Wood, 2018; Stuart, 2020). We are used
work most natural language generation; and to computers dealing with structured data
autonomous vehicles. like spreadsheets and databases. One of
strongly.
the promises of AI is to deal better with
“unstructured text” by analysing words to
identify concepts and cluster related ideas,
whether those texts be published books

13 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

and journal articles, grey literature, internal to search across whole collections as
know-how or even text harvested from datasets rather than pages, as Ridge (2019)
conversations online. Difficult to digitise puts it. Publishers are increasingly using AI
parts of texts such as tables may be easier for content enrichment, which automates
to process (Cordell, 2020). AI is improving bringing together content from different Having some
the digitisation of manuscripts through sources for the reader (67 bricks, 2017). grasp of AI
recognition of hand-written manuscripts.
Sentiment analysis even seeks to identify
As well as processing unstructured texts, techniques
other forms of information are handled will be
different types of feeling in texts.
better than before through AI technologies,
The ability of AI to analyse the content of thus computers are able to: important
texts may shift search away from being to many
1. Process inputs in the form of human
primarily through structured bibliographic
voice (e.g. voice search) and also to output information
databases to search of full text items or professional
in voice form.
whole collections using multiple potential
algorithms. The paradigm could be seen as 2. Identify the subject of images and jobs.
shifting from searching for an item (like a categorise them better, as in a search
book or online report) to read manually to engine being able to identify what an image
mining for knowledge in whole collections is of or enable searching using an image
of text data: from close reading to distant rather than a text term, or as in facial
reading. If this is the case, having some recognition, where it is able to match a
grasp of AI techniques will be important human face to a photograph. Computer
to many information professional jobs. vision is the area of computing which deals
Potentially, work information professionals with the extraction of information from
do through creating descriptive metadata images and video.
might become less necessary as
This means that computers can accept
computers perform this task, though this
a wider range of inputs as data, including
is debatable. The change will not reduce
images, sound recordings and videos.
the need to understand the ecologies of
Natural Language Processing (NLP) is
the production of information in order to
the term for turning text or audio into
conduct an effective search or the need
structured information. They can also
to critically evaluate results based on this
interact with people to receive and give
understanding. Yet as Tony Russell-Rose
data through voice or images. Chatbots
suggested in the interview for this report,
can draw on a knowledge base to interact
AI will impact on the production of texts
with a human in a more or less human
throughout the information lifecycle or
interactional style.
value chain:
Machine learning means that computers
o In producing texts. The first scholarly can be given training data from which they
text composed entirely by a computer
can identify useful patterns, without those
synthesising existing literature was
patterns necessarily being determined by
published by Springer in 2019. Publishers
a human in advance. Supervised machine
are widely using AI to manage the peer
learning is where training data that has
review processes (Thelwall, 2018).
been manually labelled is used to help the
o In acquiring texts into collections, computer to learn to recognise patterns.
e.g. in crawling for data or processing text In unsupervised machine learning, the
algorithm searches for patterns without
o In categorising texts and knowledge being given training data.
discovery. New tools will exist to
describe the content of a single text or There is nothing new to automating
a large collection. business processes, but what robotic
process automation (RPA) offers is relatively
o In curating and reuse of text. user-friendly tools to stitch together
o In consuming text. Users will be processes across different platforms.
supported to access text through AI This promises to make it easier to automate
tools such as summarisation which daily processes.
provides them quick access to content These abilities widen the range of data
or recommendation and personalisation that computers can accept, processes
whereby appropriate content is pushed they can carry through and outputs they
towards them. They may also be enabled

A REPORT FOR CILIP 14

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

can generate. It allows them to do things Equally the need for a critical awareness
in some sense akin to human thinking, about how knowledge is generated
learning, decision making and interacting. grows, and the information, knowledge
management and library profession has
The ability to train AI rests on the availability
Biases in of data. AI is premised on “big data”. As
an important part to play in this.
training data the points above illustrate, the range of Nevertheless, it is important to recognise
(and in the inputs and outputs that computers can that in many contexts these developments
computing handle has expanded to unstructured remain potential. The admittedly
text, voice and images. In this sense the fragmentary evidence suggests a great
industry definition of “data” has expanded. What interest in the technologies in all sorts
itself) librarians might call collections and more of organisations in the UK (AIIM Industry
systema- latterly might have been referred to as Watch, 2018; British Computing Society,
tically “content” can now be seen as data (Padilla 2020; Carter, 2019). But as section 3 made
et al., 2018). Books and journal articles clear, different sectors are at different
reinforce are data, as are grey literature, know-how, stages of acceptance of the technologies.
existing photo collections and archives of email or
unfair social online forums. This is not just an important
shift in language. This is a critical change
2.2 The ethics of AI
structures.
because many professional distinctions In addition to questions about the maturity
rest on differentiating types of content, of the technology, there is intense concern
such as between published books and at the societal level around the ethics of
articles in libraries, unique manuscripts in AI. Jobin et al (2019) report on no less
archives and historical objects in museums. than 84 statements attempting to define
Increasingly the boundaries blur and what constitutes ethical AI, including by
issues converge. the House of Lords, the European Union
and the OECD; the Royal Society and IEEE;
In addition, the digital shift has led to a
as well as Google, IBM and Microsoft. The
growing range of other forms of computer-
timeline in Fjeld et al. (2020) suggests that
accessible data. Online activities leave data
these concerns emerged in 2018 and 2019.
traces. Sensors embedded in objects and
So the impact of AI is surrounded by a
buildings gather more data about the world,
storm of controversy in society as a whole
including ourselves. We carry devices such
(Centre for Data Ethics and Innovation, 2020;
as smartphones that track our own activity,
Equality Task Force, 2020; Floridi & Cowls,
location and movements. Indeed, even
2019) as well as within the information
human feeling is claimed to be data for
profession itself (Association of Research
affective computing. This is associated with
Libraries, 2019; Cordell, 2020; IFLA, 2020;
a societal re-evaluation of data, with claims
Padilla, 2019). Despite this evidence of
that data is the “new oil”. Actually, this claim
concern, there is more to say because
was first made as early as 2005 by the
most of these statements originate from
mathematician involved in developing
the global North and questions remain
Tesco’s supermarket reward scheme, the
about the implementation of these codes
Clubcard. His point was that oil is valuable,
in practice (AINow, 2019). The following
but needs to be refined. Similarly, issues
paragraphs summarise some of the
of data quality, of data description, data
main concerns.
management and governance apply in the
AI context. If the data used to train an algorithm
is based on a biased sample then it will
This is in many ways very exciting from an
learn a misleading pattern. If the data we
information, knowledge management and
train computers on only reflects our past
library perspective, because an AI-driven
faulty decisions, then it will just repeat our
and data-intensive society is one based on
past biases. Computers trained on white
information and knowledge. The amount
faces do not necessarily identify black
of data available to humanity is growing
faces. This raises issues of fairness and
and in many contexts AI is making it is
non-discrimination. Much of the literature
easier to find information. The potential
points to the way that these biases in
to search across different formats of
training data (and in the computing industry
information is increasing. Computers are
itself) systematically reinforce existing unfair
better at supporting our understanding
social structures.
through interacting with us orally and
presenting new visualisations of data.

15 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Privacy is another concern. Because these A further dimension of the societal impact
technologies are embedded in other tools of AI and robots, is their potential impact
and sites as general purpose technologies, on work, including professional work.
the user may not be aware that they Section 3 already discussed the hard to
are gathering data about them, trying predict impact of the technologies on
to influence or nudge their behaviour or employment. The next section considers the
making decisions based on it. This raises implications for information and knowledge
issues of consent, the right to control over management and library profession.
how one’s data is used and the right to
be forgotten.
2.3 The threats to the profession
from AI and robots
In so far as we do have a sense that data
is being collected, AI can be creating
surveillance. Quite apart from our right to In addition to the wider issues and social
privacy, this awareness of surveillance can risks, AI and robots pose a number of
in itself change our behaviour, e.g. make us threats from a professional perspective.
reluctant to search for particular forms of One arises if AI can genuinely perform tasks
information. Surveillance impacts the right currently carried out by an information
to free expression. professional. For example, if it produces
Gathering data about us creates a summaries or indexing that are good
security risk, if it is compromised or stolen enough for their context of use, and
and appropriated. much more quickly than is possible to do
manually. Or if it can efficiently sort books
The operation of AI in personalising content and put them back on a library shelf. This
and supporting moderation has direct appears to be highly likely with very routine,
impact both on the freedom to form an predictable cognitive or manual work. It
opinion and on the right to free expression may do away with certain more routine
(Kay, 2018). work, especially those involving sifting vast
Because the technologies “learn” the amounts of data.
patterns it may not be easy for anyone, Currently most AI and robot work is
including system designers, to fully narrow. It requires a “human in the loop”.
understand and explain the outputs or Over time it may also prove to be possible
decisions they make. This opacity leads to to use AI and robots on more complex,
problems of explainability and transparency. unpredictable tasks, including those
How do we explain how the outputs of AI seemingly based on expert judgement. At
were arrived at? Who is accountable for a this point more skilled tasks may be lost in
mistake if a computer makes it? How can human jobs. For information professionals
an appeal be lodged? The shifting nature this represents another wave of fear of
of machine learning may also reduce the
The
“disintermediation”.
possibility of verifiability and replicability. information
There is also a risk that even though they profession
There are also fundamental issues are less fit for purpose, cheap to produce
around the human control over technology, and easy to use results from AI satisfy busy, have an
professional responsibility and the promotion unskilled users. There is even the risk that important
of human values (Fjeld et al., 2020). that information services are replaced by AI role in
These issues are particularly acute where even if the latter do not work so well at all, if weighing up
AI is used to make decisions that directly ill-informed decision makers fall for some of
the hype. Thus, AI implies the restructuring how ethical
shape people’s lives, influence or “nudge”
human behaviour, or wherever personal of work, in unpredicted and perhaps issues
data is involved. Indeed, some critics of AI disruptive ways. around AI
have gone much further in seeing AI as part However, because there are many apply in
of a wider movement towards surveillance challenges around AI in practice, there contexts
and the metrification of society, where only remain opportunities for information
things that can be counted are valued.
where
and knowledge professionals to
contribute. The opportunities can be more they are
Ultimately, it is vital that AI and robots
are used for social good. The information clearly understood if we examine the operating.
profession have an important role in applications of AI and robots relevant to
weighing up how these ethical issues apply information work.
in contexts where they are operating.

A REPORT FOR CILIP 16

Applications
of AI and robots
in information
work
 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Applications of AI and robots

in information work

3.
3.1 Intelligent web and mobile Some ranking is determined by
organisations paying for a better placing.
search: the need for algorithmic 4. Search engine optimisation is used to
literacy promote results in search results.

The first domain where both users and 5. Search engine results are affected by
information professionals have already data voids, keyword appropriation, filter
encountered AI is in search engines, bubbles, echo chambers, deep fakes,
through the many novel features of automated content moderation, etc.
search that have appeared there first, 6. Confirmation bias and unconscious
such as auto-suggestion, auto-correction, bias may impact our own information-
recommendation and personalisation, and seeking behaviours.
most recently voice search (Fernandez,
2016). Some of these changes have already This is an area where most information
happened or are continuing to happen. professionals are well informed and already
In many ways they are welcomed by the engaged with AI, though they might not be
information, knowledge management and used to calling it that. It is a positive finding
library profession as increasing easy and of this report that information professionals
ubiquitous access to information. are at the forefront of promoting better
understanding of AI in mobile and web
Yet with our focus on Information Literacy search and many of the same issues apply
and as Google sceptics we immediately across the range of AI applications.
understand the limits of some of these
applications of AI: encapsulated in terms
such as keyword appropriation, filter 3.2 AI interfaces to existing
bubbles, echo chambers, and deep fakes. knowledge discovery systems
A critical evaluation of search results Just as the first applications of AI that we
requires an understanding of the hidden have probably experienced are in web
biases. The term algorithmic literacy and mobile search, so the most obvious
has been coined to refer to the deeper application is in information service
understanding of the operations of search discovery systems. Such tools have existed
engines needed by users to critically for some time in third party subscription
evaluate search results (Bakke, 2020; databases. They may now be offered for
Sander, 2020). library discovery systems. One example
To have algorithmic literacy would be to application that is being publicised at the
have an awareness that: time of writing is Yewno, which is being
integrated with ExLibris. This offers a new
1. Search engines are driven by commercial way to explore library content through
motives, not simply by a desire to offer visualised concept maps (Gramatica
access to information in an unbiased way, & Pickering, 2017).
so that search results are not simply ranked
by popularity. AI also has applications in supporting the
creation of descriptive metadata within
2. There are layers of network bias conventional knowledge discovery systems
(Johnson, 2020) in how search results are (Flannery, 2020) or to identify items to be
produced and presented. weeded because of lack of use (Wagstaff
& Liu, 2018).

18 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

3.3 AI in knowledge discovery: There are many models of what a relevant

socio-technical infrastructure might look
the need for socio-technical like, such as the following (also summarised

infrastructures in table 3.3).

