Understanding Local Government Responsible AI Strategy 1729374120

Cities 155 (2024) 105502
Contents lists available at ScienceDirect
Cities
journal homepage: www.elsevier.com/locate/cities
Understanding local government responsible AI strategy: An international

municipal policy document analysis
Anne David a , Tan Yigitcanlar a,*, Kevin Desouza b, Rita Yi Man Li c , Pauline Hope Cheong d ,
Rashid Mehmood e , Juan Corchado f
a
School of Architecture and Built Environment, Queensland University of Technology, 2 George Street, Brisbane, QLD 4000, Australia
b
School of Management, Queensland University of Technology, 2 George Street, Brisbane, QLD 4000, Australia
c
Department of Economics and Finance, Hong Kong Shue Yan University, 10 Wai Tsui Cres, North Point, Hong Kong, China
d
Hugh Downs School of Human Communication, Arizona State University, Phoenix, AZ 85287, USA
e
Faculty of Computer and Information Systems, Islamic University of Madinah, Madinah 42351, Saudi Arabia
f
Department of Computer Science and Automation, University of Salamanca, 37007 Salamanca, Spain
A R T I C L E I N F O A B S T R A C T
Keywords: The burgeoning capabilities of artificial intelligence (AI) have prompted numerous local governments worldwide
Artificial intelligence to consider its integration into their operations. Nevertheless, instances of notable AI failures have heightened
Responsible innovation and technology ethical concerns, emphasising the imperative for local governments to approach the adoption of AI technologies
Responsible AI
in a responsible manner. While local government AI guidelines endeavour to incorporate characteristics of
AI policy
AI strategy
responsible innovation and technology (RIT), it remains essential to assess the extent to which these charac-
Local government teristics have been integrated into policy guidelines to facilitate more effective AI governance in the future. This
study closely examines local government policy documents (n = 26) through the lens of RIT, employing directed
content analysis with thematic data analysis software. The results reveal that: (a) Not all RIT characteristics have
been given equal consideration in these policy documents; (b) Participatory and deliberate considerations were
the most frequently mentioned responsible AI characteristics in policy documents; (c) Adaptable, explainable,
sustainable, and accountable considerations were the least present responsible AI characteristics in policy doc-
uments; (d) Many of the considerations overlapped with each other as local governments were at the early stages
of identifying them. Furthermore, the paper summarised strategies aimed at assisting local authorities in iden-
tifying their strengths and weaknesses in responsible AI characteristics, thereby facilitating their transformation
into governing entities with responsible AI practices. The study informs local government policymakers, prac-
titioners, and researchers on the critical aspects of responsible AI policymaking.
1. Introduction due to greater surveillance (Nabavi & Browne, 2023). In 2017, the
Rotterdam local government deployed AI to generate a risk score for
Although advanced technology is crucial for local governments to everyone to reduce the possibility of fraud; females, young adults, par-
deliver improved services and amenities, traditionally their adoption of ents, and people who did not speak Dutch language were often classified
these innovations has lagged behind other sectors (Yigitcanlar, 2006). as high risk and flagged for investigation (Constantras et al., 2023).
However, in recent years, there has been a notable surge in the uti- These kinds of challenges have sparked a broader discussion
lisation of AI within local government service provision (Jeffares & regarding the role and responsibility of local government while adopting
Jeffares, 2021). While AI presents opportunities for effective service AI technologies (Atkinson et al., 2017). With the growing integration of
delivery, it also engenders significant societal challenges. These chal- AI technologies into local government services, ranging from adminis-
lenges undermine social norms such as justice, fairness, safety, and trative tasks to citizen engagement platforms, it becomes imperative for
accountability, and worsens other social issues, such as loss of privacy local authorities to focus on strategies for promoting AI that benefits
* Corresponding author.
E-mail addresses: [email protected] (A. David), [email protected] (T. Yigitcanlar), [email protected] (K. Desouza), ymli@
hksyu.edu (R.Y.M. Li), [email protected] (P.H. Cheong), [email protected] (R. Mehmood), [email protected] (J. Corchado).
https://doi.org/10.1016/j.cities.2024.105502
Received 1 April 2024; Received in revised form 13 August 2024; Accepted 5 October 2024
0264-2751/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. David et al. Cities 155 (2024) 105502
rather than harms the community (Mikalef et al., 2022). 2. Literature background
Considering growing concerns about the negative impacts of tech-
nological advancements, the discourse surrounding ‘Responsible inno- The growing use of AI technology prompts concerns about its effects
vation and technology’ (RIT) has gained significant traction recently (Li, on society (Gianni et al., 2022) and necessitates following RIT charac-
Yigitcanlar, Browne, & Nili, 2023). This conversation underscores the teristics while implementing AI technologies (European Commission,
importance of ensuring that modern innovation practices yield respon- 2021). The term ‘responsible innovation’ was prominently featured in
sible technological outcomes that can be effectively integrated into so- the European Union’s Horizon 2020 research and innovation framework
ciety, thereby alleviating potential adverse effects associated with program, which ran from 2014 to 2020 (Bogner & Torgersen, 2018;
contemporary technological advancement (Bourban & Rochel, 2021). Landeweerd et al., 2015). Since then, responsible innovation charac-
Academia has become increasingly engaged in researching and teristics have been further refined and expanded upon in various aca-
developing frameworks for responsible innovation, emphasising the demic publications, policy documents, and international initiatives (Li,
importance of ethical considerations, sustainability, and inclusivity in Yigitcanlar, Nili, & Browne, 2023). Today, responsible innovation is
technological advancements (Li, Yigitcanlar, Browne, & Nili, 2023; recognised as a guiding framework for ensuring that technological ad-
Simon et al., 2019; Wang et al., 2022). Likewise, the private sector has vancements align with societal values, address pressing challenges, and
begun to prioritise responsible practices in product development, rec- promote sustainable development (Lubberink et al., 2017; Silva et al.,
ognising the value of building trust with consumers and mitigating po- 2019).
tential risks associated with emerging technologies (Imaz & Eizagirre, The characteristics of RIT encompass the use of AI technology in a
2020; Li, Yigitcanlar, Nili, & Browne, 2023). For example, leading transparent and accountable manner and a transparent and accountable
technology companies such as Amazon, Apple, Samsung, Google, manner that is consistent with user desires, organisational values, and
Microsoft, Sony, IBM, and the like have put forth a set of AI-related societal norms and laws (Mikalef et al., 2022). According to Yigitcanlar,
guiding principles (De Laat, 2021; Delecraz et al., 2022; Mikalef et al., Corchado, et al. (2021), RIT has a positive impact on our societies as it
2022) aligning to the RIT concept given their concerns about reputation prompts us to think about building desired futures as community actors
and long-term sustainability. and innovators become responsible actors and innovators through a
Although local governments attempt to integrate RIT characteristics transparent and collaborative approach (Farida et al., 2023). Concur-
into their AI policy documents, this area remains relatively underex- rently, numerous researchers have endeavoured to formulate RIT
plored. As key facilitators of community development, local govern- characteristics through the examination of AI guidelines.
ments are central to this transformative discussion (Yigitcanlar et al., For example, Jobin et al. (2019) studied 84 ethical guidelines. The
2022). They recognise that adopting a responsible approach to innova- authors outlined a set of 11 ethical characteristics to guide the devel-
tion is not only beneficial for their communities but also crucial for opment and use of technology. These include: (a) Beneficence; (b)
maintaining public trust and ensuring sustainable growth in an ever- Dignity; (c) Freedom and autonomy; (d) Justice and fairness; (e) Non-
evolving landscape (Stilgoe et al., 2020). With the community desiring maleficence (avoiding harm); (f) Privacy; (g) Responsibility; (h) Trans-
programs and services that adhere to ethical standards, the RIT concept parency; (i) Trust; (j) Solidarity; (k) Sustainability. Ryan and Stahl
has become a pivotal driver in shaping local governments’ future tech- (2020) elaborated on Jobin et al.’s (2019) responsible AI characteristics
nological development strategies (Scherer & Voegtlin, 2020). Therefore, by using 91 ethical guidelines. Simultaneously, researchers have
further investigation and reviews are necessary to capture the growing developed sector-specific responsible AI characteristics. For example,
knowledge of RIT from the perspective of local governments. This will Roberson et al. (2022) identified four dimensions of responsible inno-
help bridge the research gap and make the concept more responsive and vation in defence: (a) Anticipation; (b) Reflexivity; (c) Inclusion; (d)
adaptable to local government settings. Responsiveness.
In addressing this gap, the paper at hand conducted a comprehensive Moreover, Buhmann and Fieseler (2021) suggested three main di-
policy review to investigate insights into RIT from 26 local governments’ mensions of responsible AI: (a) The obligation to minimise harm through
AI policy documents. The methodological approach employs directed risk management in order to minimise adverse effects; (b) Responsibility
content analysis with thematic data analysis software. The study’s pri- to do good refers to better living conditions—e.g., by the sustainable
mary aim is to elucidate the level of consideration given to RIT char- development goals; (c) Governance responsibility, which is the obliga-
acteristics within local government. To achieve this, the paper poses the tion to establish and uphold international governance frameworks that
research question ‘How do local governments incorporate RIT charac- can support the first two obligations. Mikalef et al. (2022) put forward
teristics to adopt AI technology responsibly?’ The findings from this eight responsible AI dimensions: (a) Fairness; (b) Transparency; (c)
analysis offer a nuanced understanding of how local governments Accountability; (d) Robustness and safety; (e) Data governance; (f) Laws
interpret and implement the RIT concept, highlighting variations and and regulations; (g) Human oversight; (h) Societal and environmental
commonalities across different local governments. Additionally, the well-being. Likewise, Yigitcanlar, Mehmood, and Corchado (2021)
study provides an in-depth discussion of the strategies employed by local identified the critical characteristics of responsible AI in smart cities and
governments to uphold RIT principles in their AI initiatives. This dis- urban futures as: (a) Efficiency; (b) Sustainability; (c) Equitability.
cussion is intended to offer practical guidance for policymakers, aca- Nonetheless, the most comprehensive elaboration of responsible AI was
demics, and community stakeholders, advance the development of more put forward by Li, Yigitcanlar, Browne, and Nili (2023). The author
responsible and socially beneficial technological outcomes. thoroughly examined the literature and proposed a conceptual frame-
Accordingly, the paper’s structure includes an introduction to the work of responsible innovation and technology (RIT)—that also serves
research gap and the importance of RIT in local government AI adop- as a conceptual framework of responsible AI (Table 1) and was adopted
tion. The methodology section details the use of directed content anal- in this study.
ysis and thematic data analysis software to examine policy documents
from 26 local governments. Findings are presented to show how RIT 3. Research design
principles are incorporated, followed by a discussion on the strategies
employed to uphold these principles. The paper concludes with rec- 3.1. Data
ommendations for policymakers, academics, and community stake-
holders, along with reflections on the implications for future research This study tackles the research question: How do local governments
and practice in advancing responsible technological outcomes. incorporate RIT characteristics to adopt AI technology responsibly? A
local government refers to the administrative authority and governing
body responsible for managing the affairs of a specific geographical
2
Table 1 Table 1 (continued )

Characteristics of responsible AI, derived from Li, Yigitcanlar, Browne, and Nili Characteristic Description Sub- Objective
(2023). characteristic
Characteristic Description Sub- Objective Inclusive Incorporating diverse
characteristic cultures, knowledge,
Acceptable “Publicly acceptable, Ethical Uphold human rights and values is essential
ethically and freedoms, for overcoming
unproblematic, and including dignity and technological and
harmless, including privacy. Ensure that ideological lock-ins
being devoid of innovations and and improving the
prejudice and technologies do not responsiveness of
deception, oriented to reinforce social technological
supplying equitable structures that cause trajectories to societal
products and people to be requirements.
promoting fair oppressed. This can be Aligned “Deliberate in Deliberate Carefully foresee and
technology use for achieved by decision-making evaluate the potential
well-being and the encouraging practises, aligned with and opportunities and
common good.” autonomy and moral social desirability, use thoughtful
acceptability, to avoid committed to decision-making
unethical upholding human techniques to reduce
consequences. values, devoted to any adverse impacts
Equitable Eliminate systematic creating meaningful, on life on earth that
stereotyping to lessen beneficial, and are already occurring
the risks and effects of sustainable solutions or may arise in the
increased and to address the future.
sustained disparities associated difficulties Meaningful Improve the
between people and and enhance the alignment of
groups and promote quality of life.” innovative
more widespread, technology, social
democratic, and norms, and people’s
equitable innovation. wants and preferences
Harmless Ensure that products to produce expected
do not negatively and meaningful
impact the outcomes, such as
environment or addressing pressing
human health issues or societal
(including mental and requirements and
physical health) while enhancing human
minimising or well-being.
eliminating the Sustainable The environment can
negative be part of innovation,
consequences of enabling the
technology use responsible handling
through suitable of resources across the
safeguards, such as whole innovation
intrinsically safe cycle for broad
design, meaningful sustainability
human control outcomes while
methods, etc. preventing negative
Accessible “Extend the definition Adaptable Produce compelling irreversible impacts
of accessibility to and trustworthy on the earth.
create socially products that can be Trustworthy “Improve information Explainable Making decisions or
inclusive, adapted to current transparency and actions clear and
technologically technologies and technical security being able to defend
adaptive, and challenging while planning, them to gain the
economically feasible operational producing, public’s trust and
goods committed to situations. Ensure implementing, and sufficient
distributing the these products are operating procedures, comprehension are
advantages of simple to learn, use, committed to necessary for
digitisation across and maintain so that a providing successful
societies and cities broader range of users comprehensible implementation and
without obstacles.” may use them. justifications for management. This
Affordable It is preferable to choices made to bolster goes for both
assure the supply of the public’s professionals and the
high-value outputs knowledge, public.
while maintaining confidence, and trust.” Secure Assure the physical
economic viability so and digital safety and
that resources may be security of innovation
leveraged in ways that in society to reduce
do not negatively the potential harm or
impact consumers. adverse effects while
This improves the attempting to increase
conditions for more public confidence in
extensive new technologies.
implementation Transparent To increase the
scenarios. general public’s
understanding of
innovation, including
(continued on next page)
3
Table 1 (continued ) (Fig. 1).