1. A project-specific assembly of digital
Probably the most important application
content and tools to analyse it. The
of AI arises from the way it creates potential
focus here is on creating a temporary
for information services to support new
infrastructure to accomplish specific
ways to analyse content. This could be for
project goals. An example could be a digital
digital humanities researchers or social
humanities project in a research library.
scientists to explore special collections
in a research library; for lawyers to mine 2. A local content collection, such as a
information from a know-how collection; know-how collection in a law firm. This could
for researchers to mine published literature be of one type of data or multiple. Bespoke
for new insights in an era of “big publishing”; AI tools are trained to analyse this data.
for government officials mining information The focus here is the unique value from
sources on an ad hoc basis to support advice the content.
to government ministers. 3. Digitised or born digital content in the
In order for this to happen, challenges special collections of a research library.
around data/ content, choice and training The focus here is the unique cultural
of AI tools, technical infrastructure (e.g. data significance of the content, many tools can
storage) and workflows, and of forging be applied. Multiple projects work with the
collaborations and increasing skills have to collection over time, often in complementary
be solved to create an appropriate socio- ways. Outputs from the project could
technical infrastructure. Here information enhance the collection, e.g. by creating
services have a strong potential part to play. new descriptive metadata, aiding
future discovery.

TABLE 3.3 Alternatives of socio-technical infrastructures to support AI in knowledge discovery

Content Technical infrastructure Information service role

A projects-pecific assembly of digital Temporary, researcher chosen. Temporary, ad-hoc. Background support.
1 content and tools.

A local content collection, such as a Locally developed, Permanent. Assembling content, training
2 know-how collection in a law firm. unpublished content. tools, supporting usage.

Digitised or born digital content in the Local unpublished content, Temporary for each project. Providing content,
3 special collections of a research library. often digitized. supporting usage.

4 A sector wide aggregation of content

available for mining.
Published content. Collective. Signposting the service to users.

A tool/ content agnostic infrastructure Any. For storage and tools. Building and promoting
5 within which bespoke collections of content infrastructure.
for different types of analysis are stored.

A publisher or aggregator platform Publisher owned Licensed. Licensing service

6 including content and tools – subscribed to or aggregated. and supporting use.
by an information service for users.

A semi-permanent assembly of content, Information service selected Semi-permanent. Controlling an integrated

7 infrastructure, tools and people for a very from published sources. solution.
specific, ongoing purpose, such as in a living
systematic review.

A support service that focuses on advice on None. Human rather than technical. Assembling expertise
8 tools, training and supporting collaboration. and building community.

A REPORT FOR CILIP 19

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

4. A sector wide aggregation of content

available for mining (as planned at one
time by JISC). Here the content is published
rather than local or unique content. The
Living systematic reviews focus here is primarily on the content with
some ready-made tools supplied.
This case study illustrates the issues around model 7, in table 3.3. In
the health sector (and increasingly in other areas such as education 5. A tool/ content agnostic infrastructure
and agriculture) systematic reviews are an important information within which bespoke collections of content
activity. Because they are critical to guiding clinicians on healthcare, for different types of analysis are stored.
but because of the high rate of publication in this field, the use of AI The focus here is on providing a technical
is increasingly attractive (Beller et al., 2018; Elliott et al., 2014; Grames infrastructure, which users can use in
et al., 2019; Jonnalagadda et al., 2015; Millard et al., 2018). So long as different ways.
the process and algorithms in use could be clearly defined there would
also be increased transparency and reproducibility, aspects which are
6. A publisher or aggregator platform
including content and tools – subscribed
important in a health context.
to by an information service for users. This
Systematic reviews are generally considered to have the offers an integrated solution, but at a high
following steps: cost and loss of researcher independence
in choosing content and tools.
1. Identifying that a review is needed or an existing one needs
updating 7. A semi-permanent assembly of content,
infrastructure, tools and people for a
2. Identifying a search strategy, including keyword and database
very specific, ongoing purpose. This is a
selection
complete solution assembled for a specific
3. Screening for items that meet inclusion criteria organizational purpose. A good example
4. Collecting the full text of results would be a living systematic review.

5. Evaluating the items found, including assessing risk of bias 8. A support service that focuses on
in trial results advice on tools, training to use the tools
and supporting collaboration, but does not
6. Writing the synthesis provide a technical infrastructure. The focus
The concept of the living systematic review (a self-driving SR) here is on people.
encapsulates the ambition to automate the whole process, so that Thus the model could be a temporary
the review can always be up to date at the level of each guideline. project (1); an infrastructure licensed
Many of the individual steps in the process have been approached from an aggregator (6); a locally created
with a view to automating them, with some, such as collecting full infrastructure to support projects on valued
text or supporting human screening, being easiest to achieve. Tools content (3); or an infrastructure to support
have also been developed to assess risk of bias in trials or to identify content collection and analysis (5); an
sample sizes in reviewed papers. Views differ on time frame within ongoing integrated infrastructure (7); or
which the various steps in the process can be automated. It is likely something more akin to a community (8).
that there will be a human in the loop for the foreseeable future, It could focus on local content (2,3) or
and so they might be better referred to as computer assisted reviews published or licensed content (4,6,7) or
(O’Connor et al., 2019). There remain barriers to acceptance (O’Connor be content agnostic (5,8). The collection
et al., 2019), in funding of development, managing crowd sourcing of content could be made by an
and licensing issues. There are also critical ethical issues in terms of information service such as a library (2,3,7)
ensuring transparency and considering the issue of liability if an error by a third party (4,6) or by researchers (1,5).
is made based on a computer decision. It could focus on temporary (1) or long-term
In the conventional systematic review, the information professional purposes (7).
plays a critical role at the stage of understanding the topic, choosing It could focus on content; technical
databases to search, defining a precise search to ensure all relevant infrastructure or tools; on skills and
papers are found, and perhaps in retrieving results. One of the most collaboration; or purposes. Or it could seek
achievable applications of AI is to help screen hits, by allowing a user to to provide all these.
train it to sift a large number of potential hits rapidly. This means that
techniques of precise searching become less relevant. This is exactly In reality there might be a mixing of
where many information professionals have tended to fit into the these options, e.g. where some elements
process, so the effect could be to diminish their role significantly. Yet of support (8) are supplied alongside
they may be able to contribute to areas such as licensing, creating a content platform (2,3) or licensing a
workflows, training and support. publisher platform (6).

20 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Library research collections and the creation of a community of data scientists

The way that research libraries can support digital humanists be mined, notwithstanding the text and data mining exception
and other scholars is a case of how information services can in UK law. They can offer an alternative to expensive publisher
be involved in the application of AI, and one on which much and aggregator offerings (model 6 in table 3.3). This makes
literature has been published (model 3 in table 3.3) (Cordell, the library central not just for digital humanists using the
2020; Lewis et al., 2015; Maxwell et al., 2018; Oliver et al., 2019; collection, but to the widening range of disciplines that are
Padilla, 2019; Wang, 2013). Rich research collections offer many using AI. Researchers from many disciplines are starting to use
opportunities for the application of computational techniques, AI, but need training and guidance on tools to use, that libraries
and there are many benefits to such projects for libraries in terms could supply. This is analogous to traditional roles in training
of reputation, creating partnerships, improving systems, learning and advice. The library or information service might also be in
about methods and digitisation, and acquiring new descriptive a position to fill the technical “infrastructure gap” by providing
metadata as a derivative from project outcomes (Ridge, 2019). space to store text data that is being mined (model 5 in table
There are still challenges in terms of copyright and intellectual 5.3). This might be better produced at a sector level (model 4
property rights (IPR), the General Data Protection Regulation in table 3.3). Finally, as a “neutral” space the library is well-placed
(GDPR), in the cost of working at scale, of reintegrating data to bring together users interested in AI to work together and
from projects into systems. Algorithms based on contemporary support collaboration.
data do not necessarily work on historic data, so there is a need
With the use of data science techniques across disciplines
for creating a significant amount of training data. Skills gaps in
(but not necessarily rooted in using library special collections
library staff can be addressed through running internal courses
or even licensed content) the centre of gravity may have shifted
with a strong hands-on element.
to broader support of data science. Oliver et al. (2019) argue
The library or information service could be central to this that the academic library is the natural locus for supporting
activity because of its collections, but in some cases there could the spread of data science skills across disciplines combined
be other reasons. Fenlon (LIBER, 2020) sees the library as a with the promotion of open science principles. This is partly
natural focus for AI partly because of its role in licensing and to be achieved by new collaborations within and outside the
copyright is important in dealing with publishers of content to institution. It illustrates model 8 in table 3.3.

Some of the forces shaping how this service questions, saving time for more complex
might evolve would be: queries. For certain purposes the way that
the user can remain anonymous may be an
o The purposes of analysis advantage. Data from interactions can be
o Cost or type of cost mined for further information. They might
o Skills base of the information service be delivered through apps or on web sites.
and the users
Ethical issues with chatbots include how
far to make it explicit that the chatbot is not
3.4 AI interacting with users: human and also how to design them not to

new ways to connect reinforce gender stereotypes through their

naming and behaviour.
3.4.1 Conversational agents aka chatbots It is obvious that there are many potential
Conversational agents or chatbots are applications of chatbots in handling user
tools that are programmed to simulate enquires. It is clear too that information
conversation, usually through text professionals might well have an important
interfaces such as instant messaging, or role to play in building knowledge bases
voice. They are widely used on commercial for chatbots. However, despite being by
web sites to respond to initial customer repute in widespread commercial use,
queries. They have some obvious they seem to be less used in information
advantages in terms of being tireless and contexts, such as libraries. This could be
available 24/7 (McNeal & Newyear, 2013; because queries are often too complex
Vincze, 2017). They can respond at the point or because of a desire to build up
of need, reflecting users’ preference for this relationships with the user. The difficulty
style of support. They can pick up routine of ensuring that answers address the user

A REPORT FOR CILIP 21

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

need accurately or fully might be an issue 3.4.2 Voice assistants

in some information contexts, e.g. where Voice assistants, also known as smart
a high degree of accuracy is required. speakers, such as Amazon’s Alexa and
The effort required to build up a knowledge Google Home, use automatic speech
base of answers or the technical challenges recognition technology to respond to
of building them could also be obstacles. verbal requests for information. Because
There might be particular circumstances they are becoming familiar through the
in which it is advantageous to build, growing level of mass ownership, they
e.g. in very busy services. have potential to be used by information
services. Information services can add
their own sets of answers to these tools
(Geary, 2019; Shih, 2020). Voice assistants
could respond to simple requests, offer
wayfinding information or be used to
deliver skills tests. Voice search has
also been applied to search library and
The Health Service Executive’s LAMA information service catalogues. Voice
activated tools could be particularly
LAMA (Library Ask Me Anything) is a customised version of Watson valuable to certain user groups, such as
Assistant and hosted on the Health Service Executive library website the elderly. There may be privacy and
as a chat interface. It has been programmed to respond to a range of security issues and concerns arising from
library enquiries and to redirect enquirers to the relevant sections of their differential ability to understand
the library website. It has also been configured to recognise symptoms certain accents.
and disease terms and pass a search through to a point of care tool, BMJ
Best Practice. It will also recognise common drug terms and search in
Medicines Complete. The primary use case for the AI assistant was to ease 3.5 AI managing people:
staff workload, reduce duplication and free staff time for in depth research
enquiries and service development. A number of specific targeted use
questions about ethical use
cases were identified: 3.5.1 User management
1. Responding to routine enquiries about OpenAthens accounts Information services collect or have
(setup, password activation and expiry). potential access to a vast amount of data
2. Responding to introductory enquiries about library services and access. about their users. This could also be used
3. Responding and direct to further information on locating library by AI to help describe, predict or influence
resources from specific discovery sources. behaviour. However, the use of AI to
manage user experience is a far more
4. Responding to the staffed virtual desk chat service where relevant.
contentious area than analysis of content.
5. Enabling AI powered discovery search within chat interface from a list While there are ethical, safety and legal
of library resources. challenges to the use of AI in information
6. Guiding users towards appropriate resources to address their enquiry. content, these are much stronger when AI
and robots use personal data about users
These have all be achieved to some extent and are being built upon.
or seek to “nudge” people’s behaviour.
Experience in the area of library/ learning
analytics in the educational sector may
be instructive here. Libraries potentially
collect a vast amount of data about
university students such as through
Roche Products Limited: LibraryAnn turnstiles, book circulation, online resource
usage (including detailed ebook reading
LibraryAnn is a chatbot designed by a one-person librarian working behaviours), interaction with staff or
in a specialist pharmaceutical company. Their work involves answering online chat. This could be combined with
copyright enquiries and supporting the UK organisation with finding other institutional data about students,
and accessing medical and developmental resources. Because they including their demographic data,
are a one-person librarian for a global organisation, they saw the value coursework performance, detailed online
in an internal chatbot service that could handle routine questions and learning behaviour and even social media
respond continuously, including outside the librarian’s working hours. activities. Some researchers have been
LibraryAnn answers around three questions a week on chat which strong advocates for combining such data
represents about half of the weekly information requests. Maintenance to inform decisions about services and to
takes around an hour a month. demonstrate library value.