Characteristic Description Sub- Objective In the first phase, the Google search engine identification stage
characteristic included developing the keywords for policy documents searching and
its possible
inclusion and exclusion criteria. The Boolean search string was as fol-
advantages, lows: (“artificial intelligence” OR “AI”) AND (“policy*” OR “strategy*”
challenges, and OR “action plan” OR “framework” OR “report” OR “agenda” OR
effects; information “regulation” OR “roadmap”) AND (“local government” OR “municipal-
and data on the
ity” OR “council” OR “city”). The results of the initial review returned
design, generation,
execution, and 1385 publications. To successfully screen documents, the second
operation procedures screening phase, our study established inclusion and exclusion criteria
as well as future (Table 2).
planning should be Table 2 indicates that journal articles, book chapters, and conference
freely provided.
Well- “Well-governed by the Accountable The goal is to infuse
proceedings which were not in English and not national or regional-level
governed larger stakeholder the values of articles were initially removed, and records were reduced to 104. The
groups, committed to responsibility in the documents related to smart cities and digital strategies were excluded
maintaining its innovation process by from this study due to their lack of specificity regarding AI technology,
reliability and identifying, tracing,
which is essential for achieving the study’s aim.
accountability to and holding
preserve public accountable the Accordingly, 10 policy documents—i.e., Nordic Municipalities, Mil-
support and trust, entities with legal ton Keynes Council, Madrid City Council, Shenzhen Municipality, City of
which results in obligations for their Vienna, District council of Streaky Bay, Shanghai Municipality, City of
increased acceptance decisions and actions Tempe, Wroclaw Metropolitan Area, and New York City Area—were
and continued or other failures.
implementation.” Participatory Maintaining public
considered for the analysis. These documents mentioned other local
support and trust, government names as examples and references. These city or local
growing stakeholder government names were then added to the following search string, e.g.,
groups, raising (“artificial intelligence” OR AI) AND (“policy*” OR “strategy*” OR
engagement levels,
“action plan” OR “framework” OR “report” OR “agenda” OR “regula-
and fostering
relationships that are tion” OR “roadmap”) AND (“Barcelona”). The inclusion and exclusion
mutually responsive criteria (Table 2) were applied in the screening phase. Through this
are crucial for process, 10 policy documents from different cities have been found—i.
successfully e., Barcelona City Council, City of Amsterdam, City of Helsinki, Warr-
integrating
nambool city, City of Lebanon, Singapore City State, London Council,
innovation into the
complex and Kelowna City, Mexico City and Dubai City were identified for the
changing societal analysis. Singapore’s national AI strategy documents have been incor-
context. porated into this study as a city-state. A state is an independent nation
Regulated It is operated by laws,
inside a single city’s bounds (Longley, 2019).
fundamental
freedoms, and rights In the second phase, the Baidu search engine identification stage
as much as is used the Chinese keywords of 人工智能 (AI) 地方政府 (local govern-
reasonably possible, ment) 政策 (policy) or 支持 (support) 政策(policy) or 相关 (relevant) 政
which aids in 策 (policy) to find the policy documents. These strings pointed to the
addressing
official Chinese municipal websites where the AI policy was posted.
complicated issues
and achieving There were 48 websites found in total, all written in Chinese. Then, these
consensus that documents were translated into English by using Google Translate. A
encourages and study by the University of California, Los Angeles at the Medical Centre
promotes innovation
in 2021 found that Google Translator successfully translated 82.5 % of
and enhances the
broader consequences
the texts while maintaining the overall sense (Taira et al., 2021). Next,
for society. the duplications which were, irrelevant to the study aims and study area,
were removed. In addition, two Chinese-speaking researchers checked
the translation, and assessed the accuracy level to be 90 %. After these
area, such as a city, town, municipality, county, or region (Masuda et al., filters, six documents, i.e., Fuzhou Municipality, Nanjing Municipality,
2022). Local governments operate at a level of administration below the Zhengzhou Municipality, Changsha Municipality, Dalian Municipality,
national or federal government and typically have jurisdiction over and Shenyang Municipality, were considered for the analysis.
matters that directly affect the local community (Butt et al., 2021; Following the completion of all three rounds, in total 26 policy
Phillimore et al., 2022). Accordingly, the dataset was compiled by documents underwent analysis and were documented in Table 3. These
examining local government AI policy documents published by gov- documents encompassed 34 local governments, with the majority being
ernment authorities or in collaboration with private institutions. The Chinese local governments (n = 8). Published between 2018 and 2024,
search task was conducted in May 2023. As the methodological the average length of these documents is 25 pages, with individual
approach of the systematic review, the preferred reporting items for lengths varying from three to 69 pages.
systematic reviews and meta-analyses (PRISMA) protocol was adopted
(Regona et al., 2022). 3.2. Methodology
China and the US are among the leading nations driving global ad-
vancements in AI (AlShebli et al., 2023; Hine & Floridi, 2022). Content analysis is a method that systematically analyses textual
Accordingly, the selection process involved two phases and three stages: data to enable researchers to make generalisations based on the data
Stage 1 (identification), Stage 2 (screening), and Stage 3 (inclusion). categories that interest them (Haggarty, 1996, p.99). Besides, content
During the first phase, the Google search engine was used to locate analysis allows to “identify meaning by examining patterns across
English-language documents, while the Baidu search engine was various artifacts” (Fatima et al., 2020, p.180). Content analysis has been
employed in the second phase to find Chinese-language documents highly incorporated in research fields such as health (Abdelmutti &
4
Fig. 1. PRISMA selection procedure of the strategy documents.
4. Results and analysis

Table 2
Inclusion and exclusion criteria.
4.1. General observations
Inclusion Exclusion
▪ AI policy documents published by ▪ Smart city or digital strategy The policy documents explicitly state the reasons behind their
local authorities or published by documents development, which centre on protecting individual rights in all AI-
research institutes, universities, related activities undertaken within the city. As the Tempe City docu-
non-government organisations, or
private institutions
ment outlines:
▪ Relevant to the study aim or ▪ Academic journal articles,
“The City of Tempe established the Ethical Artificial Intelligence (AI)
question books, chapters, conference
proceeding Policy to affirm Tempe’s commitment to responsible and ethical use of AI
▪ In the English language ▪ Not in the English language through the principles that ensure transparency, fairness, accountability,
▪ Full-text report ▪ Not relevant to the study aim and the protection of individual rights in all AI-related activities con-
▪ Available online ▪ Not at the study level: ducted by the City of Tempe. Being intentional in our adoption and use of
Case study level: local government national or regional
AI technologies will drive innovation, support increased efficiencies in
▪
or city level
operations and improved experiences for community engagement. This
policy outlines the principles, guidelines, and procedures governing the
Hoffman-Goetz, 2009), education (Cardno, 2019), journalism (Hurwitz responsible and ethical use of AI technologies within Tempe” (Tempe
et al., 2016), and technology (Gurr & Metag, 2022). Moreover, Hsieh City, 2023, p.1).
and Shannon (2005) identified three distinct approaches to content
Meanwhile, certain policy documents specify the areas to which the
analysis: conventional, directed, and summative (Table 4).
policy guidelines will apply. For instance, the Dubai City policy guide-
Accordingly, this study follows the ‘directed content analysis’
lines will exclusively pertain to AI systems involved in making or
method, where the nodes are defined prior to and throughout the data
informing ‘significant decisions’.
analysis. To address the research question, this study adopted the nodes
and sub-nodes from Li, Yigitcanlar, Browne, and Nili’s (2023) respon- “This document is applicable only to those AI systems which make or
sible AI characteristics framework (Table 5). The content analysis used inform ‘significant decisions’ (An individually significant decision which
NVivo 1.6.1 software to analyse the qualitative and quantitative AI is deemed to either have a very large impact on an individual or to have
policy documents. especially high stakes, be especially sensitive, have the potential to cause
high loss or damage, is societally significant, or sets important precedent)–
5
Table 3
Salient characteristics of the strategy documents.
No City or local government Policy document Publisher Sector* Country Year
1 Barcelona City Council Government measures for municipal Barcelona City Council G Spain 2021
algorithms and data strategy for ethical
promotion of AI
2 New York City IT Policy: Acceptable use of Artificial NYS Office of Information G US 2024
intelligence technologies Technology Services
3 Denmark - Gladsaxe Municipality, Copenhagen Nordic municipalities work with AI Nordic Council of Ministers G Nordic 2019
Municipality, Finland - Espoo Municipality, countries
Helsinki Municipality, Iceland - Kópavogur
Municipality, Reykjavík Municipality, Norway -
Oslo Municipality, Trondheim Municipality,
Sweden - Helsingborg Municipality, Stockholm
Municipality
4 Milton Keynes Council Milton Keynes and AI report Milton Keynes Council G UK 2020
5 Madrid City Council Municipal program for AI and robotisation Madrid City Council G Spain 2021
of the City Council of Madrid
6 Shenzhen Municipal Government Shenzhen New generation AI development Shenzhen Municipal People’s G China 2019
action plan Government
7 City of Amsterdam Agenda AI: Amsterdam intelligence Local Authority of Amsterdam G Netherlands 2020
8 City of Helsinki Public AI register The City of Helsinki and Saidot GP Finland 2020
Ltd
9 City of Vienna AI Strategy City of Vienna G Austria 2019
10 Warrnambool City Council Use of Artificial Intelligence Policy 2024 Warrnambool City Council G Australia 2024
11 City of Lebanon Administrative policies & procedures – use City of Lebanon N⋅H G US 2022
of artificial intelligence
12 District Council of Streaky Bay Artificial Intelligence policy District Council of Streaky Bay G Australia 2024
13 Singapore City State National Artificial intelligence strategy Smart Nation Singapore G Singapore 2019
14 London City Council An Emerging technology charter for Mayor of London Assembly G UK 2021
London
15 Shanghai Municipality Regulations for the promotion of the Shanghai Municipal People’s G China 2022
development of the AI industry in Congress
Shanghai Municipality
16 Kelowna City Intelligent city strategy City of Kelowna G Canada 2020
17 Mexico City Towards an AI strategy in Mexico: British Embassy Mexico City, GP Mexico 2018
harnessing the AI revolution Oxford insights, C MINDS
18 Dubai City Smart Dubai – AI ethics principles & Smart Dubai G UAE 2018
guidelines
19 City of Tempe Ethical Artificial Intelligence policy City of Tempe G US 2023
20 Wroclaw Metropolitan Area AI for a good life Wroctaw 2030, Education Center GP Poland 2021
in Wrocław, Wrocław
Agglomeration Development
Agency, PWC
21 Fuzhou Municipality Implementation plan for promoting the General Office of Fuzhou G China 2020
accelerated development of new- Municipal People’s Government
generation AI
22 Nanjing Municipality Nanjing’s 14th five-year high-tech zone General Office of Nanjing G China 2022
development plan Municipal People’s Government
23 Zhengzhou Municipality Zhengzhou new generation AI industry Zhengzhou Municipality People’s G China 2021
development plan (2021–2025) Government
24 Changsha Municipality Opinions of the General Office of the General Office of Changsha G China 2019
Changsha Municipal People’s Government Municipal People’s Government
on further promoting the development of
the AI industry
25 Dalian Municipality Development plan for a new generation of Dalian Municipal People’s G China 2019
AI Government
26 Shenyang Municipality Shenyang’s new generation AI Shenyang Municipal People’s G China 2018
development plan (2018–2030) Government
*
Sectors: government (G), government and private (GP).
Table 4 Table 5
Coding difference between the three approaches, derived from Hsieh and Responsible AI’s nodes and sub-nodes characteristics.
Shannon (2005). Node Sub-nodes
Content analysis Study begins Timing of defining nodes or keywords
Acceptable Ethical, Equitable, Harmless
approaches with
Accessible Adaptable, Affordable, Inclusive
Conventional Observation The nodes, or coding categories, are Aligned Deliberate, Meaningful, Sustainable
defined during the data analysis process. Trustworthy Explainable, Secure, Transparent
Directed Theory Nodes, or coding categories, may be Well-governed Accountable, Participatory, Regulated
established beforehand and throughout the
data analysis
Summative Keywords Keywords can be determined both prior to
and during the data analysis process
6
that is, those decisions which have the potential for significant impact strive to ensure AI is free from bias and deception and does not upend
either on individuals or on society as a whole” (Dubai City, 2023, p.15). existing societal structures. The acceptability of responsible AI tech-
nology has been elaborated on, through the sub-characteristics of: (a)
Ethical; (b) Equitable; (c) Harmless (Li, Yigitcanlar, Browne, & Nili,
4.2. Quantitative content analysis 2023).
Using NVivo software’s quantitative content analysis technique, the 4.4.1. Ethical consideration
AI strategy data underwent assessment. Fig. 2 depicts a hierarchical Local government collect large amounts of resident data, which raise
chart created in NVivo illustrating all nodes and sub-nodes, amalgam- numerous ethical and safety issues—from consent to quality control to
ated from the policy documents data following the word frequency security. As cities evolve into sensor-driven environments, the ethical
analysis. Each rectangular section provides an overview of the five nodes implications of pervasive surveillance become increasingly critical
along with their respective three sub-nodes. The size and colour of the (D’Amico et al., 2020). As highlighted in the Dubai city policy report:
rectangles signify the frequency, with the largest and darkest rectangles
representing the most frequent occurrences, while the smallest and “Consumers expect that as businesses and governments continue to collect
lightest rectangles denote the least frequent ones. and store unprecedented amounts of personal data, that they will simul-
taneously be capable of keeping personal information private and safe.
4.3. Qualitative content analysis When that does not happen and data privacy is violated, the trust is
broken, and with it, there are often significant financial losses for the
After the quantitative analysis, a qualitative content analysis was organisations whose data security is breached” (Dubai City, 2023, p.
undertaken. As shown in Fig. 3, a concept map was developed to illus- 18).
trate the five nodes and their associated sub-nodes. These concept maps In responses to these concerns, local governments have shown a
were generated using NVivo software by analysing frequently encoun- commitment to protecting community data dignity and privacy through
tered words within each sub-node. The findings are expounded upon in various initiatives, which can be categorised into three primary ap-
the subsequent sections and categorised under the five AI strategy proaches. Firstly, develop or adopt ethical guidelines/frameworks. For
domains. example, Barcelona City Council has adopted the European Union’s
General Data Protection Regulation (GDPR) to legitimate personal data
4.4. AI strategy on acceptability and prevent abuse of consent (Barcelona City, 2021). London is another
local government that adheres to GDPR guidelines and further devel-
The findings on the ‘acceptability’ characteristic show how the ini- oped to ensure the trust among the partners and citizens.
tiatives taken by local government are concerned with responsibility for
widely accepted, ethically uncontroversial, and harmless AI. Initiatives
Fig. 2. Hierarchy of nodes and sub-nodes.
7
Fig. 3. Responsible AI strategy concept map.
“London requires safe, secure, and useful ways of using and sharing data Council, monitoring its level of performance and collectively taking on the
which build trust among our partners and citizens. Living the spirit, not challenges as they come” (Barcelona City, 2021, p. 23).
just the letter of UK data protection (UK GDPR) means seeing the re-
Thus, the highlighted document indicates that local governments are
quirements as deep ethical obligations as well: making it easy for a user
taking proactive steps to ensure the ethical handling of data and AI
who does not wish for their data to be used in a particular way, or wishes
technologies. These efforts align with regulatory requirements and
to opt-out at a later stage, to do so and be clear about how long the data
emphasise the essence of ethical data practices in maintaining public
will be held for” (London City, 2021, p.4).
trust and ensuring the responsible use of personal information.
Secondly, initiatives involving providing training programmes to the
employees to understand the concept of ethics: 4.4.2. Equitable consideration
Ensuring equitable representation in the implementation of AI ap-
“(a) Employees involved in the development or use of AI technologies
plications is crucial for fostering inclusivity and fairness within com-
shall receive appropriate training on the ethical considerations, risks, and
munities. The Lebanon document underscores this commitment by
best practices related to AI; (b) Continuous learning and professional
stating:
development opportunities shall be provided to employees to stay updated
on emerging AI technologies and ethical guidelines” (Streaky Bay, 2023, “Using AI should not in any way discriminate against any individual’s
p.4). race, skin colour, national or ethnic origin, cultural group, language,
gender identity or expression, sexual orientation, mental or physical
Thirdly, establishing a dedicated council tasked with overseeing the
ability, age, religious or political opinion or activity, economic status,
ethical aspects of AI technology. The local government in Mexico
immigration status, or housing status. In accordance with City Code 185
established an ethics council composed of prominent ethicists, business
Welcoming Lebanon, every action carried out by City of Lebanon em-
leaders, and representatives from the national AI oversight office. This
ployees is a reflection upon the City, and thus it is imperative to act
council was tasked with setting guidelines and limits that align with
respectfully and responsibly” (Lebanon City, 2023, p.4).
Mexican values and principles (Mexico City, 2018). Similarly, the Bar-
celona City Council has set up a cross-cutting commission to guide their The system adopted by local authorities has come under increased
strategy on ethical AI: scrutiny in recent years, as there are concerns that it may be less accu-
rate for specific groups, potentially resulting in disparate impacts or
“A cross-cutting commission will be set up, made up of representatives
harms (New York City, 2024). This disciplinary is known as ‘bias’. Bias is
from the departments or manager’s offices of all the City Council’s areas,
a sign of much more serious technical problems (Ntoutsi et al., 2020).
to guide the strategy of incorporating an ethical AI into Barcelona City
8
Lebanon City report stated that: 4.5. AI strategy on accessibility

“This can come in many forms, such as, biased training data, sampling
The findings of the ‘accessibility’ characteristics show how the ini-
bias, and stereotyping/ societal biases” (Lebanon, 2023, p.4).
tiatives taken by local government are concerned with responsibility for
Nonetheless, we cannot exclusively assert that biased output is the providing local government AI based services in a technically adaptive,
result of technical mistakes; it may also occur because of human inter- cost-effective, and culturally inclusive manner, improving the circum-
vention in the system (New York City, 2024). While AI systems excel at stance for widespread use. The accessibility of responsible AI technology
pattern recognition and finding comparable, they are inherently limited has been elaborated on, through the sub-characteristics of: (a) Adapt-
in their ability to identify genuinely better cases (Nordic Municipalities, able; (b) Affordable; (c) Inclusive (Li, Yigitcanlar, Browne, & Nili, 2023).
2019).
The Streaky Bay policy document outlines three mechanisms to 4.5.1. Adaptable consideration
mitigate bias in AI systems as: AI has the calibre to adapt to their environment by using known data
and patterns, in so doing, autonomously generate new knowledge. The
“(a) Employees shall strive to minimize biases in AI systems and algo-
City of Vienna mentioned that:
rithms that could result in unfair or discriminatory outcomes; (b) Regular
monitoring and evaluation of AI systems shall be conducted to identify “AI consists of computer systems that can act intelligently. On that ac-
and address any biases or unintended consequences; (c) In cases where count, such systems usually do not simply implement instructions that
bias is detected, employees shall take appropriate measures to rectify the have been programmed into them. Rather, they adapt their behaviour to
issue, such as retraining models or adjusting algorithms” (Streaky Bay, suit the situation on hand using the information available” (Vienna City,
2023, p.4). 2019, p.7).
Thus, the highlighted document indicates the ongoing efforts by At the same time, the Singapore policy report stated that the citizens
local government to tackle algorithmic bias and promote inclusivity. also need to be ready to adapt to the new environment when AI becomes
These initiatives take significant steps to prevent unfair or discrimina- more prevalent. Because Singapore believes:
tory outcomes by explicitly considering potential biases, adopting
“At the societal level, as part of the overall promotion of digital literacy,
comprehensive guidelines, and providing employee training.
we will raise awareness of AI, so that citizens are prepared for techno-
logical change, and are engaged in thinking about AI’s benefits and im-
4.4.3. Harmless consideration
plications for the nation’s future. At the workforce level, we will prepare
The importance of human oversight in AI applications is critical for
our professionals to adapt to new ways of working, in which workers are
safeguarding the society’s physical and mental well-being. The Barce-
augmented by AI capabilities” (Singapore, 2019, p. 13).
lona City Council report highlighted that:
The Mexico policy report clearly stated the role that local govern-
“Any technological application launched by Barcelona City Council that
ment need to play to adapt to the AI environment. It elaborates this as:
affects people must be under human control, even in circumstances where
automated learning or similar techniques allow algorithms to make “A key role for local government: to set a strategic direction, to invest in
automated decisions” (Barcelona, 2021, p.14). Mexican data and research to help build tomorrow’s AI technologies, and
to support the learning that will help citizens adapt to technological
Accordingly, AI system developers and users must assume prime
change” (Mexico, 2018, p. 38).
responsibility and carefully check from the beginning of the process on:
(a) whether the available data is suitable for the intended analysis; (b) At the same time, one of the characteristics of the adaptable is us-
whether clear and understandable information is provided in explaining ability for a broader range of people Li, Yigitcanlar, Browne, and Nili
how an AI system works; (c) why it issues a given recommendation or (2023). The Kelowna City policy document stated that:
takes a given decision in a specific situation and which data, if any, is
“We need to make sure that all residents have fast and reliable internet
generated; (d) it is especially important to check whether the initial data
access, as well as the knowledge and skills to use it” (Kelowna City,
itself contains any bias to preclude self-fulfilling prophecies (Vienna,
2020, p.25).
2019).
Additionally, the Helsinki City policy document outlines how the The Dubai report stated that even though they have highest internet
degree of human control can influence AI functionality: access, there is major gap in digital skill:
“The human oversight section should provide information on how humans “It is a global leader in 5G adoption and has one of the highest internet
are in the loop or control, ensuring that your system does not undermine penetration rates in the region, with 100% of the population having
human autonomy or cause other adverse effects. Depending on your internet access still, there is a major talent gap when it comes to the digital
system’s autonomy level, this can be realised in many ways” (Helsinki skills in demand in the market” (Dubai, 2023, p. 114–114).
City, 2020, p.9).
Thus, the highlighted document indicates that, it is not only about
Similarly, the Tempe report emphasises the critical necessity of how AI is trained to adapt to the environment but also how people are
delineating the precise roles and duties of human operators involved in ready to understand and adapt the AI for their daily lives.
AI systems:
4.5.2. Affordable consideration
“We will clearly define the roles and responsibilities of human operators
Improving financial accessibility is one of the strategies adopted by
in utilizing AI systems, including their training requirements and obliga-
many local governments to explore the potential of AI (Valle-Cruz et al.,
tions for monitoring, and intervening in system decisions. We will
2021). These strategies can be investigated in three categories. Firstly,
encourage collaboration between humans and AI systems, leveraging the
local government itself to allocate the funds. For example, the Changsha
strengths of both to enhance decision-making processes and ensure that
Municipality policy document said that:
ultimate control remains with humans” (Tempe city, 2023, p.1).
“Set up a special fund for the development of the AI industry, with an
Thus, the reviewed policy document emphasised that human over-
annual expenditure of 500 million yuan, which is used to support the
sight is crucial for preventing AI from negatively impacting society’s
development of the city’s AI industry, focusing on supporting industrial
well-being. Ensuring human control, defining clear roles, and moni-
agglomeration, enterprise introduction and training, R&D innovation,
toring data for biases are essential steps to safeguard AI systems.
9
demonstration applications, talent training, financial innovation, and “The City of Lebanon will provide employees with access to a variety of
ecological optimisation” (Changsha Municipality, 2019, p.1). training resources to promote AI literacy and will assist in providing the
necessary knowledge in order to safely, efficiently, and responsibly use AI
Secondly, collaborating with private institutions. Cities such as
technology in the workplace” (Lebanon City, 2023, p.5).