22 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

AI could enhance this, e.g. by analysing There may be information contexts

chat log data, or be used to nudge where such AI applications feels safer,
students towards useful content or to again pointing to the differential impact
personalise material for users based on of AI across information, knowledge
their patterns of previous use. In practice, management and library sectors. For Just as the first
both library and learning analytics seem to example, monitoring social media applications
be underdeveloped. responses to public library services through
of AI that we
tools such as sentiment analysis might be
There could be a number of explanations have probably
deemed a legitimate use, if one regards
of this. There are practical obstacles such
social media content as in some sense experienced
as data silos and lack of interoperability
published. Murphy and Villaspesa’s (2020) are in web
(Oakleaf, 2018). Tsai and Gasevic (2017)
toolkit for museums suggests applications
identify a number of key organisational and mobile
to analyse TripAdvisor experiences or
barriers to the use of learning analytics, search, so the
modelling use of museum space (French
including the difficulty institutionally
& Villaespesa, 2019). These may have some most obvious
of leading such a profound change
application in public library settings. Text application is
management process, the fact that teaching
mining of survey data is uncontroversial
staff are not trained in using them and a in information
because permission has already been given
lack of evidence for a return on investment service
in consenting to participate in the survey
in the technologies. However, from a library
(Moore, 2017). Much of the barrier might be discovery
perspective, what may be critical are the
cultural. In the legal sector one use of AI is systems.
ethical issues:
in more accurately capturing lawyers’ time.
o The surveillance implied by monitoring
library behaviours may inhibit students 3.5.2 Human resource management
from searching for particular topics and Information services such as libraries
so freedom of expression. often have their own staff, and in some
o The benefits mostly accrue to the cases their own relatively autonomous HR
university in terms of efficiencies and sales, function. Although it is not central to this
rather than the students (Jones & Salo, study, it is worth reflecting on the potential
2018). uses of AI in the management of staff,
partly because they reinforce the concerns
o Consent seems rarely to be obtained. reflected in the previous sub-section.
Students do not seem to be aware that Faggella (2019) identifies some of the
learning analytics are being used by possible AI applications in HR, including
their teachers. in assessing applicants, on-boarding,
o Brooke Robertshaw and Asher (2019) monitoring engagement and predicting
argue that the rigour of the data analysis attrition. Yet Tambe et al. (2019) point to
in library analytics studies is suspect and a number of challenges to applying AI to
that the benefits in lessons learned do not HR, including the difficulty of quantifying
outweigh the ethical risks. aspects such as human performance, the
small amount of data most organisations
o Local policies rarely fully reflect ethical have on which to train AI (and this may well
concerns (Jones et al., 2020) and studies be deemed to instantiate past biases in
have rarely been approved by local ethical recruitment), and the fact that hiring and
review boards (Jones, 2019). firing decisions have such a strong effect
Jones comments (2019: 419): “Pursuing on human lives, meaning that the ethical
socio-technical practices that collect, issues become acute. While not a central
aggregate, analyse, and act on data concern for this report, the ethics of AI
revealing students’ intellectual behaviors in HRM is an important dimension of the
and interests is antithetical to the library issues in using AI.
profession’s commitments to user privacy
and intellectual freedom.”
This foregrounds the ethical dilemmas
around AI, further explored in the
next section.

A REPORT FOR CILIP 23

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

3.5.3 Marketing have to be repeated many times. It is used

The promotion of information services for tasks such as data entry and validation,
such as libraries is so important that a file manipulation, web scraping, data
brief sub-section on the application of quality checking, etc. An area of particular
AI in marketing was felt to be justified. interest is in automating information
Ma and Sun (2020) describe a customer’s governance compliance, e.g. by creating
journey towards a purchase, unknowingly alerts when improper or suspicious activity
shaped by AI: at the stage of an initial is detected (AIIM, n.d., 2019). Applications
search (results of which are partly are particularly popular in the financial
influenced by company advertising), sector. They ensure accuracy, increase data
through customised content displayed on quality and standardisation, but they are
an e-commerce site, by follow up ads and flexible enough to be quickly reconfigured.
through chatbot interactions. AI enables The software is increasingly easy to use,
advertising to be more interactive, with Microsoft recently launching Power
personalised and targeted. AI will be used Automate as part of the Office suite.
strategically throughout the marketing Library-specific applications in bibliometrics
process (Huang & Rust, 2020). There is work are described by Tam (2020). RPA
potential for the power of such methods could complement other AI applications
to be used by information professionals in to automate parts of workflows (Syed et
promoting services. Equally, they could be al., 2020).
perceived to be dimensions of the issue
around algorithmic literacy discussed RPA relieves the burden of very routine
in 3.1. administrative tasks. This could clearly
lead to loss of tasks and so perhaps jobs,
at least routine ones. It could also lead to
3.6 Robotic process job enrichment or free up time for staff
automation (RPA) to undertake higher level tasks. It clearly
creates a role for those who carry out
RPA is software that mimics the “path and maintain the automation system;
taken by a human through a range of some information professionals might be
computer applications when performing involved in this work. Schlegel and Kraus
certain tasks in a business process” (Syed (2020) identify the skills required to work
et al., 2020: 1). In that sense it works in a on automating tasks.
way akin to a macro, except across multiple
software applications. A simple example
would be to produce a software robot to
3.7 Smart spaces and robotics
repeat the process of taking an item from 3.7.1 Smart spaces
a list, using it to search on a web site,
A smart building uses sensors to collect
copying the text of the search result and
data about its status and usage and may
putting it into a spreadsheet or sending
use this to control lighting and heating,
an email including it. RPA automates
offer services such as wayfinding, or even
repetitive, routine back office tasks, that
power interventions that seek to shape user
behaviour (Hoy, 2016; Min-Allah & Alrashed,
2020). There are obvious benefits in terms
of efficiency savings and sustainability, but
potentially also in terms of providing the
basis for new information services.
The smart building concept is clearly
The ad hoc application of AI to a specific relevant to information services such
information challenge as libraries, because in many cases they
remain significant physical buildings.
An interesting example is described in the Government Information
Sensors can monitor movement, check
Group blog (2020). In this case a vast amount of data was being sent
occupancy or measure sound levels. This
daily at the beginning of the COVID pandemic to the Foreign and
could help understand how users are
Commonwealth Office crisis email from the UK’s 270 overseas missions.
moving through or settling in a library,
A team of information professionals used Power Automate to extract
with a view to improving its design. It could
data from emails and a wide range of resources into a COVID-19 library
help predict occupancy, and so help users
which could be accessed by many different stakeholders.
decide when to come to the library.

24 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

It could offer services such as to allow Thus, it is easy to imagine robots being
users to book spaces or for wayfinding. It employed in library buildings for:
could be used to nudge behaviour, such as o Cleaning
to encourage users who have not moved
for a long time to take a break, for their
o Delivery, including by drones (Nath, 2018)
well-being. The further one moves into the o Building maintenance
realms of seeking to influence behaviour, o Robotic furniture (such as furniture that
the more smart technologies raise the types automatically adjusts to the height of the
of ethical concerns discussed in section 5.5. user or shelves that move down to the user)
The smart library might be part of a smart
or intelligent campus and ultimately a smart
city, with data being integrated across these
3.8 4Th industrial revolution
different levels (JISC, 2019a). Many scenarios automation: the need for AI,
of use are imagined in JISC (2019) and the
Intelligent Campus blog (intelligentcampus. robotics and data literacy
jiscinvolve.org/wp). Some are as simple as AI and robots are being applied in a very wide
providing integrated travel information. range of social and organisational settings:
Others are more futuristic. from supermarkets to governments.

3.7.2 Robots and libraries One important implication for all information
Vlachos et al. (2020) propose that there are professionals is that a basic form of literacy
three main uses of robots in a library context: that they need is understanding of AI and
for navigation, book location and placement; robots: they need AI and robotics literacy.
as information desks; and for learning (see Algorithmic literacy could be seen as a subset
also Tella, 2020; Zheng, 2019). Reflecting on of this. Given its dependence on big data, AI
the first usage, we can acknowledge that literacy also implies data literacy. As citizens
book returns sorters have existed for some and employees, information professionals
time. Other uses relating to traditional book need this understanding themselves. More
stock, are to: specifically, AI, robotics and data literacy
is something that libraries can play a role
1. Retrieve books within large book in promoting.
storage facilities (automated storage
retrieval systems)
2. Locate books on standard shelves
3. Scan shelves to identify misplaced
and lost books (Liau, 2019)
4. Move large quantities of books
(Liau, 2019)
An embodied robot could also be used to Create Crates
welcome or inform users. A telepresence Wakefield public libraries run a service called Create Crates, circulating
robot enables a remote user to navigate maker materials to libraries that have neither the space or budget to
around a space (Guth & Vander Meer, 2017). accommodate permanent makerspaces. Each of the crates is broadly
An example of a learning use of robots based on a Science, Technology, Engineering or Maths (STEM) subject
would be to build a robot, perhaps from and allows branches to deliver fun, creative, STEM-based activities
a kit, as an educational exercise within an that support young people’s educational and social development while
information service or makerspace. Robots encouraging innovation, critical thinking, imagination and inventive
such as drones can be made available as a problem solving without the need for permanently dedicated space or
borrowable item. a budget for materials. Each crate includes the core materials necessary
to deliver STEM-based activities and ranges from low-tech items such
It seems likely that robots developed for
as modelling clay and Lego to high-tech items such as Raspberry Pis,
very generic tasks in other sectors, rather
electronic kits and a 3D printer. There are four types of positive impact:
than robots for library specific tasks, will
firstly, in enthusing children and their parents in STEM, including,
be the first to reach a scale of production
potentially, robots and AI; secondl,y in presenting the library service
where they become cheap enough for
as a place where people are encouraged to create, experiment and
widespread use.
explore ideas; thirdly, in boosting library staff skills and confidence in
dealing with topics such as AI; and fourthly in creating collaborations
with new partners.

A REPORT FOR CILIP 25

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

This is an obvious role for public libraries Appendices 1 and 2 summarise respectively
since their constituency is the whole Long and Magerko’s (2020) definition of the
population. Makerspaces are the ideal dimensions of AI literacy (which includes
place to promote in-depth exploration, understanding of robotics) and Prado
but less intensive introductions would also and Marzo’s (2013) data literacy definition
be relevant. But given the wide and deep (others are available).
impacts of AI and data, all information
services could arguably integrate AI
and data literacy into their information
literacy training.

26 A REPORT FOR CILIP

The challenges
and opportunities
 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

The challenges
and opportunities

As the previous section shows, AI There is also a lack of proven evidence

and robots have potential to reshape of benefit: do they actually do something
information work in many ways, including users or organisations need? Information
some fundamental, even disruptive ones. professionals are well positioned to
Yet often the conversations for this report understand what is really needed by
had the mood of the “trough of disillusion” users. There are still questions around
in the Gartner hype cycle. While the where the technology can make a return
technologies hold great future promise, on investment, be that through business
there are recognised to be many challenges transformation or simply through
to realising these soon; some technical, efficiencies. Yet sometimes the decision
some organisational and some ethical. makers who invest in AI do not see through
Although we are talking about a range of the hype and are attracted by the “shiny
different applications of AI in some very tech”. This can create the pattern of a
different contexts and in which information technical solutionism: technologies in
professionals are positioned differently, search of a problem.
there seems to be some commonality in
In some other sectors, such as the
what these challenges are.
academic sector, there seems to be some
These challenges can also be seen interest across all disciplines in applications
as opportunities for information and such as machine learning to analyse
knowledge professionals to step forward published texts or other data collections,
and contribute or even lead within but it seems to remain unclear which of
their context. the computational techniques are relevant
(Maxwell et al., 2018).
4.1 The hype challenge: the need In other contexts it is hard to find attention

for understanding and a vision

for such developments, when there are
many competing priorities (Carter, 2018).
There remains scepticism that many of the The hype challenge creates an opportunity
products being labelled as AI are truly novel for information professionals to play a role
or can fully deliver on vendors’ promises. in developing public and organisational
They are often perceived to be familiar understanding of AI and data, to identify
technologies rebadged. If they do offer where the real benefits of AI lie and so
Information something novel it is more limited than have an informed vision for how they
professionals the claim. How proprietary systems work can be implemented in ways aligned to
often act as is often a secret. As Callister (2020: 209) societal need and organisational mission
a bridge or comments in the legal context: and strategy.
translator “The best we librarians can do in the face The lack of vision is not universal. One
between of uncertainty [about what approaches to exception to this picture seems to be the
AI vendors are actually using] is to teach health sector, where the NHS is seeing its
different our users about the limitations of these future as very much digital.
groups, such systems, disillusioning them of computer
as IT services intelligence doing the work for them—at
and domain least for now. If anything, AI is a tool and,
one day perhaps—assuming a humanistic
experts. techno-central vision—a partner.”