London and New York took bold steps to encourage early-stage ideas by
providing venture capital funding, which opened up the private sector to This deliberation expands from AI professionals to non-AI pro-
work collaboratively with the public sector. It paved the way for several fessionals. For instance, Milton Keynes organizes monthly AI expert
start-ups instead of depending on public funds (Yigitcanlar, Sabatini- forums for the local IT community to develop their knowledge and skills.
Marques, Da-Costa, Kamruzzaman and Ioppolo, 2019; London City, Each of these forums features three distinct content types: (a) Motiva-
2021; New York City, 2024). tional keynote speakers addressing the most critical AI topics of the day;
Thirdly, providing subsidies to use the internal resources to improve (b) Coding workshops and machine learning sessions; (c) Roundtable
the city’s function. For example: discussions with AI practitioners (Milton Keynes, 2020).
Singapore focused on two types of training strategies to strengthen
“Encourage enterprises, colleges and universities, and scientific research
the AI initiatives in future:
institutions to fully play their technological and resource advantages to
build carriers such as maker spaces, incubation bases, and ‘double “ (a) Train more Singaporeans for high-quality AI jobs - Singapore will
innovation’ platforms. Each platform will be given a financial subsidy of establish more local talent pipelines to increase the quantity and quality of
up to 2 million yuan [~US$280K]” (Changsha Municipality, 2019, our AI workforce in the longer-term; (b) Teach basic computing skills and
p.2). computational thinking to all - As AI becomes more accessible and
prevalent, Singapore will need to develop AI-ready individuals who can
Thus, the reviewed policy documents emphasised that local gov-
understand and apply AI in their respective domains” (Singapore, 2019,
ernments are actively investing in dedicated funds, fostering public-
55–59).
private partnerships, and providing subsidies to enhance the integra-
tion of AI into their services, ultimately aiming to create more efficient Similarly, the Dubai report stated that knowledge training and ups-
and accessible public services for their communities. killing have created a parallel argument on job loss of the local gov-
ernment employees; meanwhile, it is paving the way to create new jobs
4.5.3. Inclusive consideration and further emphasised:
Limited policy documents have distinctly emphasised the signifi-
“While job loss through automation is a real phenomenon, so too is job
cance of diversity considerations. Kelowna policy document stated that
creation through the implementation of emerging technologies. Thus, the
inclusivity and diversity enhance strength and innovation of the com-
need to upskill workers for the digital age becomes of critical importance,
munity, and highlighted that:
so that current workers whose jobs either no longer exist, or have been
“Through the Imagine Kelowna engagement process (a vision created by partially automated, can be redeployed into new digital jobs” (Dubai,
our community), residents recognized that inclusivity and diversity make 2023, p.13).
us stronger and more innovative. And that the wellbeing of Kelowna de-
Thus, the highlighted document underscores the critical role of ed-
pends on the shared responsibility of individuals, businesses, government,
ucation and training in successfully adopting and integrating AI tech-
academia and community organizations coming together to address some
nologies. By investing in comprehensive AI education and fostering a
of the complex issues facing our community” (Kelowna, 2020, p.11).
culture of continuous learning, local governments ensure their work-
The London policy document examined diversity not only from the force is well-equipped to navigate the evolving technological landscape.
user’s perspective but also from the maker’s standpoint:
4.6.2. Meaningful consideration
“Hiring a more diverse and inclusive workforce and benefit from the lived
The significance of incorporating public perception and needs into
experience of Londoners in the team” (London City, 2021, p.4).
the development of AI applications was highlighted in policy docu-
“If certain groups are excluded from the benefit of new technologies, the ments. For instance, the Lebanon City document elaborated as:
technology should have a clear reason why it does not serve these groups”
“City employees should make efforts to gather the questions, concerns,
(London City, 2021, p.4).
and other types of feedback from Lebanon residents regarding AI systems
Thus, these highlighted policy documents advocate for a collabora- used by the City of Lebanon. By doing so, this can enhance fairness and
tive approach to create a more equitable and innovative society. inclusion in the responsible decision- making process” (Lebanon City,
2023, p.3).
4.6. AI strategy on alignment
However, it is evident that local authorities are struggling to incor-
porate public requirements due to the complex, diverse and dynamic
The findings of the ‘alignment’ characteristics show how the initia-
societies characterised by uncertainties. An example of this challenge
tives taken by the local government are aligned with social desirability,
was found in the Barcelona City Council report:
providing sufficient knowledge and environmental consideration to
solve the challenges accompanied by technology to improve the life and “Up to 30% of potential beneficiaries of public aid are not applying for
well-being of citizens using AI. The aligned characteristic of responsible them because they are unaware of their existence. If the Administration
AI technology has been elaborated through the sub-characteristics of: (a) has a good understanding of the reality of its citizens based on data,
Deliberate; (b) Meaningful; (c) Sustainable (Li, Yigitcanlar, Browne, & gathered with prior consent, it will be capable of anticipating and offering
Nili, 2023). citizens a certain service without having to wait for citizens to request it”
(Barcelona, 2021, p.6).
4.6.1. Deliberate consideration
Accordingly, local authorities are using communication channels to
Effective utilisation of the advantages offered by innovation and new
obtain feedback and perceptions from members of their communities.
technologies is contingent upon possessing the appropriate knowledge
The Barcelona document stated that communication platforms will be
(Lind et al., 2020). Knowledge not only increases willingness to adopt AI
used to reinforce the governing principles of transparency and the right
but also provides an understanding of its benefits and potential draw-
to information. It underlined that:
backs (Jarrahi et al., 2022). The Lebanon policy document specifically
noted:
10
“Barcelona City Council will make use of existing communication characteristics of: (a) Explainable; (b) Secure; (c) Transparent (Li,
channels with citizens to strengthen the governing principles on trans- Yigitcanlar, Browne, & Nili, 2023).
parency and right to information. This communication will serve a double
purpose: to provide information and resolve queries and disputes over the 4.7.1. Explainable consideration
use of AI in public services and to contribute to the democratisation of Explainability articulates the reasoning behind an AI system’s deci-
universal access to knowledge and innovation. Citizen communication sion, suggestion, or prediction (Grennan et al., 2020). However, the
will be essential for resolving disputes that the social players may pose ‘Blackbox problem’ is often cited in connection with AI, where it refers
relating to the use of AI systems in the public services provided by Bar- to having trouble explaining how the algorithms achieve their results,
celona City Council” (Barcelona City, 2021, p.22). often because they are very advanced and complex (Gryz & Rojszczak,
2021).
These findings underscore the necessity of proactive approaches in
For instance, the Nordic Municipality emphasised the importance of
incorporating public needs and preferences into AI development,
explaining algorithmic workings and their purposes to citizens (Nordic
ensuring that AI systems are responsive and aligned with the complex-
municipality, 2019). As preliminary strategy local authorities often
ities of societal demands.
resort documentation to explain the AI process. Barcelona City Council
plans to create a technical guide detailing the processes and consider-
4.6.3. Sustainable consideration
ations for promoting ethical AI within the council, and stated:
Innovation and technology should be committed to pursuing greater
sustainability outcomes, which should prioritise environmental sus- “A technical guide will be created detailing and exemplifying all the
tainability as a crucial step in the innovation process and social and processes and aspects that must be considered for promoting an ethical AI
economic sustainability (Li, Yigitcanlar, Browne, & Nili, 2023). in Barcelona City Council” (Barcelona City Council, 2021, p.20).
Accordingly, policy documents mentioned the AI project initiatives
Similarly, the City Council of Amsterdam and Helsinki produced an
aimed to safeguard the environment. For example, the Sounds of New
AI register, it is:
York City (SONYC) project was initiated to reduce urban noise pollution.
The low-power sensors record street sounds, and machine learning “A standardised, searchable, and archivable way to document the de-
processes the audio and trains the sensors to recognise and distinguish cisions and assumptions made in developing, implementing, managing
between various types of noise, such as music, jackhammers, yelling, and ultimately dismantling an algorithm. when applicable, explainability,
barking or sirens (New York City, 2024). Dalian and Fuzhou munici- can be given for public debate, independent auditors, and individuals
palities introduced an innovative ocean project. It researched and citizens. For civil society, it is a window into the artificial intelligence
implemented marine robot applications in the fields of maritime patrol, systems used by a government organisation” (Helsinki City, 2020, p.3).
environmental monitoring, fishing boat monitoring, and typhoon
Thus, this highlighted document emphasis the critical role of
disaster prevention. The initiative improved and guaranteed services in
explainability in addressing the black box problem and fostering public
the city’s sea area route and boosted disaster prevention and mitigation
trust in AI technologies.
capabilities, in addition to providing a batch of integrated new products
and services (Dalian Municipality, 2019; Fuzhou Municipality, 2020).
4.7.2. Secure consideration
Meanwhile, the Mayor of London has set a goal for the city to achieve
Policy documents highlight the importance of safety and security in
net-zero carbon emissions by 2030, while also expressing support for the
the application of AI within local government. These documents
public spaces contributing to London’s vibrant character. Consequently,
emphasise both internal and external aspects of AI security.
three strategies have been proposed:
Internal security is about protecting the data as more sensitive
“(a) Follow the Business Climate Guidance in the counting for the entire because it identifies or pertains to a sensitive subject, such as a person’s
life-cycle impact of the technology, including its design, use and eventual personal information (Barcelona City Council, 2021). These circum-
retirement. This may include de-carbonising production, considering en- stances might involve more stringent requirements for the protection of
ergy use in data processing and an end-of-life plan for hardware; (b) the data as well as stricter restrictions on its usage (New York, 2021).
through our planning policies we encourage the use of smart technologies Nonetheless, only few local authorities have taken initiative to
to gather data on the performance of the built and natural environment, secure the data. For example, Shanghai is having the municipal cyber-
including water and energy consumption, waste, air quality, noise, and security and information department to coordinate the cybersecurity
congestion. Non-personal data of this kind in public places can be shared and informatisation work in the development of the AI industry and take
with the London Datastore to ensure pan-London insights can be gained; responsibility for relevant network information content governance and
(c) over many decades, London has evolved, resulting in an extraordinary supervision management work (Shanghai, 2022).
web of distinctive residential streets, squares, markets, parks, offices, and External security, on the other hand, involves ensuring that AI ap-
industrial and creative spaces. Consideration should be given to the plications operate safely within their surrounding environment. For
impact on neighbourhood amenity of any hardware installed in the public example, Fuzhou, Nanjing, Zhengzhou, and Dalian municipality docu-
realm, and having a plan for its removal and responsible disposal when it ments stated that securing the leadership, talent protection, increasing
reaches the end of its useful life” (London City, 2021, p.5). financial security, improving policy guarantees, and environmental
protection are mentioned as the measures to protect the standard of AI
Thus, these initiatives underscore the critical role of AI in promoting
(Dalian Municipality, 2019; Fuzhou Municipality, 2020; Zhengzhou
sustainable development and safeguarding the environment for future
Municipality, 2021; Nanjing Municipality, 2022).
generations.
Thus, the findings underscore that the safety and security of AI in-
novations in society require a combination of both physical and digital
4.7. AI strategy on trustworthiness protections to minimise the risk of adverse consequences from these
technologies.
The findings of the ‘trustworthiness’ characteristics show how the
initiatives taken by local government are concerned with responsibility 4.7.3. Transparency consideration
for information transparency management, the level of understanding Transparency is considered as first principle of trustworthiness
and to what extent information is secured to minimise the risk of (Grimmelikhuijsen et al., 2013). The ability to analyse the operation of
advance consequences in adoption of AI. The trustworthy characteristic the system is made possible by transparency, which empowers people to
of responsible AI technology has been elaborated on, through the sub- act as responsible, independent auditors and civil society activists
11
(Helsinki, 2020). will proactively engage in ongoing monitoring, evaluation, and auditing to
Few local authorities follow the ‘open government’ or ‘open by ensure compliance with ethical standards” (Tempe City, 2023, p.1).
default’ approach. This is known as providing open access to informa-
Thus, this contrast illustrates varying approaches to accountability in
tion about how a government is run. In other words, making data open
AI governance, ranging from specific role assignments to comprehensive
and available is a standard practice (Kelowna City, 2020; Dubai City,
responsibility for outcomes and ethical compliance.
2023). Open data is a crucial enabler of economic innovation (OECD,
2020).