28 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

4.2 The ownership challenge: Because it is a new market, there is little

understanding of the criteria to apply in
the need for collaboration procurement (Carter, 2019). The players
are many and unfamiliar.
Potentially linked to an uncertain vision,
Within
it remains unclear where AI sits within In the academic sector there is a
the organisation. It might be deemed to proliferation of free and open source tools, individual
be best located in IT. But it could be led but it is hard to navigate to which are the organisations
by a KM team, by business development best for particular tasks. the amount of
or operations, or perhaps it is a strategic
In fact, the procurement could also be data to feed
project. Some interviewees thought AI sat
of data as much as algorithms. AI is perhaps
purely with specialist data scientists.
Although UK law has an exception for not enough
This creates an opportunity both because
text and data mining for non-commercial to produce
of what information professionals can
research, there remain complexities around
offer from their own capabilities in any effective
the law, e.g. if working in a collaboration,
collaborations, but also because they results. This
how far partners can use licensed content
often act as a bridge or translator between
(Caspers et al., 2015; Kelly, 2016; LIBER, may place a
different groups, such as IT services and
2020; McNeice, 2017b, 2017a). Publishers limit on what
domain experts (Henke et al., 2018).
and aggregators have been accused of
is possible.
In the academic sector, there was a using techniques to block or inhibit the
sense of a widening circle of researchers legal right to mine, but the law is sufficiently
interested in using AI. This is likely to grow ambiguous to allow them to do this.
as more AI applications are applied in
The procurement challenge creates an
the economy and it needs to be taught
opportunity for information professionals
to students not just used in research.
to play a role in procuring both systems
However, there is a lack of coordinated
and data, drawing on their expertise in
support to do this. There was perceived
procuring systems, licensing content, and
to be a role in bringing academics
their understanding of copyright and IPR.
from different departments together,
as the library remains perceived as a
neutral space. 4.4 The data challenges:
In this respect specialisms like KM, with the need for data stewardship
its focus on building collaborations and
AI requires data. The internet giants
sharing knowledge, are likely to be very
such as Google and Amazon have vast
much needed.
quantities of data to train their algorithms.
The ownership challenge creates an This is where AI is already being applied
opportunity for information professionals successfully. But within individual
to play a role as bridges and translators, organisations the amount of data to feed
spanning between the technologists and AI is much less, perhaps not enough to
organisational purposes (Henke et al., produce effective results. This may place a
2018; Open Data Institute, 2020b). limit on what is possible. In the commercial
sector organisations are unlikely to
4.3 The procurement challenge: collaborate and share data. Interestingly,
one respondent to Carter (2019: 101) talked
the need for expertise in about the way that GDPR was “making big

licensing, copyright and IPR data disappear”. Even in the public sector
there also remain challenges around the
AI solutions (and data) often carry a openness of data.
significant premium (Carter, 2019).
In all sectors data that is available is often
Cost is certainly an issue in the public fragmented and poorly described and
sector. The vendor solutions are perceived organised. But for it to be used by AI it does
to be very expensive and there is an need to be managed better. This implies
interest in open source and do-it-yourself a need for organisations to improve their
solutions, though of course these have data management in order to leverage AI.
different types of resource implications.
In the academic sector, AI could be applied
Data processing and infrastructure costs
to historic special collections, but these
of AI could be high.
are very heterogenous (Cordell, 2020).

A REPORT FOR CILIP 29

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

The algorithms for image recognition AI. But this implies understanding how to
trained on contemporary images do not motivate potential participants. So it is far
necessarily work for historical images. As from being purely an IT infrastructure that
a result, off-the-shelf algorithms cannot be is needed.
AI has only easily applied. Training data is needed and
The workflow and infrastructure challenges
been adopted in fact, in all contexts, new types of analysis
imply the need for information architects.
on particular genre of text will require a
in a few lengthy process of training.
specific areas
In the research library context the
4.6 The post-implementation
at any scale. provenance of the collections being challenges: the need for support,
The unex- used as data has to be understood for
plainability the results of analysis through AI to be training and promotion
and “opacity” understandable. “Biases” in how historic Gaining acceptance for AI remains an issue
collections were built up and also in what to be tackled. AI has only been adopted in a
of AI is has been chosen to be digitised in the few specific areas at any scale.
problematic past need to be taken into account in
In many sectors, there are strong cultural
for its interpreting analysis.
barriers to adopting AI because many
acceptance AI also creates derivative data. For example, professional groups feel threatened by it
and use. in a research library context, AI use on and motivated to be doubtful that it can
digitised collections creates new descriptive deliver promises to automate tasks that
data. Reintegrating this into discovery have always before been seen as requiring
systems poses issues of quality. human skills to undertake.
The data challenges create an opportunity The unexplainability and “opacity” of AI
for information professionals in terms of is problematic for its acceptance and use
data governance and management and (Burrell, 2016). If it is hard to understand
long term preservation that might be how a result was produced, there will be
summarised in the term ‘data stewardship’. scepticism about how far to trust it. Many
This builds on existing expertise in the information uses require replication: this
information profession in information can be hard where machine learning is
governance and information description involved. In other contexts, there may be
(taxonomy building, use of standards). a temptation to rely too heavily on the
Since the data is often information content, outputs of AI without real understanding
collection skills are also highly relevant. of how it works and its limitations (Baker,
2018). Critical information literacy
4.5 The technical infrastructure remains relevant.

and workflow challenges: the There are many training needs (Carter,
2018). Domain experts need basic training
need for information architecting in how to use AI. AI is not advanced enough
to work on its own: there remains a need
In many industries current processes are
for there to be a human in the loop. There
complex and have strong local adaptations,
is also a need for trusted intermediaries
involve lots of human judgement and
who can interpret results. In the academic
remain somewhat undocumented. They
sector, while there appears to be widening
have not been looked at with an eye to
interest in using AI across all disciplines,
routinise them. In the commercial sector,
the skills are lacking to do the analysis and
offshored activities may be more routinised
interpret the results.
and so more open to AI applications, such
as RPA. There is much to be learned about So, there are emerging needs for support,
the implementation challenges (Pelz-Shapre training and the promotion of AI. These
& Kompella, 2019). post-implementation challenges create an
opportunity for information professionals
In the research library context, building
to facilitate the development of AI and its
workflows for AI systems is also a challenge.
implementation, e.g. through support,
It may be that libraries themselves have to
training and promotion. This builds
create an infrastructure for text and data
on existing roles of the same sort in
mining. In the special collection context, it
supporting information use and promoting
is also recognised that there is a value in
information services.
integrating crowdsourcing, e.g. to help train

30 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

4.7 The ethical use challenge: There is a great potential for bias in
training data which can lead to algorithms
the need for responsible use simply repeating past errors. The way that

rooted in professional values historic collections have been created

reflected past social values which may Information
and ethics now make them problematic, even deeply professionals'
problematic because they privilege certain
Previous sections (2.2 and 3.5) surfaced voices and reinforce the marginalisation
understanding
the cherished societal and professional of others (Cordell, 2020; Padilla, 2019). of the nature
ethical values that could be endangered Even digitisation initiatives in retrospect of data
by misuse of AI and robots: such as may be deemed to have reflected
fairness and non-discrimination, privacy, puts them
problematic biases.
the right to freedom of thought and in a strong
freedom of expression, explainability In many ways the issues are addressed by position to
and transparency, human control over the values and ethical principles articulated
by CILIP (2018) such as around privacy and evaluate the
technology, professional responsibility
and the promotion of human values. avoidance of bias. The development of less obvious
information skills and information literacy risks, such
Information professionals play a vital role resonate with the need for explainable
when they reflect on how these issues play as created by
AI and to ensure freedom of thought
out in the context where they are working. and expression. The profession has an datafication.
Their understanding of the nature of data important role to play in foregrounding
puts them in a strong position to evaluate ethical issues and finding ethical and safe
the less obvious risks, such as created ways to use AI and robots for the benefit
by datafication. of society. Yet one could argue that more
Privacy is not easily protected where work needs to be done to explore how the
masses of data is being joined up, even profession’s enduring values and principles
if it has been anonymised. Where data is apply to the challenges of AI, particularly
collected about individuals there are risks in terms of accountability, explainability
of creating an atmosphere of surveillance. and transparency.

A REPORT FOR CILIP 31

Competencies
needed to take the
opportunities presented
by AI and robots
 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Competencies needed
to take the opportunities
presented by AI and robots

Chapter 3 identified a number of

specific applications of AI and robots of
5.1 The relevance of existing
relevance to information professionals. information professional
AI in knowledge discovery is particularly
important. Section 6 identified the competencies
challenges and opportunities for Table 5.1 identifies the main activities
information professionals in enabling associated with AI and robots, particularly
organisations to benefit from new AI in knowledge discovery, identifies
technologies, particularly AI. This chapter the impacts information professionals
considers the implications for skills. It is can have, and maps them to the PKSB
particularly closely tied to the opportunity and longstanding areas of information
presented by AI in knowledge discovery. work, building on Cox et al. (2019) and
Carter (2020).

TABLE 5.1 The relevance of existing information professional competencies

Area of activity Impact PKSB Role extended* Challenges

Understanding and vision

Increasing understanding Organisations maximise Teaching and training skills, Information Lack of current
of AI, robots and data. the effective and safe use supporting users (Literacies literacy and understanding
of AI and robots. and learning). digital skills amongst users.
training.

Leadership – having The organisation Leadership skills, strategic Leadership. Hard to predict what
a vision for how the positions itself to benefit thinking and evaluation the technologies
information service/ strategically from the (Leadership and advocacy), will really do and in
organisation / society technologies, it identifies Business planning (Strategy, complex environment
can best use AI and where the technology planning and management) where main benefit
be involved in wider can have an impact in and communicating with will be; need a clear
AI use. either transforming stakeholders (Customer understanding of
the business or simply focus, service design the potential.
through improved and marketing).
efficiency.

Collaboration

Partnership creation Rich and supportive Partnership development Collaboration. Differing expectations;
for multi-disciplinary collaborations (Leadership and advocacy) internal politics.
teams of domain experts, come together to and communicating with
IT and other services reap the benefits stakeholders (Customer
to support AI. of the technologies. focus, service design
and marketing).

*Traditional information role being extended.

33 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Area of activity Impact PKSB Role extended* Challenges

Procurement

Copyright – The issues around IPR Information rights Copyright role. The law may lag behind
understanding what are managed. (Information governance technologies.
is legally allowed. and compliance).

Licensing of content/ Appropriate licenses are Copyright, intellectual Licensing New layer of concerns
data – negotiating/ negotiated and their property and licensing of content. in examining licences.
checking licences and terms are understood (Information governance
publicising licence terms. by users. and compliance).

Locating content. Necessary content Collection management Collection Anticipates what

is collected. (Collection management building. researchers need –
and development). unlikely?
Content contains
multiple layers of bias.

Licensing of tools The process of choosing Contract management Procurement Novel types of system
or algorithms. and procuring tools (Strategy, planning and of systems. (or data?) require new
successfully identifies management). understanding; wide
effective and safe range of tools available;
technologies. complex issues in
vendor relations
(Carter, 2019).

Data stewardship

Description of The nature and “biases” Cataloguing and resource Creating

provenance of data in data collections description; ontologies; and managing
sources and derivatives. are understood and metadata (Organising metadata.
considered within knowledge and information).
analysis.

Data governance. The risk and benefit Information governance Information Ensuring compliance.
from data are effectively and compliance. governance.
managed.

Curation and Long term benefits Digitisation, curation and Digital Technological
preservation of content are derived from data preservation (Records preservation. uncertainty; lack of
created by AI such as produced by AI. management and archiving). a digital preservation
derivatives. culture.

Data analysis, including Data is used for useful Data analytics (Using Domain experts, data
visualisation. analytic descriptions and and exploiting knowledge scientists and AI experts
predictions. and information). more likely to lead on
this but some level of
understanding useful.

*Traditional information role being extended.

A REPORT FOR CILIP 34

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Area of activity Impact PKSB Role extended* Challenges

Technical infrastructure
Infrastructure Effective, reliable, secure Information architecture Management IT team more obvious
building/ procurement, and safe technologies are (Organising knowledge and of information to manage.
e.g. storage for scale assembled for AI to work information), information content specific
and heterogeneity within. retrieval (Using and infrastructure.
of data. exploiting knowledge
and information), data
management (Knowledge
and information
management), ICT skills (IT
and communication).

Workflow design, Workable workflows Information architecture Procurement IT team more obvious
including reintegration are created. (Organising knowledge of systems. to manage.
of derived data into and information), data
discovery systems. management (Knowledge
and information
management), ICT skills
(IT and communication).

Support, training and promotion

Skills development Users are supported Teaching and training Training. Diverse ways machine
for users. to acquire the skills skills, supporting users learning used makes
they need to manage (Literacies and learning). this challenging.
data and use AI tools.
Users such as
researchers may not
expect to come to the
library for this training.

Curation and Long term benefits Digitisation, curation and Digital Technological
preservation of content are derived from data preservation (Records preservation. uncertainty; lack of
created by AI such as produced by AI. management and archiving) a digital preservation
derivatives. culture.

Marketing new tools Users are aware of the Strategic marketing Marketing of Finding bandwidth.
to users. changing technology (Customer focus, service new information
options available. design and marketing). services and
products.

Responsible use

Ethics and values – Professional values Access to knowledge, Access to The way familiar issues
ensuring that AI uses and ethics are respected intellectual freedom, knowledge. are manifested may
are ethical and in in every aspect of information skills and be shifting.
tune with human and technology use. information literacy
professional values.
Research ethics
(Research skills).

*Traditional information role being extended.