4.8.2. Participatory consideration
Meanwhile, the City of Lebanon adheres to a mechanism aimed at
The benefits of operating as a local AI ecosystem were stressed in
maintaining transparency by providing disclosure when AI technology
these local government policy documents. These advantages span from
solely develops content.
promoting well-being, creating equity and creating community oppor-
“City employees shall make efforts to disclose the use of AI systems in the tunities to sharing the time, effort, cost, and risk (Nordic Municipalities,
workplace when and where appropriate. Employees must be open to 2019). The collaboration includes a wide range of stakeholders: (a)
sharing information, when appropriate, about the data being used, algo- Communities; (b) Industry; (c) Other government organisations; (d)
rithms that are applied, and decision- making criteria. When disclosing Research institutions; (d) Civil society organisations.
the use of AI in the workplace, employees should utilize the guidance For example, Amsterdam City committed itself fully to reinforcing
provided in this policy. However, a process that is fully automated without the AI ecosystem, which concerns: (a) Investors; (b) Research in-
human intervention would need to provide appropriate disclosures. For stitutions; (c) Talent; (d) the Business community (Amsterdam City,
example, if an AI was programmed to read an agenda and then sum- 2020).
marize the contents to post on social media without any human oversight, In tandem, Tempe and Nordic policy documents emphasised the
the post would need to include that the text was AI-generated” (Lebanon importance of the collaboration as:
City, 2023, p.2).
“We will actively engage with stakeholders, such as users, experts, poli-
Thus, the highlighted document emphasises that by making AI pro- cymakers, and community organizations, to solicit diverse perspectives
cesses visible and comprehensible, local government can build and and feedback on the AI technology under consideration or currently used.
maintain public trust. This ultimately facilitates the successful integra- We will seek to foster collaboration and share knowledge to address
tion of technologies into public services. ethical challenges and ensure that AI technologies benefit society as a
whole” (Tempe City, 2023, p.2).
4.8. AI strategy on well-governance
“Conduct joint research and development projects. Several municipalities
were working on similar projects. It seems appropriate to pool the re-
The findings of the ‘well-governance’ characteristics show how the
sources. For example, the municipalities could work together to develop
initiatives taken by the local government are concerned with re-
solutions for artificial intelligence so that the risk and costs are shared”
sponsibility for following legal guidelines and governance requirements
(Nordic municipality, 2019, p.38).
throughout the technology’s lifecycle and for offering explicit account-
ability measures to force participants to think about and uphold Yet collaborative practices between organisations do not always go
responsible ideals. The trustworthy characteristic of responsible AI has smoothly (Benton, 2013). Several municipalities raised some issues that
been elaborated through the sub-characteristics of: (a) Accountable; (b) they thought it sensible to be aware of when working collaboratively.
Participatory; (c) Regulated (Li, Yigitcanlar, Browne, & Nili, 2023). This raises some important questions such as:
“How do you ensure that you only share data with parties that you trust,
4.8.1. Accountable consideration
and that they only receive the data that you want to share?”
In the context of AI, accountability means taking responsibility for
(Amsterdam, 2020, p. 14) and “should technology companies develop
the outcomes, judgements, or effects that result from applying a system
algorithms based on free use of public data and then sell the algorithms to
or model (New York City, 2024). Policy documents stated the role of the
public actors? Who will own the algorithms after they are fully devel-
actor’s accountability in the AI adoption process. For example, the
oped?” (Nordic Municipality, 2019, p. 24).
report of the City Council of Helsinki put forward that:
Thus, the highlighted document stress how collaboration often faces
“The accountability section helps you assign the responsibilities for your
challenges, such as ensuring data security and managing intellectual
system. Accountable people ensure the system is developed and main-
property, which require careful consideration and clear agreements to
tained responsibly throughout its entire lifecycle. We should look at the
address effectively.
internal actors and the external partners with a key role in either the
development or maintenance of the system. We suggest having in place at
4.8.3. Regulated consideration
least the following details in the accountability section: (a) Who are the
Policy documents pointed to the importance of having a proper: (a)
responsible organisations and departments? (b) How can the responsible
Regulatory framework; (b) Regulatory body; (c) Types of regulatory
parties (contact person, email, and phone) be contacted? (c) Which
approaches. For example:
external partners and suppliers are involved? Besides, we recommend
assigning a responsible manager, technical owner, and possible external “A clear legal framework is needed to guarantee the security. It is not
auditors for every system in the register” (Helsinki City, 2020, p.7). without significance to sort out the ethical issues regarding what AI is and
what it is not” (Wrocław City, 2021, p.30).
In contrast, municipalities like Tempe and Warrnambool take a more
comprehensive approach by assuming full responsibility for the actions Besides, Shanghai has the local government cybersecurity and in-
and impacts of AI technology. It is stated that: formation department to take charge of relevant network information
content, governance, and supervision management tasks, and coordi-
“We will establish standards for accountability and oversight throughout
nate the cybersecurity and information-technology work for the devel-
the selection, design, development, and deployment of AI technology. We
opment of the AI sector (Shanghai Municipality, 2022). The document of
will be responsible for the actions and impacts of the AI technology we use,
Amsterdam City proposed the following approaches:
and we will implement strategies to identify, mitigate and rectify any
potential harms or unintended consequences resulting from their use. We “(a) Legislation and regulation; (b) development of auditing procedures
for algorithms; (c) formulation of better (stricter) tendering requirements;
12
(d) greater in-house development capacity in the public sector so that essential throughout the AI adoption process to oversee the AI system
critical system need no longer be purchased from commercial providers” because skilled human judgment improves performance and balances
(Amsterdam City, 2020, p.22). fundamental rights (Desouza et al., 2020; Wirtz et al., 2018). Mean-
while, top policymakers—typically not technically trained—are at an
Despite these varied approaches, AI challenges existing policy and
increased risk of getting a ‘Blackbox’ nature as technological complexity
regulations to account for future issues emergent after AI devel-
rises (Marcucci et al., 2022). This raises concerns about the features of
opments—i.e., “there is also a concern that artificial intelligence will lead to
responsibility such as transparency and explainability.
changes in legislation rather than the legislation providing the framework for
Moreover, Fatima et al.’s (2020) study on national strategic AI plans
the development, for example if issues with the legislation come to light after
indicated that explainability is the most discussed concept at the na-
developing artificial intelligence” (Nordic Municipality, 2019, p.29).
tional level. However, it is the least mentioned characteristic on the local
Additionally, the Dubai policy report stated that: “Even if such policies
government scale. There may be several reasons why explainability is
were to be developed, regulatory bodies would not have the capacity to
frequently discussed nationally. Nonetheless, it appears that trans-
effectively implement them or monitor their implementation thus delegating
parency and explainability overlap at the local level as some researchers
the burden to companies who are already struggling to navigate the current
have used these terms interchangeably (Brand, 2022). Explainability is
regulatory environment” (Dubai, 2023, p.116).
not the same as complete algorithm transparency; explainability re-
Hence, these highlight the necessity for adaptive, forward-thinking
quires a balance between openness and transparency, and local au-
regulatory mechanisms that can evolve alongside technological ad-
thorities must clearly define when and how AI is used to interact with
vancements to ensure responsible AI adoption.
citizens and make decisions about their access to and distribution of
services (Robinson, 2020).
5. Findings and discussion
It is crucial to recognise the distinct approaches taken by national
and local-level AI responsible strategies. National-level strategies tend to
The significance of RIT cannot be overstated, as it plays a pivotal role
focus more on implementing specific instruments, while local govern-
in mitigating the adverse repercussions associated with AI technologies
ment policies often emphasise characteristics or principles. This differ-
and fostering public trust in AI systems. This endeavour is perpetual,
ence stems from the varying roles and responsibilities at each
demanding ceaseless vigilance, assessment, and enhancement of AI
governance level. National governments typically employ a range of
systems to ensure their alignment with society’s values and
instruments to govern AI responsibly, as identified by Djeffal et al.
anticipations.
(2022): (a) Direct regulation; (b) Governmental strategies; (c) Public
This study advocates that AI policy documents are a rich source of
investment; (d) Financial incentives; (e) Institutions; (f) Networks; (g)
information for local government to comprehensively understand: (a)
Data & monitoring; (h) Certification & labels; (i) Outreach & literacy. In
How AI-related strategies have been proposed or implemented by the
contrast, Liebig et al. (2022) found that subnational strategies, while
local authority; (b) Local government’s ability on a human, technical,
also using instruments, place greater emphasis on principles and char-
legal, and financial level to adopt an AI project and work on the capacity
acteristics such as: (a) Responsibility of state; (b) Leadership claims; (c)
gaps; (c) Whether AI projects are initiated by a private and/or public
Ethical principles, alongside more practical measures like; (d) Network/
institution and their level of functionality.
cooperation; (e) Competition; (f) Citizen consultation & civil society; (g)
The findings insist on developing collaboration (participatory)
Legislation; (h) Investment; (i) Certification.
locally. Collaboration is not a new concept for local government (Warm,
This distinction highlights a key relationship between national and
2011), but AI adoption in local government is new. Besides, it is a
local-level strategies: national strategies often provide the overarching
complex process due to the variety of processes of which AI technology
framework and instruments, while local strategies focus on imple-
consists of (Vogl et al., 2020). Research findings indicate that local
menting and adapting these to specific contexts, emphasising ethical
governments lack legal, financial, human, and technical resources to
principles and citizen engagement. For instance, a national strategy
implement AI technology, which is fundamental for the local govern-
might establish broad AI regulations and investment plans, while local
ment to adopt AI technology. Accordingly, most local authorities are
strategies translate these into concrete ethical guidelines and
working on developing a ‘triple helix partnership’ method to implement
community-based initiatives. Both levels work in tandem to build trust,
AI applications. It is a collaboration between local government, in-
ensure transparency, and embed ethical principles into the AI
dustry, and academic institutions. “The intention is that each partner
ecosystem, as noted by Taeihagh (2021) and Hagendorff (2020). This
contributes with unique resources and functional capabilities that the
complementary approach allows for a comprehensive governance
other lacks” (De Lima Figueiredo et al., 2023, p.854). This dependency
structure that addresses AI development and deployment at multiple
helps to make the collaboration stronger (O’Donnell, 2012). On the
levels, combining top-down policy instruments with bottom-up imple-
other hand, in the long run, the organisation’s changing nature and
mentation of responsible AI characteristics. This revised version
institutional context could lead to trust issues (Brem & Radziwon, 2017).
explicitly addresses the relationship between national- and local-level
Since the local government is responsible for the public, it could be
responsible AI strategies, highlighting their different focuses and how
particularly impacted. We believe that being accountable for the role
they complement each other in the overall governance of AI.
that each collaborator plays in the adoption process could be more
Nevertheless, the implementation of this complementary approach
credible. But accountable is one of the least mentioned characteristics
faces significant challenges, particularly at the local level. Another sig-
among the policy documents.
nificant finding is that the need to have an appropriate legal framework
The study findings indicate that adaptability is one of the least
is underlined in half of the policy documents. Nevertheless, the research
mentioned characteristics, whereas deliberation is the highest. Policy
findings indicated that none of the local authorities have one. Municipal
documents validate that AI literacy among technical and non-technical
governments have significantly expanded the use of AI (Herath & Mittal,
professionals is vital for adopting AI. For non-technical professionals,
2022). This spans from simple automating tasks to sending reminders to
AI literacy helps to be aware of what AI does, how local governments
police and judiciary as a form of support for the legal system. Local
utilize these tools, the advantages, and disadvantages of AI, and how
governments have come far without a proper AI regulatory framework
decision-makers protect citizens’ rights when utilizing AI, and so on.