35 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

As section 3.3 established, what AI will This consists of:

look like in different contexts could be
quite different. The application could be
o Comfort and fluency with
computational systems
very specific to information service work,
e.g. use of AI on a special collection in a o Metacognitive skills in learning about
research library or an intranet in a law firm new computational resources
or data warehouse run by the information o Fluency in incremental tailoring
service. But it could also relate to wider and combining of applications for
organisational use of AI in knowledge work evolving needs
with impacts on information work. This o A sense of applications as ongoing
work would be led by knowledge engineers co-designed artefacts rather than
with their own, mostly technical skillset technological givens
(Anton et al., 2020). In the first case the
information service will need to have the
o A sense of the feasibility of potential
design options
vision and make the decisions. In other
cases this may be happening elsewhere, This articulates a positioning of an
but that is all the more reason for the information professional who is
information service to be involved in the comfortable with technologies, learns
project. It may be an individual information to use them quickly and plays roles in
professional seeking to get involved in an envisioning their future use, commissions
intelligent automation project. systems, customises systems for local
needs and generally bridges between
In many cases this work feels like an
users’ needs, organisational purposes, and
extension of some familiar information
the perspectives of developers. For many
professional roles, e.g. work around
information professionals computational
collections, copyright and metadata
sense may be useful, depending on how
management. As it is likely to be a
technical their role is. This goes beyond the
collaborative effort there is an emphasis
orientation towards IT skills described in
on multi-professional working. The final
the current PKSB, which focuses on specific
column surfaces issues such as how
information service technologies and skills
it is problematic to convert traditional
characteristic of a power user. It also aligns
competencies to work in this new area,
to JISC’s useful digital capabilities model
including because other professional
(JISC, 2019b). Certainly, how the profession
groups have more obvious expertise.
positions itself in relation to computing is
critical, though it was equally a common
5.2 Computational thinking, opinion among the experts consulted for
the report that there was no likelihood
fusion skills and soft skills or need for information professionals to
Many participants in the study thought become computer scientists.
that it would be useful if information Another perspective is offered by
professionals “got their hands dirty” with AI Daugherty and Wilson (2018). Given the
tools, exploring how to use them to better ongoing requirement for AI and robots
understand potential applications. AI tools to work together with humans, they see
such as for NLP and computer vision have a need for eight “fusion skills” to allow
become much easier to use out of the box organisations to permit computers and
in the last five years, so this is now possible. people to work together more effectively
Some thought a deeper understanding (see Appendix 3: Fusion skills). Their
was required. A classic articulation of this last skill of “relentlessly reimagining” is
is the notion of “computational thinking”, interesting in placing emphasis on the
which seeks to capture at an abstract need to reinvent processes to reflect
level what type of thinking is needed to transformed processes, not simply to
solve problems using computers (Barr automate old ones.
et al., 2011; Wing, 2006). Another useful
conceptualisation of how information While computational thinking or sense,
professionals might position themselves data science skills and even fusion skills
in relation to computing is “computational may become more important due to AI
sense” (Twidale & Nichols, 2006). and robots, there is also a plausible case
to say that the skills that will gain in value

A REPORT FOR CILIP 36

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

will be those which computers remain many tools available will be useful, be that
poor at. WEF (Gray, 2016; World Economic to support users or to apply them directly
Forum, 2020) suggest that the following are to their own professional work.
areas needed:
In terms of exploiting data science methods
o Complex problem solving for information service work, there will be
o Critical thinking strong competition for people with these
o Creativity competencies, so it may be hard to recruit
or even retain talent in this area (Markow
o People management et al., 2017).
o Coordinating with others
The Edison project (Demchenko et al.,
o Emotional intelligence 2017) explored skillsets in the broad
o Judgement and decision making data science area and drew a useful
o Service orientation distinction between data analytics (see
o Negotiation Appendix 4: Data Science Data Analysis
skills), data management skills (Appendix
o Cognitive flexibility 5: Data Science Data Management skills)
This listing reflects the ability to apply and a number of other skillsets, such as
intelligence in a broad way (in contrast engineering for data science. In reality, as
to the narrow AI achieved by computers) they recognize, roles might sit across these
and also to have the soft, human skills divides: data science is a spectrum (Burton
that computers are unlikely to develop in et al., 2017). Information professionals need
the near future. In this context the skills to be more “data savvy” in general and
many information professionals possess some information professionals will acquire
in collaboration, influencing and negotiating data analytics skills. However, it seems
seem to be very valuable. Arguably the reasonable to position most information
PKSB does not currently fully represent professionals as most likely to work more
some of these higher order cognitive skills; in the data management than data analysis
for example, complex problem solving, area, as defined by the Edison project.
critical thinking and creativity are not
As has already been established, the
presently identified as professional skills.
current wave of excitement around AI was
The information profession is so diverse preceded by a wave of interest in “big data”,
and positioned so differently that it is because AI requires large volumes of data.
hard to generalise about what skills will As has also been explained, the range of
be needed, but computational sense, data what is called data has expanded. In AI
science skills, fusion skills and these soft there is often training data, input data
skills suggest directions of travel. and derived data to manage. As a result,
a core aspect of most AI applications is
5.3 Data science and data data governance. Since data is closely
related to information, data governance
stewardship and management feels very much within
the scope of information professionals.
Use of AI tools to support data analysis has
They can certainly contribute a unique
created a new form of analyst role, the data
perspective in the context of multi-
scientist. Data science uses computational
stakeholder data governance (Wendehorst,
methods (such as those discussed in 4.1
2020). Articulating this clearly is important
above) to derive new knowledge from data.
to taking the opportunities presented by
Data scientists tend to have a combination
AI. Developments in some information
of statistical skills, computational skills and
sectors, such as around research data
domain knowledge. Some might sit more
management in the academic sector, give
at the analysis end, some more at the end
us some sense of what this might look
of business need, and some bridge or
like. A lot here may turn on how involved
translate between the two (Henke et al.,
information professionals are already in
2018; Open Data Institute, 2020a). Data
the data management in their organisation.
visualisation is also part of what they do.
In some cases the data is library collections
This might be an area where information
or know-how managed by an information
professionals need to reposition their skills
service. In other cases there may be other
base. Some understanding of the principles
stakeholders who currently manage data.
involved and hands-on experience of the

37 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

This recognition should be qualified o An emphasis in many, though not all

by remembering that there are other contexts, on the value of sharing data,
professional groups who would claim the even of open data, including based on
data governance realm, notably computer FAIR principles. The value of commitment
scientists, IT staff, data analysts and data to standards, unique identifiers and
scientists. One organisation of importance interoperability is highly relevant.
in this context is DAMA. Appendix 6 quotes
an overview of DAMA’s Data Management
o Ethics might be given more emphasis
than security.
Body of Knowledge (DMBoK2) (DAMA
International, 2012). The emphasis in DAMA o An emphasis on a lifecycle of
DMBok2 is on the low-level management of data, including a concern with long
data for immediate use, remaining rather term preservation.
detached from what the data is about
Thus, information professionals’ approach
or its provenance. In a number of ways
to data is somewhat different from that
information professionals might have a
of DAMA. This kind of appreciation is
rather different view on data governance
better articulated in the Edison project’s
that might include:
(Demchenko et al., 2017) definition of
o Emphasis on the provenance of data Data Science Data Management because
as created in a particular context and it mentions issues such as ethics and data
needing to be used in ways based on an curation (Appendix 5). Several participants
understanding of it having originated in for this report recognised that librarians
that particular context. have an instinctive understanding of
issues relating to the nature of data.
o Linked to this, a critical understanding To capture the unique information
of the limits of any data set: its “bias”. The
professional perspective on these matters
term “data” perhaps implies its “truthiness”
the use of the term ‘data stewardship’ is
that it is “a single, direct, objective
suggested. Currently the PKSB locates
representation of a measurable reality”
data management under Knowledge and
(Fiore-Gartland & Neff, 2015: 476). Actually,
Information Management, but states
this is not true of any form of data, but we
that by data we mean structured, usually
particularly know that texts do not have
numerical data. This needs to be extended
this quality. Individual texts are “biased” by
to reflect the societal re-evaluation of
who authored them, in the sense that they
data and to articulate the information
obviously represent views of an author or
profession’s unique perspective of
authors. Collections also inevitably have
data stewardship.
their “biases” that need to be understood
in interpreting them (Cordell, 2020; Padilla, The data librarianship literature explores
2019). There is a history of how they were the way that one branch of the information
created and selected, but when texts profession has adapted to work in the
become seen as data, this sense that bias area of data management and curation,
needs to be explained and managed may chiefly for research data, hence the other
be lost – just as there may be a danger of term used to define work in this area,
forgetting the bias, the inherent limits, of Research Data Management (Cox, Kennan,
the algorithms used by AI to analyse them et al., 2019; Federer, 2018; Semeler et al.,
(Mordell, 2019). 2019; Tammaro et al., 2019). As well as
capturing the technical skills required, the
o Emphasis on subject description of literature also addresses the generic skills
the content of data, on categorisation,
and personal characteristics needed by the
taxonomies, etc. Computers are
professional working in this area.
increasingly able to do this, but in many
contexts building on human concepts.

A REPORT FOR CILIP 38

Strengths and
vulnerabilities
of the profession
 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Strengths and vulnerabilities

of the profession

The fundamental strength of information contribute to identifying useful applications

professionals in taking up the opportunities of AI, though this might look very different
created by AI is that, as the previous in different contexts.
chapter above indicates, they align to roles
6. Many of the soft skills that information
they already play. The more similar to
professionals excel at are all the more
traditional roles new activities are, the more
relevant because computers are not likely
probable the competencies already exist;
to replicate them in the near future.
and the more likely that other stakeholders
immediately understand that there is a role 7. In terms of adapting to the new
for information professionals. technologies, in many sectors information
professionals are active in sharing good
practice across different organisations,
6.1 Strengths so are able to learn from each other,
Reinforcing this fundamental point we can collectively. The shared professional
reflect that: identity among information, knowledge
management and library professionals
1. The profession’s work in understanding is an asset, whereas other groups, such
how to navigate across complex ecologies
as data scientists, lack this coherent
of information and knowledge production
professional community.
and to critically evaluate content is all the
more important in a data and information Engagement with AI and robot applications
rich world. reflects a desire by information
professionals to ensure that we remain
2. Web and mobile search has been one seen as a modern, future-facing and
of the first areas to see applications of AI.
digital profession.
Thus, information professionals have a
profound, critical grasp of the underlying
issues from promoting user understanding 6.1 Vulnerabilities
of the value and limits of search engines
Nevertheless, there are areas where the
as an aspect of information literacy. The
profession can be seen as vulnerable:
same kind of issues are apparent in other
AI applications. 1. While aligned to many roles which
The information professionals already play,
strength of 3. Much of the latest development in AI there is a need to translate relevant skills
is around creating, using and especially
information to the new context of AI. This repositioning
describing textual content, as well as other
requires effort and experiment on
professionals forms of content, so closely relates to
behalf of organisations and individuals.
in taking information professionals’ work.
Organisations in which information
up the 4. Because data is critical to AI, our general professionals are embedded will not always
opportunities understanding of information governance see how they are relevant, especially if they
and management more generally is highly work in something called a library. This
created by AI
relevant, albeit sometimes it has to be returns us to the longstanding dilemma
is that they translated to be applied to a new context. about the naming of the profession: there
align to roles We have a broad understanding of this as is respect for the label of librarian, but little
they already data stewardship. public understanding of what it means.
There is a collective need to assert the value
play. 5. Because there is a strong emphasis
and relevance of the work of information
on information professional work as a
and knowledge management and library
service for users aligned to organisational
professionals. It is a call often heard, but is
purposes, we are in a good position to
nevertheless true.

40 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

2. As chapter 5 showed, there is a strong 8. Organisations need to find ways to

relation between roles in AI and existing experiment with AI and technologies
aspects of the PKSB. Yet there were that sometimes fail. Some information
clearly areas where the PKSB could be organisations such as some libraries could
strengthened to reflect a repositioning be thought of as geared to repeating The more
of the profession. routine tasks on a massive scale. Creating immediate
organisational spaces for innovation issue is CPD
3. As noted above (in section 2.2, 3.5,
within these structures requires different
4.7) professional values and ethics are for current
approaches (Cook & Van der Veer Martens,
highly relevant to the use of AI but may professionals
2019; Rowley, 2011).
need some re-articulation to address
> See recommendation 4. who will be
the challenges AI creates, particularly
around accountability, explainability 9. Information professionals are busy operating
and transparency. and resources are stretched. What they over the next
> See recommendation 1. already do is important. There are many decades
other new things to be learned, beyond
4. The profession’s positioning towards as AI in its
those covered in this report. It has to be
technology may also need to shift. Many many forms
understood for which organisations and
participants saw a need for information
individuals AI is to be identified as a priority. gradually
professionals to get their “hands dirty”
Individual professionals have to seek out becomes
trying out the technologies so that they
opportunities to try to get to grips with the
could develop a feel for their potential common
new technologies in real world applications,
applications for users. Computational practice.
and to share their experiences.
thinking or computational sense is not built
> See recommendations 7 & 8.
into the heart of the profession.
> See recommendation 7. 10. While knowledge sharing and collective
learning about professional practice is
5. Linked to this, data analysis skills,
strong within some sectors of information
especially those using computational
work, this is not the case in all, and it is
methods, in data science are at a premium;
weaker between sectors. AI is an agenda
it will be hard for information services to
across the profession and the conversation
recruit in these areas.
could also include adjacent professions
6. At the same time it is unlikely for most such as archivists and museum curators,
information professionals to get involved because in many ways foundational
in writing algorithms, writing code or doing distinctions between types of content
heavy duty data analysis. So while there are eroding.
is a pull in the direction of embedding > See recommendations 3, 6 & 8.
computational sense and data science as
core professional competencies, there is an
11. Learning providers, such as library
and information schools, who train
equal pull towards the development of the
new professionals need to give them
fusion skills to work alongside AI and the
a fundamental grounding in AI, although
soft skills that computers are unlikely to be
it is hard to squeeze this into already
able to reproduce in the foreseeable future.
packed curricula.
7. While information and knowledge > See recommendations 9, 10 & 11.
professionals have a potential role
to play in AI, there are other groups
12. Curricula need to be updated, but the
more immediate issue is CPD for current
who have a big part to play such as
professionals who will be operating over
IT specialists, data analysts and data
the next decades as AI in its many forms
scientists. In organisations where AI is a
gradually becomes common practice.
core competency it is likely that specialist
> See recommendation 12.
expertise will occupy most of the space
around AI. There are also others who
claim the data management space (e.g.
DAMA). CILIP’s constituency is very broad
and it is necessary to articulate a vision
that resonates in all the different contexts,
yet identifies information professionals’
unique contribution.
> See recommendation 1.

A REPORT FOR CILIP 41

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Recommendations

The technologies discussed in this report,

though not without their challenges,
For information services
present a major opportunity for social and libraries
good, as the profession learns more about
how to use them and supports users to Recommendation 4
engage productively and safely with them. Leaders in information services and
In responding to the opportunities and libraries need to create organisational
challenges there are a number of areas structures within which experiment is
of potential action for stakeholders. possible and within which individual
learning is supported and encouraged.

For CILIP Recommendation 5

Information services and libraries should
Recommendation 1 actively engage with AI to explore the
It would be productive if CILIP could potential benefit to users.
facilitate a collective articulation of how
the profession can contribute to AI, such Recommendation 6
as through the role in AI and data literacy, Where possible, information services
through the notion of data stewardship, should actively share their knowledge
and the relevance of its values and ethics across the wider profession.
to this new context. The current review of
the PKSB could strengthen its alignments
to the needs of AI and robots, especially
For individual information
in relation to computational sense, data professionals
science, data stewardship and the soft
skills needed to work effectively with AI Recommendation 7
and robots. Facilitating discussion within In the spirit of the professional obligation
the profession about how CILIP’s ethics to keep up to date, individuals need to
and value statements apply to the specific be inquisitive and willing to explore the
case of AI and robots would help in new technologies.
strengthening the profession’s stance.
Recommendation 8
Recommendation 2 In the spirit of the professional obligation
This would be supported if CILIP could to engage with colleagues, individuals
identify pathfinder organisations and need to share their understanding and
individuals who demonstrate how AI and promote their vision of the relevance of
robots can be introduced for the benefit the profession.
of users and organisations.

Recommendation 3
CILIP with other professional bodies
can foster knowledge sharing across the
profession, and with adjacent professions,
through events and curating relevant
open learning resources. CILIP special
interest groups offer a locus for developing
communities of practice to support learning
about the new technologies.

42 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

For educational institutions For other training providers

at all levels Recommendation 12
There is a need both for taster courses
Recommendation 9
and more in-depth training in relevant
Training for new entrants to the profession
AI applications.
needs to encompass an understanding
of AI and how it may be applied in
information contexts. For research
Recommendation 10 Recommendation 13
There could be a greater stress on More research is needed on the adoption
developing computational sense, data of AI, the organisational structures that
science, fusion skills and also on the types support it and on the impact of AI on
of soft skills in greater demand because information, knowledge management
computers cannot carry them out. There and library professional roles.
could be a greater stress placed on
data stewardship.

Recommendation 11
There is a need for hands-on experiences
with AI applications in practice.

A REPORT FOR CILIP 43

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

References

67 bricks. (2017). Content Enrichment : An essential Burton, M., Lyon, L., Erdmann, C., & Tijerina, B.
strategic capability for every publisher. www.67bricks. (2017). Shifting to Data Savvy: The Future of Data
com/content-enrichment-an-essential-strategic- Science In Libraries. d-scholarship.pitt.edu/33891.
capability-for-every-publisher. Callister, P. D. (2020). Law, artificial intelligence,
AIIM. (n.d.). Robotic Process Automation ( RPA ) for and natural language processing: A funny thing
Information Management Professionals: Is RPA Right happened on the way to my search results.
for My Business? info.aiim.org/is-rpa-right-for-my- Law Library Journal, 112(2), 161–212. ssrn.com/
business. abstract=3712306.
AIIM. (2019). 10 Things You Need to Know About Carter, D. (2018). How real is the impact of artificial
Process Services and RPA. info.aiim.org/10- intelligence? The business information survey
things-you-need-to-know-about-process- 2018. Business Information Review, 35(3), 99–115.
automation-and-rpa. doi.org/10.1177/0266382118790150.
AIIM Industry Watch. (2018). Leveraging Deep Carter, D. (2019). New technologies and new
Learning and Machine Learning Capabilities. data sources: The business information survey
info.aiim.org/leveraging-deep-learning-and- 2019. Business Information Review, 36(3), 95–105.
machine-learning-capabilities. doi.org/10.1177/0266382119871848.
AINow. (2019). 2019 report. ainowinstitute.org/ Carter, D. (2020). Regulation and ethics in
AI_Now_2019_Report.pdf. artificial intelligence and machine learning
Anderson, C. B., & Craiglow, H. A. (2017). Text technologies: Where are we now? Who is
mining in business libraries. Journal of Business and responsible? Can the information professional play
Finance Librarianship, 22(2), 149–165. doi.org/10.108 a role? Business Information Review, 37(2), 60–68.
0/08963568.2017.1285749. doi.org/10.1177/0266382120923962.
Anton, E., Behne, A., & Teuteberg, F. (2020). Caspers, M., Guibault, L., McNeice, K., Piperidis,
The humans behind artificial intelligence – an S., Pouli, K., Eskevich, M., & Gavrilidou, M. (2015).
operationalisation of AI competencies. Twenty-Eighth Reducing barriers and increasing uptake of text and
European Conference on Information Systems data mining for research Baseline report of policies
(ECIS2020), June. aisel.aisnet.org/ecis2020_rp/141. and barriers of TDM in Europe ( extended version
of D3 . 3). project.futuretdm.eu/wp-content/
Association of Research Libraries. (2019). Research uploads/2017/05/FutureTDM_D3.3-Baseline-Report-
Library Issues RLI 299: Ethics of Artificial Intelligence. of-Policies-and-Barriers-of-TDM-in-Europe.pdf.
publications.arl.org/rli299.
Centre for Data Ethics and Innovation. (2020).
Baker, J. J. (2018). 2018: A legal research odyssey: Review into bias in algorithmic decision-making.
Artificial intelligence as disruptor. Law Library assets.publishing.service.gov.uk/government/
Journal, 110(1), 5–30. doi.org/10.2139/ssrn.2978703. uploads/system/uploads/attachment_data/
Bakke, A. (2020). Everyday Googling: Results file/938850/Main_Report_-_CDEI_Review_into_Bias_
of an Observational Study and Applications for in_Algorithmic_Decision-Making.pdf.
Teaching Algorithmic Literacy. Computers and CILIP. (2018). Ethical Framework. www.cilip.org.uk/
Composition, 57, 102577. doi.org/10.1016/j. general/custom.asp?page=ethics.
compcom.2020.102577.
Cook, M., & Van der Veer Martens, B. (2019).
Barr, D., Harrison, J., & Conery, L. (2011). Managing Exploratory Units in Academic Libraries.
Computational Thinking: A Digital Age Skill for Journal of Library Administration, 59(6), 606–628.
Everyone. Learning and Leading with Technology, doi.org/10.1080/01930826.2019.1626647.
38(6), 20–23. eric.ed.gov/?id=EJ918910.
Coombs, C., Hislop, D., Taneva, S. K., & Barnard,
Beller, E., Clark, J., Tsafnat, G., Adams, C., Diehl, S. (2020). The strategic impacts of Intelligent
H., Lund, H., Ouzzani, M., Thayer, K., Thomas, J., Automation for knowledge and service work:
Turner, T., Xia, J., Robinson, K., Glasziou, P., Ahtirschi, An interdisciplinary review. Journal of Strategic
O., Christensen, R., Elliott, J., Graziosi, S., Kuiper, J., Information Systems, July 2017, 101600.
Moustgaard, R., … Wedel-Heinen, I. (2018). Making doi.org/10.1016/j.jsis.2020.101600.
progress with the automation of systematic reviews:
Principles of the International Collaboration for Cordell, R. (2020). Machine Learning + Libraries:
the Automation of Systematic Reviews (ICASR). A report on the state of the field. labs.loc.gov/static/
Systematic Reviews, 7(1), 1–7. doi.org/10.1186/ labs/work/reports/Cordell-LOC-ML-report.pdf?.
s13643-018-0740-7. Cox, A. M. (2021). Exploring the impact of Artificial
British Computing Society. (2020). AI and your job. Intelligence and robots on Higher Education
doi.org/10.1093/itnow/bwaa063. through literature-based design fictions.
International Journal of Educational Technology in
Brooke Robertshaw, M., & Asher, A. (2019). Higher Education. doi.org/doi.org/10.1186/s41239-
Unethical numbers? A meta-analysis of library 020-00237-8.
learning analytics studies. Library Trends, 68(1),
76–101. doi.org/10.1353/lib.2019.0031. Cox, A. M., Kennan, M. A., Lyon, L., Pinfield, S.,
& Sbaffi, L. (2019). Maturing research data services
Burrell, J. (2016). How the machine ‘thinks’: and the transformation of academic libraries.
Understanding opacity in machine learning Journal of Documentation, 75(6). doi.org/10.1108/JD-
algorithms. Big Data and Society, 3(1), 1–12. 12-2018-0211.
doi.org/10.1177/2053951715622512.
Cox, A. M., Pinfield, S., & Rutter, S. (2019). The
intelligent library. Library Hi Tech, 37(3), 418–435.
doi.org/doi.org/10.1108/LHT-08-2018-0105.