(Wu & Liu, 2023). When an ethical concern arises, the local government
This facilitates community participation in the AI adoption process
halts the project. For example, in 2017, the Pittsburgh Bureau of Police
which helps reflect the community’s needs. For technical professionals,
and the Department of Innovation and Performance experimented with
it helps to uphold the communities’ ethics, equity, and diversity from the
a predictive policing algorithmic system in partnership with Carnegie
initial stage of the AI adoption process. Notably, skilled humans are
Mellon University’s Metro21 (Smart Cities Institute). That system is
13
currently suspended due to the issues related to data and transparency. resources of various districts (new districts) and the Shenzhen-Shantou
These emerging ethical issues require local governments to develop a Special Cooperation Zone, promote the establishment of artificial intelli-
policy framework. Still, the research findings indicate a ‘responsibility gence development alliances, and promote the smooth implementation”
gap’ of who is responsible and liable for the ethical issues of AI. (Shenzhen Municipality, 2019, p.19); (c) Accelerated innovation - Joint
In addition to ethical considerations, AI governance must also work will be carried out with the world of AI research and managers of sci-
address environmental sustainability. AI for sustainability addresses entific infrastructures to boost applied research and technological-innovation
environmental issues and spans several fields of study (Yigitcanlar, projects”(Barcelona City, 2021, p.23); (d) Knowledge sharing – “The
2021). As a powerful and promising technology for environmental sus- universities work with local businesses and MK Council to trial projects,
tainability, AI leaves a significant carbon footprint that directly affects produce innovation and create knowledge transfer partnerships” (Milton
the environment (Nishant et al., 2020). Nevertheless, few documents Keynes, 2020, p.14); (e) Risk mitigation – “significant potential benefits
have considered the negative impact on the environment. This may be also in collaboration on identifying human-rights related risks and finding
because the environmental issues of AI still have not come to the fore- good ways to address those, via greater visibility into AI portfolios of different
front of the public attention as much as ethical issues. The lack of government entities” (Helsinki City, 2020, p.10).
awareness of AI could be one potential reason for this. Also, policy Stakeholder collaboration fosters trust, accountability, and legiti-
documents have prioritised social, economic, and legal factors over macy in AI governance processes, leading to greater public acceptance
environmental factors. Policies and regulations, private industry initia- and support. Ultimately, by embracing stakeholder collaboration, local
tives on AI (e.g., Google, Amazon, and Microsoft addressed sustain- governments can navigate the intricacies of AI adoption more effectively
ability through commitments to be carbon-free or negative by 2030), and realise AI’s full potential to benefit society while minimising
reporting carbon consumption level on the technical documents, and adverse impacts.
providing sustainability certification are a few actions followed at the
national level to protect the environmental sustainability (OECD.AI, 5.1.2. Capacity development
2023). Despite the commitments that have been made to meet envi- Capacity development refers to enhancing individuals or organisa-
ronmental goals at the national level, the degree of such commitments at tions’ abilities, skills, knowledge, and resources to achieve their goals
the local level is comparatively low (Arbolino et al., 2018). effectively and sustainably (Anuoluwapo Durokifa et al., 2023). Funding
In sum, these findings underscore the critical implications for the is essential for increasing AI activities (Fatima et al., 2020), and local
integration of AI in local government. Collaborative efforts involving government financial constraints limit or postpone AI adoption (Volmer,
both industry and academia must be established, with clearly defined 2021). AI needs a large amount of technical and human infrastructure
roles, a focus on trust-building, and the implementation of robust throughout the AI process, and the costs are high (Desouza et al., 2020;
accountability mechanisms. Striking a balance between ethical consid- Mikhaylov et al., 2018).
erations and the adaptability required for AI advancements is para- Meantime, larger technological capacity is vital for local govern-
mount, necessitating continuous capacity-building initiatives. ments to successfully adopt AI (David et al., 2023). This entails having
Distinguishing between transparency and explainability in AI in- robust technological infrastructure encompassing hardware, software,
teractions in local government policy documents is essential and calls for and networks to support AI initiatives. With a solid foundation in place,
establishing clear guidelines and developing industry standards. Incor- local governments can effectively deploy AI systems and applications,
porating environmental sustainability into AI policies is a pressing ensuring they have the necessary computational power, storage capa-
concern, and this endeavour involves comprehensive awareness cam- bilities, and network bandwidth to handle complex AI tasks.
paigns, regulatory measures, and commitments to sustainability prin-
“Set up a special fund of 10 million yuan [~USD1.4M] every year to
ciples. The formulation of specialised legal frameworks holds significant
carry out the major municipal-level science and technology project of
weight in addressing ethical concerns and ensuring the responsible
“R&D and application of key artificial intelligence technologies“ and
adoption of AI in municipal contexts. Creating well-structured ethics and
support the joint implementation of industry-university-research in-
accountability frameworks is equally imperative, as they define roles
stitutes, with a subsidy of 1 million yuan for each project” (Fuzhou
and bridge the so-called ‘responsibility gap’. Moreover, this study re-
Municipality, 2020, p.26).
veals that for AI technology to uphold RIT characteristics which of those
key local government strategies need to be prioritised. These are pre- Additionally, local governments should focus on educating the
sented in the following section. public about the intended use cases for AI in addressing public chal-
Furthermore, this study emphasises the importance of prioritising lenges, how the technology will be implemented, the anticipated goals,
specific key local government strategies to ensure that AI technology and the specific steps that will be taken to achieve those objectives
consistently upholds RIT characteristics. By identifying these critical (OECD, 2020). “AI ethics training - Elements of AI supporting people to learn
strategies, the study provides a roadmap for local governments to what AI is and Ethics of AI how to start thinking about AI from an ethical
effectively integrate responsible AI practices. These strategies are pre- point of view” (London City, 2021, p.6).
sented in the following section. Publics will benefit from increasing awareness and knowledge about
AI applications (Robinson, 2020). Nonetheless, the diversity of audi-
5.1. Local government responsible AI key strategy priority areas ences must be acknowledged; this diversity must be accommodated
through various teaching methodologies as per levels of digital literacy
5.1.1. Stakeholder collaboration among different generations. Meanwhile, providing proper internet
Collaboration plays a crucial role in maximising the benefits of AI accessibility and knowledge is vital for shared input to enable the
technology adoption for local governments, enabling them to harness community to understand and reflect on their requirements. It is also
the full potential of AI while addressing associated challenges more important to provide training programs, workshops, and certification
effectively (Mikhaylov et al., 2018). The fact that they do not share the courses to local government employees to develop, implement, and
same goals or timelines makes it difficult (Benton, 2013). However, in manage AI systems effectively.
the end, each collaborator receives benefits and support for the city’s
“Cyber Services in coordination with the TRC will provide training op-
practical functionalities (Mikhaylov et al., 2018). Ideally, local gov-
portunities for City employees to provide them with the necessary
ernment receive multiple benefits through the collaboration: (a) Ethical
knowledge to use AI technology safely and appropriately. Employees who
influences – “We will seek to foster collaboration and share knowledge to
utilize AI in their work shall be trained on those systems” (Lebanon City,
address ethical challenges and ensure that AI technologies benefit society as a
2023, p.7).
whole” (Tempe City, 2023, p.2); (b) Resource pooling – “Coordinate the
14
5.1.3. Data governance “Follow the Business Climate Guidance in the counting for the entire life-
Data governance ensures that data is managed effectively, securely, cycle impact of the technology, including its design, use and eventual
and ethically, throughout the AI adoption lifecycle (Janssen et al., retirement. This may include de-carbonising production, considering en-
2020). ergy use in data processing and an end-of-life plan for hardware”
The risk of insufficient demographic representation occurs when the (London City, 2021, p.5).
AI does not adequately reflect the diversity and different groups are not
Furthermore, there are numerous places in the AI adoption process
represented evenly (Cheng et al., 2022). Population exclusion damages
where problems can be raised. Local authorities ensure that the entire AI
the AI system’s performance and fairness. If the sample size is too small
ecosystem is secure before the AI application is released to the public
or there are too many data points from one group, there are two possible
(Arnold et al., 2019; Ryan & Stahl, 2020) and secured in terms of
causes of these imbalances (Ntoutsi et al., 2020). It is critical for the local
financially stable, legally protected, sufficient technical and human re-
government to recognise these biases at the data collection stage itself
sources, and so on. Local government ensures the importance of having
due to the adverse consequences of these biases (Cheng et al., 2022;
trained experts involved throughout the process. Unfair outcomes are
Mehrabi et al., 2021).
more likely since algorithms cannot detect bias, overrepresentation,
“Ensuring that the collection of any personal data is justified and mindful imbalance, geographic misalignment, and other data input issues
that even if you don’t mean to identify people in your smart places pro- (Cheng et al., 2022). This could be more dangerous in the absence of
jects, if you can distinguish them from other individuals it will trigger data human oversight. Addressing this responsible characteristic by policy-
protection law; maintaining high quality data standards with comple- makers could lead to successful AI adoption by the local government.
mentary metadata so it is clear where the data comes from and how it was
collected” (London City, 2021, p.5).
5.2. Limitations of the study
Also, gathering information about individuals without consent ac-
celerates privacy violations (Herschel & Miori, 2017; Mai, 2016). The While the findings of this research contribute to the field, it is
local government ensures safety protocols to monitor AI systems which important to recognise its limitations. (a) This study focused exclusively
are likely to raise ethical issues (Shneiderman, 2020). on AI-specific policy documents published by local governments or
developed in collaboration with private institutions. It did not include
“Ensuring the accuracy of data put into the AI system and the AI system’s
broader smart city or digital strategies that might contain AI-related
outputs; Disposal of the data once the purpose of using the data has been
content. This narrow focus may have excluded relevant AI strategies
fulfilled, when possible, in compliance with applicable state and federal
embedded within wider technological frameworks; (b) the research
laws; Providing data subjects with control and transparency in relation to
analysed only 26 local government policy documents. This limited
data processing” (New York City, 2024, p.4).
sample size may not fully represent the diverse range of AI policies
across all local governments globally. This approach may lead to
5.1.4. Regulatory compliance incomplete information or unintended prejudices in data analysis. A
Regulatory compliance refers to the process of ensuring that an larger sample could provide a more comprehensive and accurate
organisation adheres to relevant laws, regulations, guidelines, and perspective on local government AI policy trends. Furthermore, it is
standards that are applicable to its operations (Hashmi et al., 2018). important to note that many local governments may have AI initiatives
Before implementing AI technology for municipal services, local au- or practices without related policy documents, which could result in an
thorities conduct a comprehensive evaluation of the current legislation, underrepresentation of actual AI efforts at the local level; (c) the coding
and regulations related to AI. This spans national laws and regulations, of policy documents was based on the authors’ interpretation of Li,
human data privacy and protection regulations, cross-sectoral regula- Yigitcanlar, Browne, and Nili’s (2023) RIT framework. While this
tions, and so on. approach provided consistency, it’s important to acknowledge that
alternative interpretations of the RIT framework exist. Different in-
“We will comply with applicable laws, standards and regulations related
terpretations could potentially lead to variations in the categorization
to AI and data protection. We will actively support the development of
and analysis of the policy content; (d) the documents varied significantly
ethical AI regulations and frameworks that align with our principles and
in format and level of detail, spanning from detailed depictions to
contribute to the responsible use of AI technology” (Tempe City, 2023,
concise summaries, which constrained the depth of analysis and un-
p.2).
derstanding; (e) AI is a fast-changing area, and policies may quickly
Meanwhile, understanding the distribution of responsibility will be become outdated. The review might not capture the most recent de-
particularly useful for local governments to implement the application velopments or changes in local government approaches; (f) different
of AI (Bedué & Fritzsche, 2022). Accountability issues arise if re- local governments may use varying terms or definitions for similar
sponsibilities are unclear (Santoni de Sio & Mecacci, 2021). Local gov- concepts. This limitation highlights the lack of standardisation in AI
ernments must take the lead in distributing the responsibilities among policy language across different local governments, making comparative
the stakeholders to make it a shared commitment. analyses challenging.
“We will clearly define the roles and responsibilities of human operators
6. Conclusion
in utilizing AI systems, including their training requirements and obliga-
tions for monitoring, and intervening in system decisions” (Tempe City,
In the context of local government, this study breaks new ground by
2023, p.1).
examining the emerging concept of RIT from a local government
Furthermore, local governments have recognised AI’s potential to perspective, drawing practical insights from policy documents. It is
address several environmental issues (Son et al., 2023), yet it is note- crucial to acknowledge that the findings from our limited sample of 26
worthy that AI also causes environmental effects due to a large amount local government policy documents from number of countries may not
of energy consumption (Vinuesa et al., 2020). Thus, local governments be interpreted as universally applicable to all local governments
develop clear and enforceable rules and standards on the development, worldwide. The intricate diversity in governance structures, cultural
application, and eventual disposal of AI technologies to mitigate these contexts, and technological readiness across different regions precludes
possible negative environmental impacts of AI. such broad generalisations. Instead, these findings highlight the
responsible AI characteristics and approaches adopted by a myriad of
local governments, revealing how they interpret, adapt, and implement
15
within their specific contexts. to develop or refine their responsible AI policies, ensuring they
Accordingly, the research indicates that local governments are comprehensively address key issues.
increasingly prioritising RIT within their policy frameworks. This shift ▪ Practitioners could use the study’s findings to identify and
in focus among local governments reflects a growing awareness of the adopt best practices from leading municipalities, enhancing
need to balance technological advancement with societal values and their approach to responsible AI implementation.
expectations. Unlike previous studies that may have focused on national ▪ The study could assist local governments in crafting more
or private sector approaches, this research highlights how local gov- effective risk mitigation strategies for AI deployment, informed
ernments are adapting RIT principles to address the unique challenges by the common challenges identified across various policy
and opportunities presented by AI at the local level. documents.