44 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

DAMA International. (2012). DAMA-DMBOK2 Glanville, J. (2019). Text mining for

Framework. In The Data Management Association. information specialists. In P. Levay & J. Craven
dama.org/sites/default/files/download/DAMA- (Eds.), Systematic Searching : Practical Ideas
DMBOK2-Framework-V2-20140317-FINAL.pdf. for Improving Results (pp. 147–170). Facet.
Daugherty, P. R., & Wilson, H. J. (2018). Human + doi.org/10.29085/9781783303755.008.
Machine: Reimagining work in the age of AI. Harvard Glanville, J., & Wood, H. (2018). Text Mining
Business Review Press. Opportunities: White Paper. www.cadth.ca/text-
Demchenko, Y., Belloum, A., & Wiktorski, T. (2017). mining-opportunities-white-paper.
EDISON Data Science Framework: Part 1. Data Science Gramatica, R., & Pickering, R. (2017). Start-up
Competence Framework (CF-DS) Release 2 (Issue July). story Yewno: an AI-driven path to a knowledge-
edison-project.eu/sites/edison-project.eu/files/ based future The launch of Yewno. UKSG Insights,
filefield_paths/edison_cf-ds-release2-v08_0.pdf. 30(2), 107–111. insights.uksg.org/articles/10.1629/
Eliot, B., Andrews, W., & Hare, J. (2020). A framework uksg.369.
for applying AI in the enterprise. Grames, E. M., Stillman, A. N., Tingley, M. W., &
Elliott, J. H., Turner, T., Clavisi, O., Thomas, J., Elphick, C. S. (2019). An automated approach to
Higgins, J. P. T., Mavergames, C., & Gruen, R. L. identifying search terms for systematic reviews
(2014). Living systematic reviews: An emerging using keyword co-occurrence networks. Methods
opportunity to narrow the evidence-practice gap. in Ecology and Evolution, 10(10), 1645–1654.
PLoS Medicine, 11(2), 1–6. doi.org/10.1371/journal. doi.org/10.1111/2041-210X.13268.
pmed.1001603. Gray, A. (2016). The 10 skills you need to thrive in
Equality Task Force. (2020). Mind the gap: How the Fourth Industrial Revolution. World Economic
to fill the equality and AI accountability gap in an Forum. www.weforum.org/agenda/2016/01/the-10-
automated world. www.ifow.org/publications/mind- skills-you-need-to-thrive-in-the-fourth-industrial-
the-gap-the-final-report-of-the-equality-task-force. revolution.
Faggella, D. (2019). Machine learning in Human Group, C. G. I. (2020). GIG AWARD 2020 : Foreign and
Resources – applications and trends. Emerj. emerj. commonwealth office’s knowledge and information
com/ai-sector-overviews/machine-learning-in- team. cilipgig.blogspot.com/2020/11/gig-award-
human-resources. 2020-foreign-and-commonwealth.html.
Faggella, D. (2020). AI in Law and Legal Practice – Guth, L. M., & Vander Meer, P. (2017). Telepresence
A Comprehensive View of 35 Current Applications. robotics in an academic library: A study of exposure
Emerj. emerj.com/ai-sector-overviews/ai-in-law- and adaptation among patrons and employees.
legal-practice-current-applications. Library Hi Tech, 35(3), 408–420. doi.org/10.1108/LHT-
03-2017-0059.
Federer, L. (2018). Defining data librarianship:
A survey of competencies, skills, and training. Henke, N., Levine, J., & McInerney, P. (2018).
Journal of the Medical Library Association, 106(3), Analytics translator: The new must-have role.
294–303. doi.org/10.5195/jmla.2018.306. Harvard Business Review.
Fernandez, P. (2016). “Through the looking Holmes, W., Bialik, M., & Fadel, C. (2019).
glass: envisioning new library technologies”: How Artificial intelligence in education. Center for
artificial intelligence will impact libraries. Library curriculum redesign.
Hi Tech News, 33(5), 5–8. doi.org/10.1108/LHTN- Hoy, M. B. (2016). Smart Buildings: An Introduction
05-2016-0024. to the Library of the Future. Medical Reference
Fiore-Gartland, B., & Neff, G. (2015). Services Quarterly, 35(3), 326–331. doi.org/10.1080/0
Communication, mediation, and the expectations 2763869.2016.1189787.
of data: Data valences across health and Huang, M. H., & Rust, R. T. (2020). A strategic
wellness communities. International Journal of framework for artificial intelligence in marketing.
Communication, 9, 1466–1484. ijoc.org/index.php/ Journal of the Academy of Marketing Science, 4.
ijoc/article/view/2830. doi.org/10.1007/s11747-020-00749-9.
Fjeld, J., Achten, N., Hilligoss, H., Nagy, A., & IFLA. (2020). IFLA Statement on Libraries and
Srikumar, M. (2020). Principled artificial intelligence: Artificial Intelligence. www.ifla.org/files/assets/
Mapping consensus in ethical and rights-based faife/ifla_statement_on_libraries_and_artificial_
approaches to principles for AI. cyber.harvard.edu/ intelligence.pdf.
publication/2020/principled-ai. JISC. (2019a). Guide to the intelligent campus: Using
Flannery, J. (2020). Using NLP to Generate MARC data to make smarter use of your university or college
Summary Fields for Notre Dame ’s Catholic estate. repository.jisc.ac.uk/6882/1/Intelligent_
Pamphlets. International Journal of Librarianship, Campus_Guide.pdf.
5(1), 20–35. doi.org/10.23974/ijol.2020.vol5.1.158. JISC. (2019b). Jisc digital capabilities framework: The
Floridi, L., & Cowls, J. (2019). A unified six elements defined. repository.jisc.ac.uk/7278/1/
framework of five principles for AI in society. BDCP-DC-Framework-Individual-6E-110319.pdf.
Harvard Data Science Review, 1(1), 1–15. Jobin, A., Ienca, M., & Vayena, E. (2019). The global
doi.org/10.1162/99608f92.8cd550d1. landscape of AI ethics guidelines. Nature Machine
French, A., & Villaespesa, E. (2019). AI, Visitor Intelligence, 1(9), 389–399. doi.org/10.1038/s42256-
Experience, and Museum Operations: A Closer 019-0088-2.
Look at the Possible. In Humanizing the Digital: Johnson, J. D. (2020). Theorizing Network Bias
Unproceedings from the MCN 2018 Conference and Teaching Mêtic Invention in Online Search.
(pp. 101–113). Computers and Composition, 56 (August 2018),
Frey, C. B., & Osborne, M. A. (2017). The future 102573. doi.org/10.1016/j.compcom.2020.102573.
of employment: How susceptible are jobs to Jones, K. M. L. (2019). “Just because you can doesn’t
computerisation? Technological Forecasting and mean you should”: Practitioner perceptions of
Social Change, 114, 254–280. doi.org/10.1016/j. learning analytics ethics. Portal, 19(3), 407–428.
techfore.2016.08.019. doi.org/10.1353/pla.2019.0025.
Geary, D. (2019). How to bring AI into your library.
Infotoday, September, 32–36.

A REPORT FOR CILIP 45

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Jones, K. M. L., Briney, K. A., Goben, A., Salo, D., McNeal, M., & Newyear, D. (2013). Introducing
Asher, A., & Perry, M. R. (2020). A comprehensive chatbots in libraries. Library Technology Reports,
primer to library learning analytics practices, 49(8), 5–10.
initiatives, and privacy issues. College and Research McNeice, K. (2017a). Skills and education: The need
Libraries, 81(3), 570–591. doi.org/10.5860/ for strategic thinking. FutureTDM Symposium.
crl.81.3.570.
McNeice, K. (2017b). Text and Data
Jones, K. M. L., & Salo, D. (2018). Learning Mining: The Changing Nature of Research.
analytics and the academic library: Professional FutureTDM Symposium.
ethics commitments at a crossroads. College and
Research Libraries, 79(3), 304–323. doi.org/10.5860/ Millard, T., Synnot, A., Elliott, J., & Turner, T.
crl.79.3.304. (2018). Results from the evaluation of the pilot living
systematic reviews. community.cochrane.org/sites/
Jonnalagadda, S. R., Goyal, P., & Huffman, M. D. default/files/uploads/inline-files/Transform/201905
(2015). Automating data extraction in systematic LSR_pilot_evaluation_report.pdf.
reviews: A systematic review. Systematic Reviews,
4(78). doi.org/10.1186/s13643-015-0066-7. Min-Allah, N., & Alrashed, S. (2020). Smart
campus—A sketch. Sustainable Cities and Society, 59.
Kay, D. (2018). Report on Artificial Intelligence doi.org/10.1016/j.scs.2020.102231.
technologies and implications for freedom of
expression and the information environment. Moore, M. T. (2017). Constructing a sentiment
www.ohchr.org/EN/Issues/FreedomOpinion/Pages/ analysis model for LibQUAL+ comments.
ReportGA73.aspx. Performance Measurement and Metrics, 18(1), 78–87.
doi.org/10.1108/PMM-07-2016-0031.
Kelly, J. (2016). THE TEXT AND DATA MINING
COPYRIGHT EXCEPTION: BENEFITS AND Mordell, D. (2019). Critical questions for archives as
IMPLICATIONS FOR UK HIGHER EDUCATION. (big) data. Archivaria, 87(5), 140–161. archivaria.ca/
www.jisc.ac.uk/guides/text-and-data-mining- index.php/archivaria/article/view/13673.
copyright-exception. Murphy, O., & Villaespesa, E. (2020). The Museums +
Kroski, E. (2020). Law librarianship in the age of AI. Ai Network. themuseumsainetwork.files.wordpress.
American Library Association. com/2020/02/20190317_museums-and-ai-toolkit_
rl_web.pdf.
Lewis, V., Spiro, L., Wang, X., & Cawthorne, J.
E. (2015). Building Expertise to Support Digital Nath, F. (2018). Library drone delivery programme:
Scholarship: A Global Perspective. In CLIR Publication A study. DESIDOC Journal of Library and Information
(Issue 168). www.clir.org/pubs/reports/pub168/ Technology, 38(5), 350–354. doi.org/10.14429/
pub168. djlit.38.5.12892.
Liau, Y. chin. (2019). Transforming library operation Neary, M. A., & Chen, S. X. (2017). Artificial
with robotics. August, 1–5. creativecommons.org/ intelligence: Legal research and law librarians.
licenses/by/4.0. AALL Spectrum 21, 5(6), 16–20. lawdigitalcommons.
bc.edu/lsfp/1073.
LIBER. (2020). Research Libraries: How You Can
Support Text and Data Mining. NHSX. (2019). Artificial Intelligence: How to get it
right. Putting policy into practice for safe data-driven
Long, D., & Magerko, B. (2020). What is innovation in health and care. www.nhsx.nhs.uk/
AI literacy? Competencies and design media/documents/NHSX_AI_report.pdf.
considerations. Conference on Human Factors
in Computing Systems - Proceedings, 1–16. O’Connor, A. M., Tsafnat, G., Gilbert, S. B., Thayer,
doi.org/10.1145/3313831.3376727. K. A., Shemilt, I., Thomas, J., Glasziou, P., & Wolfe,
M. S. (2019). Still moving toward automation of
Lowendahl, J.-M., & Calhoun Williams, K. the systematic review process: A summary of
(2018). 5 best practices for artificial intelligence discussions at the third meeting of the International
in higher education. Collaboration for Automation of Systematic
Ma, L., & Sun, B. (2020). Machine learning and AI Reviews (ICASR). Systematic Reviews, 8(1), 1–5.
in marketing – Connecting computing power to doi.org/10.1186/s13643-019-0975-y.
human insights. International Journal of Research Oakleaf, M. (2018). The problems and promise of
in Marketing, 37(3), 481–504. doi.org/10.1016/j. learning analytics for increasing and demonstrating
ijresmar.2020.04.005. library value and impact. Information and Learning
Markow, W., Braganza, S., Taska, B., Miller, S. M., Science, 119(1–2), 16–24. doi.org/10.1108/ILS-08-
& Hughes, D. (2017). The Quant Crunch: How the 2017-0080.
Demand for Data Science Skills is Disrupting the Job OECD. (2019). Artificial intelligence in society.
Market. www.burning-glass.com/research-project/ www.oecd-ilibrary.org/science-and-technology/
quant-crunch-data-science-job-market. artificial-intelligence-in-society_eedfee77-en.
Maxwell, D., Norton, H., & Wu, J. (2018). The Data Oliver, J. C., Kollen, C., Hickson, B., & Rios, F. (2019).
Science Opportunity: Crafting a Holistic Strategy. Data Science Support at the Academic Library.
Journal of Library Administration, 58(2), 111–127. Journal of Library Administration, 59(3), 241–257.
doi.org/10.1080/01930826.2017.1412704. doi.org/10.1080/01930826.2019.1583015.
Mckinsey Analytics. (2019). Global AI Survey: AI Open Data Institute. (2020a). Data Skills Framework.
proves its worth, but few scale impact. In McKinsey theodi.org/article/data-skills-framework.
&Company (Issue November). www.mckinsey.com/
featured-insights/artificial-intelligence/global-ai- Open Data Institute. (2020b). Getting data right:
survey-ai-proves-its-worth-but-few-scale-impact. Perspectives on the UK National Data Strategy 2020.
library1.nida.ac.th/termpaper6/sd/2554/19755.pdf.
McKinsey Global Institute. (2018). Skill Shift:
Automation and the Future of the Workforce Padilla, T. (2019). Responsible operations: Data
(Discussion Paper, May 2018). In McKinsey science, machine learning, and AI in libraries. In
&Company (Issue May). www.mckinsey.com/~/ OCLC. doi.org/10.25333/xk7z-9g97.
media/McKinsey/Featured Insights/Future of Pelz-Shapre, A., & Kompella, K. (2019). Practical
Organizations/Skill shift Automation and the future artificial intelligence: An enterprise playbook.
of the workforce/MGI-Skill-Shift-Automation-and- Deep Publishing.
future-of-the-workforce-May-2018.ashx.