Key findings from the review of local government policy documents ▪ Additionally, the findings could inform strategies for better
reveal that not all RIT characteristics have been given equal consider- engagement with citizens, private sector partners, and other
ation. While participatory and deliberate considerations are frequently stakeholders in the development and implementation of AI
mentioned, adaptable, sustainable, affordable and inclusive character- policies.
istics are often overlooked. Policy documents reveal that local govern-
ments often lack the necessary legal, financial, human, and technical In conclusion, the significance of RIT in local government AI stra-
resources to implement AI effectively. This resource gap necessitates the tegies is paramount, as it is crucial for mitigating potential negative
development of collaborative partnerships. Such collaborations can pool impacts and building public trust. Future research should delve deeper
unique resources and capabilities, thereby strengthening the overall AI into these aspects, offering additional insights into the formulation of
adoption process (Chowdhury et al., 2023). A critical finding of this responsible AI policies at the local government level.
study is the significant gap between the recognition of need and actual
implementation of legal frameworks for AI governance at the local level. CRediT authorship contribution statement
While half of the examined policy documents underline the necessity for
an appropriate legal framework, our research reveals that none of the Anne David: Writing – original draft, Investigation, Formal analysis,
local authorities have actually established one. Data curation. Tan Yigitcanlar: Writing – review & editing, Supervi-
This discrepancies between acknowledgement and action represent a sion, Funding acquisition, Conceptualization. Kevin Desouza: Writing –
crucial area for future development in local government AI strategies. review & editing, Supervision. Rita Yi Man Li: Writing – review &
Accordingly, findings indicate that local governments are increasingly editing. Pauline Hope Cheong: Writing – review & editing. Rashid
adopting strategies of stakeholder collaboration, capacity development, Mehmood: Writing – review & editing. Juan Corchado: Writing – re-
robust data governance, and regulatory compliance to uphold RIT view & editing.
characteristics in their AI initiatives. By focusing on these strategies,
local governments navigate the complexities of AI adoption more
Funding
effectively, ensuring that AI technologies benefit society while mini-
mising adverse impacts. The study also emphasises the need for AI lit-
This research was funded by the Australian Research Council Dis-
eracy among technical and non-technical professionals. For non-
covery Grant Scheme, grant number DP220101255.
technical staff, understanding AI’s functions, benefits, and risks is
crucial for informed decision-making and public engagement. For
technical professionals, AI literacy ensures adherence to community Declaration of competing interest
ethics, equity, and diversity. Transparency and explainability are key to
building public trust (Felzmann et al., 2020), requiring clear guidelines The authors declare no conflict of interest.
and standards to delineate these concepts. Also, the findings indicate
that the environmental impact of AI technologies is an area that requires
Acknowledgements
greater attention. Despite national-level commitments to sustainability,
local governments have yet to integrate environmental considerations
The authors thank the editor and anonymous referees for their
into their AI policies fully.
invaluable comments on an earlier version of the manuscript.
Despite the earlier mentioned limitations, this research establishes a
foundation for the discourse on responsible AI adoption within local
Data availability
governments. By examining municipal policy documents, we developed
a more practical and detailed review that can serve as a tool for local
Data will be made available on request.
authorities, practitioners, researchers, and other stakeholders to eval-
uate the responsibility of technological outcomes. Specifically for
References
research implications:
Abdelmutti, N., & Hoffman-Goetz, L. (2009). Risk messages about HPV, cervical cancer,
▪ This study has the potential to contribute significantly to the and the HPV vaccine Gardasil: A content analysis of Canadian and U.S. national
development of a comprehensive theoretical framework for newspaper articles. Women & Health, 49(5), 422–440.
AlShebli, B., Memon, S., Evans, J., & Rahwan, T. (2023). China and the U.S. produce more
responsible AI adoption in local governments, addressing a
impactful AI research when collaborating. arXiv preprint arXiv:2304.11123.
notable gap in the existing literature. Amsterdam City. (2020). Agenda AI: Amsterdam intelligence. https://assets.amsterdam.
▪ Future research could focus on comparative analyses across nl/publish/pages/1018100/agenda_ai_eng_1.pdf.
Arbolino, R., Carlucci, F., De Simone, L., Ioppolo, G., & Yigitcanlar, T. (2018). The policy
different municipalities, regions, or countries to identify best
diffusion of environmental performance in the European countries. Ecological
practices and variations in responsible AI strategies. Indicators, 89, 130–138.
▪ Longitudinal studies could be conducted to track the evolution Arnold, M., Piorkowski, D., Reimer, D., Richards, J., Tsay, J., Varshney, K., … Olteanu, A.
of AI policies in local governments over time, offering valuable (2019). Factsheets: Increasing trust in AI services through supplier’s declarations of
conformity. IBM Journal of Research and Development, 63(4/5), 1–13.
insights into the processes of policy development and Atkinson, K., Baroni, P., Giacomin, M., Hunter, A., Prakken, H., Reed, C., … Villata, S.
adaptation. (2017). Towards artificial argumentation. AI Magazine, 38(3), 25–36.
▪ The findings from this study could have substantial practical Barcelona City Council. (2021). Government measure for municipal algorithms and data
strategy for an ethical promotion of artificial intelligence. https://ajuntament.barce
implications. Local governments could leverage these insights lona.cat/digital/sites/default/files/mesura_de_govern_intel_ligencia_artificial_eng.
pdf.
16
Bedué, P., & Fritzsche, A. (2022). Can we trust AI? An empirical investigation of trust Grennan, L., Kremer, A., Singla, A., & Zipparo, P. (2020). Why businesses need
requirements and guide to successful AI adoption. Journal of Enterprise Information explainable AI—And how to deliver it. https://www.mckinsey.
Management, 35(2), 530–549. com/capabilities/quantumblack/our-insights/wh
Benton, J. (2013). Local government collaboration. State and Local Government Review, 45 y-businesses-need-explainable-ai-and-how-to-deliver-it#/.
(4), 220–223. Grimmelikhuijsen, S., Porumbescu, G., Hong, B., & Im, T. (2013). The effect of
Bogner, A., & Torgersen, H. (2018). Precaution, responsible innovation and beyond–in transparency on trust in government: A cross-national comparative experiment.
search of a sustainable agricultural biotechnology policy. Frontiers in Plant Science, 9, Public Administration Review, 73(4), 575–586. https://doi.org/10.1111/puar.12047
Article 421461. Gryz, J., & Rojszczak, M. (2021). Black box algorithms and the rights of individuals: No
Bourban, M., & Rochel, J. (2021). Synergies in innovation: Lessons learnt from easy solution to the “explainability” problem. Internet Policy Review, 10(2), 1–24.
innovation ethics for responsible innovation. Philosophy & Technology, 34(2), Gurr, G., & Metag, J. (2022). Content analysis in the research field of technology
373–394. coverage. In Standardisierte Inhaltsanalyse in der
Brand, D. (2022). Responsible artificial intelligence in government: Development of a Kommunikationswissenschaft–standardized content analysis in communication research:
legal framework for South Africa. JeDEM - EJournal of EDemocracy and Open Ein Handbuch-A handbook (pp. 239–247). Wiesbaden: Springer Fachmedien
Government, 14(1), 130–150. Wiesbaden.
Brem, A., & Radziwon, A. (2017). Efficient triple Helix collaboration fostering local niche Hagendorff, T. (2020). The ethics of AI ethics: an evaluation of guidelines. Minds and
innovation projects – A case from Denmark. Technological Forecasting and Social Machines, 30(1), 99–120. https://doi.org/10.1007/s11023-020-09517-8
Change, 123, 130–141. Haggarty, L. (1996). What is content analysis? Medical Teacher, 18(2), 99–101.
Buhmann, A., & Fieseler, C. (2021). Towards a deliberative framework for responsible Hashmi, M., Governatori, G., Lam, H., & Wynn, M. (2018). Are we done with business
innovation in artificial intelligence. Technology in Society, 64, Article 101475. process compliance: State of the art and challenges ahead. Knowledge and Information
Butt, A., Kroen, A., Steele, W., & Dühr, S. (2021). Local government co-ordination: Systems, 57(1), 79–133.
Metropolitan governance in twenty-first century Australia. AHURI Final Report, 352. Helsinki City. (2020). Public AI register. https://algoritmeregister.amsterdam.nl/wp-co
https://doi.org/10.18408/ahuri5323001 ntent/uploads/White-Paper.pdf.
Cardno, C. (2019). Policy document analysis: A practical educational leadership tool and Herath, H., & Mittal, M. (2022). Adoption of artificial intelligence in smart cities: A
a qualitative research method. Educational Administration: Theory and Practice, 24(4), comprehensive review. International Journal of Information Management Data Insights,
623–640. 2(1), Article 100076.
Changsha Municipality. (2019). Opinions of the general office of the Changsha municipal Herschel, R., & Miori, V. (2017). Ethics & big data. Technology in Society, 49(0160-791X),
people’s government on further promoting the development of the artificial 31–36.
intelligence industry. http://www.tianxin.gov.cn/zwgk8/bmxxgkml/qgyhxxhj51/ Hine, E., & Floridi, L. (2022). Artificial intelligence with American values and Chinese
720603/1170827/201907/t20190704_7364599.html. characteristics: A comparative analysis of American and Chinese governmental AI
Cheng, X., Lin, X., Shen, X., Zarifis, A., & Mou, J. (2022). The dark sides of AI. Electronic policies. AI & Society.. https://doi.org/10.1007/s00146-022-01499-8
Markets, 32(1), 11–15. Hsieh, H., & Shannon, S. (2005). Three approaches to qualitative content analysis.
Chowdhury, S., Dey, P., Joel-Edgar, S., Bhattacharya, S., Rodriguez-Espindola, O., Qualitative Health Research, 15(9), 1277–1288.
Abadie, A., & Truong, L. (2023). Unlocking the value of artificial intelligence in Hurwitz, L., Alvarez, A., Lauricella, A., Rousse, T., Montague, H., & Wartella, E. (2016).
human resource management through AI capability framework. Human Resource Content analysis across new media platforms: Methodological considerations for
Management Review, 33(1), Article 100899. https://doi.org/10.1016/j. capturing media-rich data. New Media & Society, 20(2), 532–548.
hrmr.2022.100899 Imaz, O., & Eizagirre, A. (2020). Responsible innovation for sustainable development
Constantras, E., Geiger, G., Braun, J.-C., Mehrotra, D., & Aung, H. (2023). Inside the goals in business: An agenda for cooperative firms. Sustainability, 12(17), 6948.
suspicion machine. WIRED. https://www.wired.com/story/welfare-state-algorithms Janssen, M., Brous, P., Estevez, E., Barbosa, L., & Janowski, T. (2020). Data governance:
. Organizing data for trustworthy artificial intelligence. Government Information
Dalian Municipality. (2019). Development plan for a new generation of artificial Quarterly, 37(3), Article 101493.
intelligence. https://www.ndrc.gov.cn/fggz/cxhgjsfz/dfjz/201910/t20191025_119 Jarrahi, M. H., Askay, D., Eshraghi, A., & Smith, P. (2022). Artificial intelligence and
5518.html. knowledge management: A partnership between human and AI. Business Horizons, 66
D’Amico, G., L’Abbate, P., Liao, W., Yigitcanlar, T., & Ioppolo, G. (2020). Understanding (1). https://doi.org/10.1016/j.bushor.2022.03.002
sensor cities: Insights from technology giant company driven smart urbanism Jeffares, S., & Jeffares, S. (2021). Artificial intelligence and frontline public service. The
practices. Sensors, 20(16), 4391. virtual public servant: Artificial intelligence and frontline work. https://doi.org/
David, A., Yigitcanlar, T., Li, R., Corchado, J., Cheong, P., Mossberger, K., & 10.1007/978-3-030-54084-5_1
Mehmood, R. (2023). Understanding local government digital technology adoption Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines.
strategies: A PRISMA review. Sustainability, 15(12), 9645. Nature Machine Intelligence, 1(9), 389–399.
De Laat, P. (2021). Companies committed to responsible AI: From principles towards Kelowna City. (2020). Intelligent city strategy. https://www.kelowna.ca/sites/files/1/d
implementation and regulation? Philosophy & Technology, 34(4), 1135–1193. ocs/related/intelligent_city_strategy_-_city_of_kelowna.pdf.
De Lima Figueiredo, N., Fernandes, C., & Abrantes, J. (2023). Triple helix model: Landeweerd, L., Townend, D., Mesman, J., & Van Hoyweghen, I. (2015). Reflections on
Cooperation in knowledge creation. Journal of the Knowledge Economy, 14(2), different governance styles in regulating science: A contribution to ‘responsible
854–878. research and innovation’. Life sciences, society and policy, 11, 1–22.