46 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Prado, J. C., & Marzal, M. Á. (2013). Incorporating Tammaro, A. M., Matusiak, K. K., Sposito, F. A.,
data literacy into information literacy programs: & Casarosa, V. (2019). Data curator’s roles and
Core competencies and contents. Libri, 63(2), responsibilities: An international perspective. Libri,
123–134. doi.org/10.1515/libri-2013-0010. 69(2), 89–104. doi.org/10.1515/libri-2018-0090.
Ridge, M. (2019). Operationalising AI at a national Tella, A. (2020). Robots are coming to the libraries:
library. Museums + AI Network. www.slideshare.net/ Are librarians ready to accommodate them? Library
miaridge/operationalising-ai-at-a-national-library. Hi Tech News, 37(8), 13–17. doi.org/10.1108/LHTN-
Rowley, J. (2011). Innovation for survival: From 05-2020-0047.
cooperation to collaboration. In Advances in Thelwall, M. (2018). Artificial Intelligence,
Librarianship (Vol. 34). Emerald Group Publishing Automation and Peer Review. In Jisc. link.springer.
Ltd. doi.org/10.1108/S0065-2830(2011)0000034013. com/article/10.1007/BF02016664.
Sander, I. (2020). Critical big data literacy tools— Tsai, Y. S., & Gasevic, D. (2017). Learning analytics
Engaging citizens and promoting empowered in higher education - Challenges and policies: A
internet usage. Data & Policy, 2. doi.org/10.1017/ review of eight learning analytics policies. ACM
dap.2020.5. International Conference Proceeding Series, 233–242.
Schlegel, D., & Kraus, P. (2020). Robotic process doi.org/10.1145/3027385.3027400.
automation as an emerging career opportunity: Twidale, M. B., & Nichols, D. M. (2006).
An analysis of required qualifications and skills. Computational sense: The role of technology in the
13th Annual Conference of the EuroMed Academy education of digital librarians. researchcommons.
of Business, 1083–1097. publikationen.reutlingen- waikato.ac.nz/handle/10289/40.
university.de/frontdoor/deliver/index/docId/2815/ Vincze, J. (2017). Virtual reference librarians
file/2815.pdf. (Chatbots). Library Hi Tech News, 34(4), 5–8.
Semeler, A. R., Pinto, A. L., & Rozados, H. B. doi.org/10.1108/LHTN-03-2017-0016.
F. (2019). Data science in data librarianship: Vlachos, E., Faber Hansen, A., & Povl Holck, J.
Core competencies of a data librarian. Journal (2020). A robot in the library. In M. Rauterberg
of Librarianship and Information Science, 51(3), (Ed.), Culture and Computing (pp. 312-). Springer.
771–780. doi.org/10.1177/0961000617742465. doi.org/10.1007/978-3-030-50267-6.
Shih, W. (2020). Voice revolution. Library Wagstaff, K. L., & Liu, G. Z. (2018). Automated
Technology Reports, 56(4), 5–13. classification to improve the efficiency of
Smith, A. (2016). Big Data technology, evolving weeding library collections. Journal of Academic
knowledge skills and emerging roles. Legal Librarianship, 44(2), 238–247. doi.org/10.1016/j.
Information Management, 16(4), 219–224. doi.org/ acalib.2018.02.001.
doi.org/10.1017/S1472669616000499. Wang, M. (2013). Supporting the research process
Stanfill, M. H., & Marc, D. T. (2019). Health through expanded library data services. Program,
information management: Implications of artificial 47(3), 282–303. doi.org/10.1108/PROG-04-2012-
intelligence on healthcare data and information 0010.
management. Yearbook of Medical Informatics, 28(1), Wendehorst, C. (2020). Framework paper for GPAI
56–64. doi.org/10.1055/s-0039-1677913. work on data governance.
Stuart, D. (2020). Data science for information Wing, J. M. (2006). Computational thinking.
professionals. Facet Publishing. Communications of the ACM, 49(3), 33–35.
Syed, R., Suriadi, S., Adams, M., Bandara, W., doi.org/10.1145/1999747.1999811.
Leemans, S. J. J., Ouyang, C., ter Hofstede, A. H. M., World Economic Forum. (2020). The 10 skills you
van de Weerd, I., Wynn, M. T., & Reijers, H. A. (2020). need to thrive in the Fourth Industrial Revolution.
Robotic Process Automation: Contemporary themes www.weforum.org/agenda/2016/01/the-10-
and challenges. Computers in Industry, 115, 103162. skills-you-need-to-thrive-in-the-fourth-industrial-
doi.org/10.1016/j.compind.2019.103162. revolution.
Tam, W. (2020). The rise of the robots: Using robotic Zheng, W. (2019). How do library staff view librarian
process automation in bibliometrics. tv.theiet. robotics? Librarian staff’s ignored humanistic views
org/?videoid=14133. on the impact and threat of robotics adoption. IFLA
Tambe, P., Cappelli, P., & Yakubovich, V. (2019). WLIC 2019, 1–17. library.ifla.org/2751/1/s02-2019-
Artificial intelligence in human resources wang-en.pdf.
management: Challenges and a path forward.
California Management Review, 61(4), 15–42.
doi.org/10.1177/0008125619867910.

A REPORT FOR CILIP 47

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Appendixes

APPENDIX 1 10. Human Role in AI

Recognize that humans play an important
AI literacy, according role in programming, choosing models, and
fine-tuning AI systems.
to Long and Magerko (2020)
11. Data Literacy
1. Recognizing AI
Understand basic data literacy concepts such
Distinguish between technological artifacts as those outlined in Prado and Marzal (2013).
that use and do not use AI.
12. Learning from Data
2. Understanding Intelligence
Recognize that computers often learn from
Critically analyze and discuss features data (including one’s own data).
that make an entity “intelligent”, including
discussing differences between human, 13. Critically Interpreting Data
animal, and machine intelligence. Understand that data cannot be taken
at face-value and requires interpretation.
3. Interdisciplinarity
Describe how the training examples
Recognize that there are many ways to think provided in an initial dataset can affect
about and develop “intelligent” machines. the results of an algorithm.
Identify a variety of technologies that use
AI, including technology spanning cognitive 14. Action & Reaction
systems, robotics, and ML. Understand that some AI systems have
the ability to physically act on the world.
4. General vs. Narrow
This action can be directed by higher-level
Distinguish between general and narrow AI. reasoning (e.g. walking along a planned
5. AI’s Strengths & Weaknesses path) or it can be reactive (e.g. jumping
backwards to avoid a sensed obstacle).
Identify problem types that AI excels at
and problems that are more challenging 15. Sensors
for AI. Use this information to determine Understand what sensors are, recognize
when it is appropriate to use AI and when that computers perceive the world using
to leverage human skills. sensors, and identify sensors on a variety
6. Imagine Future AI of devices. Recognize that different sensors
support different types of representation
Imagine possible future applications of AI
and reasoning about the world.
and consider the effects of such applications
on the world. 16. Ethics
7. Representations Identify and describe different perspectives
on the key ethical issues surrounding AI
Understand what a knowledge
(i.e. privacy, employment,
representation is and describe some
examples of knowledge representations. misinformation, the singularity, ethical
decision making, diversity, bias,
8. Decision-Making transparency, accountability).
Recognize and describe examples of how
computers reason and make decisions. 17. Programmability
Understand that agents are programmable.
9. Machine Learning Steps
Understand the steps involved in machine Note that Data literacy is explicitly mentioned
learning and the practices and challenges and Prado and Marzal (2013) cited as a source
that each step entails. for this subset of AI literacy.

48 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

APPENDIX 2 Evaluating data

Data literacy, as defined

Competency: learners need to be able to
evaluate data critically
by Prado and Marzal (2013) Contents: Data evaluation criteria (including
authorship, method of obtaining and
1. Understanding data
analyzing data, comparability, inference
What is data? and data summaries).
Competency: learners need to know what is
meant by data and be aware of the various 4. Managing data
possible types of data. Data and metadata collection
and management
Contents: Data definition; Types of data
(depending on origin, format, usage license Competency: learners need to be aware
and so on). of the need to save the data selected or
generated and of descriptive or other data
Data in society associated therewith, for due identification,
Competency: learners need to be aware management and subsequent reuse.
of the role of data in society, how they are Contents: Metadata; Reference
generated and by whom, and their possible management tools; Databases; Data
applications, as well as the implications of management repositories: policies
their use. and practices.
Contents: Data producers and consumers;
5. Using data
Data lifecycle; Data applications: their
impact on science and society; Copyright Data handling
and licenses influencing data reuse. Competency: learners need to be able to
prepare data for analysis, analyze them in
2. Finding and/or obtaining data keeping with the results sought and know
Data sources how to use the necessary tools.
Competency: learners need to be aware Contents: Data conversion; Handling data
of the possible data sources, be able to analysis tools, both locally (Excel, R, SPSS,
evaluate them and select the ones most Stata or similar) and on-line.
relevant to an informational need or a
given problem. Producing elements for data synthesis
Contents: Data sources; Criteria for Competency: learners need to be able to
assessing data sources. synthesize and represent the results of
data analysis in ways suited to the nature
Obtaining data of the data, their purpose and the audience
Competency: learners need to be able targeted in the inquiry.
to detect when a given problem or need Contents: Choosing suitable data
cannot be (totally or partially) solved with representation methods (tables, graphs or
the existing data and, as appropriate, similar); Handling tools (built into analytical
undertake research to obtain new data. tools or stand-alone applications such as
Contents: Main research methods for Gapminder, Visual.ly or IBM’s Many Eyes).
obtaining original data.
Ethical use of data
3. Reading, interpreting and Competency: learners need to make ethical
evaluating data use of data, acknowledging the source
Reading and interpreting data when obtained or formulated by others,
and making sure that used methods
Competency: learners need to be aware
are deployed and results interpreted
of the various forms in which data can be
transparently and honestly.
presented (written, numerical or graphic),
and their respective conventions, and be Contents: What is the ethical use of data;
able to interpret them. How to cite data sources.
Contents: Ways to present and
represent data.

A REPORT FOR CILIP 49

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

APPENDIX 3 APPENDIX 4
Fusion Skills Edison project
(Daugherty & Wilson, 2018) (Demchenko et al., 2017):
Rehumanizing time Data science data analytics
The ability to increase the time available for DSDA01 Effectively use variety of data
distinctly human tasks like interpersonal analytics techniques, such as Machine
interactions, creativity, and decision making Learning (including supervised,
in a reimagined business process. unsupervised, semisupervised learning),
Data Mining, Prescriptive and Predictive
Responsible normalizing
Analytics, for complex data analysis through
The act of responsibly shaping the purpose
the whole data lifecycle.
and perception of human-machine
interaction as it relates to individuals, DSDA02 Apply designated quantitative
businesses, and society. techniques, including statistics, time series
analysis, optimization, and simulation to
Judgment integration deploy appropriate models for analysis
The judgment-based ability to decide and prediction.
a course of action when a machine is
DSDA03 Identify, extract, and pull together
uncertain about what to do.
available and pertinent heterogeneous
Intelligent interrogation data, including modern data sources
Knowing how best to ask questions such as social media data, open data,
of AI, across levels of abstraction, to get governmental data.
the insights we need. DSDA04 Understand and use different
performance and accuracy metrics
Bot-based empowerment for model validation in analytics
Working well with AI agents to projects, hypothesis testing, and
extend your capabilities, and create information retrieval.
superpowers in business processes and
professional careers. DSDA05 Develop required data analytics
for organizational tasks, integrate data
Holistic melding analytics and processing applications
The ability to develop robust mental into organization workflow and business
models of AI agents to improve business processes to enable agile decision making.
process outcomes.
DSDA06 Visualise results of data
Reciprocal apprenticing analysis, design dashboard and use
1. Performing tasks alongside AI agents storytelling methods.
so they can learn new skills;
2. On-the-job training for people
so they can work well within
AI-enhanced processes.

Relentless reimagining
The rigorous discipline of creating new
processes and business models from
scratch, rather than simply automating
old processes.

50 A REPORT FOR CILIP

RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

APPENDIX 5 APPENDIX 6
Edison project DAMA DMBok2
(Demchenko et al., 2017): (DAMA International, 2012)
Data science data management Data governance
DSDM Develop and implement data planning, oversight, and control over
management strategy for data collection, management of data and the use of data and
storage, preservation, and availability for data-related resources. While we understand
further processing. that governance covers ‘processes’,
not ‘things’, the common term is Data
DSDM01 Develop and implement data Governance, and so we will use this term.
management strategy for data collection,
storage, preservation, and availability for Data architecture
further processing. the overall structure of data and
DSDM02 Develop and implement relevant data-related resources as an integral part
data models, define metadata using of the enterprise architecture.
common standards and practices, for Data modeling & design
different data sources in variety of scientific
analysis, design, building, testing,
and industry domains.
and maintenance.
DSDM03 Integrate heterogeneous data
from multiple source and provide them Data storage & operations
for further analysis and use. structured physical data assets storage
deployment and management.
DSDM04 Maintain historical information
on data handling, including reference to Data security
published data and corresponding data ensuring privacy, confidentiality and
sources (data provenance). appropriate access.
DSDM05 Ensure data quality, accessibility, Data Integration & Interoperability –
interoperability, compliance to standards, acquisition, extraction, transformation,
and publication (data curation). movement, delivery, replication, federation,
DSDM06 Develop and manage/supervise virtualization and operational support.
policies on data protection, privacy, IPR
Documents & content
and ethical issues in data management.
storing, protecting, indexing, and enabling
access to data found in unstructured
sources (electronic files and physical
records), and making this data available
for integration and interoperability with
structured (database) data.

Reference & master data

managing shared data to reduce redundancy
and ensure better data quality through
standardized definition and use of data values.

Data warehousing & business intelligence

managing analytical data processing and
enabling access to decision support data
for reporting and analysis.

Metadata
collecting, categorizing, maintaining,
integrating, controlling, managing,
and delivering metadata.

Data quality
defining, monitoring, maintaining data
integrity, and improving data quality.

A REPORT FOR CILIP 51

 RESEARCH REPORT: THE IMPACT OF AI, MACHINE LEARNING, AUTOMATION AND ROBOTICS ON THE INFORMATION PROFESSIONS

Resources
In addition to the extensive references quoted
in the report, of particular value for information
professionals interested in this area, the following
introductory learning resources are recommended:
Stuart, D. (2020). Practical Data Science for
Information Professionals. Facet Publishing.
Foster project, www.fosteropenscience.eu.
The carpentries, carpentries.org.
AI4K12,
github.com/touretzkyds/ai4k12/wiki.
Alan Turning Institute, www.turing.ac.uk.
ADA Lovelace Institute,
www.adalovelaceinstitute.org.

Case study sources

The Health Service Executive’s LAMA (p 22)
Laura Rooney Ferris, Library Resources Manager;
Digital Knowledge Service, Health Service Executive
(HSE), National Health Library & Knowledge Service,
[email protected].

Roche Products Limited: LibraryAnn (p 22)

Arthur Robbins, Library, Information & Knowledge
Services Manager, Roche Products Ltd,
[email protected].

Create Crates (p 25)

Jonathan Clayton, Library Officer, Wakefield Libraries,
[email protected].

52 A REPORT FOR CILIP

CILIP: the Chartered Institute of Library
and Information Professionals
7 Ridgmount Street
London WC1E 7AE
cilip.org.uk

Registered charity no: 313014