Delecraz, S., Eltarr, L., Becuwe, M., Bouxin, H., Boutin, N., & Oullier, O. (2022). Lebanon City. (2023). Administrative policies & procedures: Use of artificial intelligence.
Responsible artificial intelligence in human resources technology: An innovative City of Lebanon, New Hampshire. https://portal.laserfiche.com/Portal/DocView.asp
inclusive and fair by design matching algorithm for job recruitment purposes. x?id=242719&repo=r-5a00306c.
Journal of Responsible Technology, 11, Article 100041. Li, W., Yigitcanlar, T., Browne, W., & Nili, A. (2023). The making of responsible
Desouza, K., Dawson, G., & Chenok, D. (2020). Designing, developing, and deploying innovation and technology: An overview and framework. Smart Cities, 6(4),
artificial intelligence systems: Lessons from and for the public sector. Business 1996–2034.
Horizons, 63(2), 205–213. Li, W., Yigitcanlar, T., Nili, A., & Browne, W. (2023). Tech giants’ responsible innovation
Djeffal, C., Siewert, M. B., & Wurster, S. (2022). Role of the state and responsibility in and technology strategy: An international policy review. Smart Cities, 6(6),
governing artificial intelligence: A comparative analysis of AI strategies. Journal of 3454–3492.
European Public Policy, 29(11), 1–23. https://doi.org/10.1080/ Liebig, L., Güttel, L., Jobin, A., & Katzenbach, C. (2022). Subnational AI policy: Shaping
13501763.2022.2094987 AI in a multi-level governance system. AI & SOCIETY.. https://doi.org/10.1007/
Dubai City. (2023). Smart Dubai: AI ethic principles and guidelines. Smart Dubai. https: s00146-022-01561-5
//www.digitaldubai.ae/docs/default-source/ai-principles-resources/ai-ethics.pdf. Lind, J., Pelger, S., & Jakobsson, A. (2020). Students’ knowledge of emerging technology
Durokifa, A., Johnson, U., Ramlachan, P., Mdhluli, S., & Magqirana, A. (2023). and sustainability through a design activity in technology education. International
Strengthening developmental local government in South Africa: The role of capacity Journal of Technology and Design Education.. https://doi.org/10.1007/s10798-020-
building. Journal of Public Administration, 58(2), 491–508. 09604-y
European Commission. (2021). Ethics guidelines for trustworthy AI. https://digital-strate London City. (2021). An emerging technology charter for London. Mayor of London
gy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai. assembly. https://www.london.gov.uk/publications/emerging-technology-chart
Farida, I., Ningsih, W., Lutfiani, N., Aini, Q., & Harahap, E. P. (2023). Responsible urban er-london.
innovation working with local authorities a framework for artificial intelligence (AI). Longley, R. (2019). What is a city state? Definition and modern examples. https://www.
Scientific Journal of Informatics, 10(2), 121–126. thoughtco.com/what-is-a-city-state-4689289#:~:text=Robert%20Longley%20is%
Fatima, S., Desouza, K., & Dawson, G. (2020). National strategic artificial intelligence 20a%20U.S.
plans: A multi-dimensional analysis. Economic Analysis and Policy, 67, 178–194. Lubberink, R., Blok, V., van Ophem, J., & Omta, O. (2017). Lessons for responsible
Felzmann, H., Fosch-Villaronga, E., Lutz, C., & Tamò-Larrieux, A. (2020). Towards innovation in the business context: A systematic literature review of responsible.
transparency by design for artificial intelligence. Science and Engineering Ethics, 26 Social and Sustainable Innovation Practices. Sustainability, 9(5), 721.
(6), 3333–3361. https://doi.org/10.1007/s11948-020-00276-4 Mai, J. (2016). Big data privacy: The datafication of personal information. The
Fuzhou Municipality. (2020). Implementation plan for promoting the accelerated Information Society, 32(3), 192–199.
development of new generation artificial intelligence. https://www.fuzhou.gov. Marcucci, S., Kalkar, U., & Verhulst, S. (2022). AI localism in practice: Examining how cities
cn/zfxxgkzl/szfbmjxsqxxgk/szfbmxxgk/fzsrmzfbgt/zfxxgkml/gmjjhshfzghzxghqygh govern AI. Available at SSRN 4284013.
jxgzc_2571/202002/t20200219_3201644.htm. Masuda, H., Kawakubo, S., Okitasari, M., & Morita, K. (2022). Exploring the role of local
Gianni, R., Lehtinen, S., & Nieminen, M. (2022). Governance of responsible AI: From governments as intermediaries to facilitate partnerships for the sustainable
ethical guidelines to cooperative policies. Frontiers in Computer Science, 4, Article development goals. Sustainable Cities and Society, 82, Article 103883.
873437.
17
Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2021). A survey on Simon, D., Canzler, W., Kuhlmann, S., & Stamm, J. (2019). Handbook on science and public
bias and fairness in machine learning. ACM Computing Surveys, 54(6), 1–35. policy. Edward Elgar.
Mexico City. (2018). Towards an AI strategy in Mexico: Harnessing the AI revolution. htt Singapore City State. (2019). National Artificial intelligence strategy: Advancing our
p://go.wizeline.com/rs/571-SRN-279/images/Towards-an-AI-strategy-in-Mexico. smart nation journey. https://www.smartnation.gov.sg/files/publications/nationa
pdf. l-ai-strategy.pdf.
Mikalef, P., Conboy, K., Lundström, J., & Popovič, A. (2022). Thinking responsibly about Son, T., Weedon, Z., Yigitcanlar, T., Sanchez, T., Corchado, J., & Mehmood, R. (2023).
responsible AI and ‘the dark side’of AI. European Journal of Information Systems, 31 Algorithmic urban planning for smart and sustainable development: Systematic
(3), 257–268. review of the literature. Sustainable Cities and Society, 94, Article 104562.
Mikhaylov, S., Esteve, M., & Campion, A. (2018). Artificial intelligence for the public Stilgoe, J., Owen, R., & Macnaghten, P. (2020). Developing a framework for responsible
sector: Opportunities and challenges of cross-sector collaboration. Philosophical innovation. In The Ethics of Nanotechnology, Geoengineering, and Clean Energy (pp.
Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 347–359). Routledge.
376(2128), Article 20170357. Streaky Bay. (2023). Artificial Intelligence Policy. District council of Streaky Bay.
Milton Keynes. (2020). Milton Keynes and artificial intelligence report. https://www.mi https://www.streakybay.sa.gov.au/__data/assets/pdf_file/0027/1448037/
lton-keynes.gov.uk/sites/default/files/2021-12/MK%20Council%20AI%20Report% DCSB-IFM-13.08-Artificial-Intelligence-Policy.pdf.
20Final%2017062020%20Compressed_compressed%20%281%29%20%281%29. Taeihagh, A. (2021). Governance of artificial intelligence. Policy and Society, 40(2),
pdf. 137–157.
Nabavi, E., & Browne, C. (2023). Leverage zones in responsible AI: Towards a systems Taira, B., Kreger, V., Orue, A., & Diamond, L. (2021). A pragmatic assessment of google
thinking conceptualization. Humanities and Social Sciences Communications, 10(1), translate for emergency department instructions. Journal of General Internal Medicine,
1–9. 36(11), 3361–3365.
Nanjing Municipality. (2022). Nanjing’s "14th five-year" high-tech zone development Tempe City. (2023). Ethical artificial intelligence (AI) policy. Tempe city council. htt
plan. http://www.nanjing.gov.cn/xxgkn/zt/ghxxgk_70036/ssw/qygh_70040/20220 ps://tempe.hylandcloud.com/AgendaOnline/Documents/ViewDocument/ETHICAL
3/t20220311_3316741.html. %20ARTIFICAL%20INTELLIGENCE%20POLICY.DOCX.pdf?meetingId=1451&doc
New York City. (2024). IT policy: Acceptable use of artificial intelligence technologies. NYS umentType=Agenda&itemId=5692&publishId=9354&isSection=false.
Office of Information Technology Services. https://its.ny.gov/system/files/docume Valle-Cruz, D., Fernandez-Cortez, V., & Gil-Garcia, J. (2021). From E-budgeting to smart
nts/2024/01/nys-p24-001-acceptable-use-of-artificial-intelligence-technologies-_1. budgeting: Exploring the potential of artificial intelligence in government decision-
pdf. making for resource allocation. Government Information Quarterly, 39(2), Article
Nishant, R., Kennedy, M., & Corbett, J. (2020). Artificial intelligence for sustainability: 101644.
Challenges, opportunities, and a research agenda. International Journal of Information Vienna City. (2019). Artificial intelligence strategy: digital agenda Vienna. https://digit
Management, 53(53), Article 102104. ales.wien.gv.at/wp-content/uploads/sites/47/2020/04/PO19-00224-Strategiepapie
Nordic Municipalities. Nordic municipalities work with artificial intelligence. Nordic r_Ue-en.pdf.
Council of Ministers. https://doi.org/10.6027/no2019-062. Vinuesa, R., Azizpour, H., Leite, I., Balaam, M., Dignum, V., Domisch, S., et al. (2020).
Ntoutsi, E., Fafalios, P., Gadiraju, U., Iosifidis, V., Nejdl, W., Vidal, M., et al. (2020). Bias The role of artificial intelligence in achieving the sustainable development goals.
in data-driven artificial intelligence systems—An introductory survey. Wiley Nature Communications, 11(1), 1–10.
Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 10(3), Article e1356. Vogl, T., Seidelin, C., Ganesh, B., & Bright, J. (2020). Smart technology and the
O’Donnell, O. (2012). Strategic collaboration in local government a review of emergence of algorithmic bureaucracy: Artificial intelligence in UK local authorities.
international examples of strategic collaboration in local government. https://www. Public Administration Review, 80(6), 946–961.
ipa.ie/_fileUpload/Documents/StrategicCollaboration.pdf. Volmer, E. (2021). Artificial intelligence in local governmental agencies. http://resolver.
OECD. (2020). OECD Public Integrity Handbook. https://doi.org/10.1787/ac8ed8e8-en tudelft.nl/uuid:e0937bfd-cf55-47b6-aa6e-21cf67602490.
OECD.AI. (2023). Will businesses or laws and regulations ever prioritise environmental Wang, Y., Zhang, N., & Zhao, X. (2022). Understanding the determinants in the different
sustainability for AI systems? Oecd.ai. Retrieved September 13, 2023, from https:// government AI adoption stages: Evidence of local government chatbots in China.
oecd.ai/en/wonk/businesses-regulations-environmental-sustainability. Social Science Computer Review, 40(2), 534–554.
Phillimore, J., Fenna, A., Pandey, K., Drage, J., Osaghae, E., Powell, D., & Walker, A. Warm, D. (2011). Local government collaboration for a new decade: Risk, trust, and
(2022). Perspectives on local government’s place in federal systems and central-local effectiveness. State and Local Government Review, 43(1), 60–65.
relations. Commonwealth Journal of Local Governance, 27, 125–157. Wirtz, B., Weyerer, J., & Geyer, C. (2018). Artificial intelligence and the public
Regona, M., Yigitcanlar, T., Xia, B., & Li, R. (2022). Opportunities and adoption sector—Applications and challenges. International Journal of Public Administration, 42
challenges of AI in the construction industry: A PRISMA review. Journal of Open (7), 596–615.
Innovation: Technology, Market, and Complexity, 8(1), 45. Wrocław Metropolitan Area. (2021). Artificial intelligence for a good life. https://www.
Roberson, T., Bornstein, S., Liivoja, R., Ng, S., Scholz, J., & Devitt, K. (2022). A method eurometrex.org/wp-content/uploads/2021/11/publication_AI_2021_ENG.pdf.
for ethical AI in defence: A case study on developing trustworthy autonomous Wu, W., & Liu, S. (2023). Dilemma of the artificial intelligence regulatory landscape.
systems. Journal of Responsible Technology, 11, Article 100036. Commu- nications of the ACM, 66(9), 28–31.
Robinson, S. C. (2020). Trust, transparency, and openness: How inclusion of cultural Yigitcanlar, T. (2006). Australian local governments’ practice and prospects with online
values shapes Nordic national public policy strategies for artificial intelligence (AI). planning. URISA Journal, 18(2), 7–17.
Technology in Society, 63, Article 101421. Yigitcanlar, T. (2021). Greening the artificial intelligence for a sustainable planet: An
Ryan, M., & Stahl, B. (2020). Artificial intelligence ethics guidelines for developers and editorial commentary. Sustainability, 13(24), Article 13508.
users: Clarifying their content and normative implications. Journal of Information, Yigitcanlar, T., Sabatini-Marques, J., Da-Costa, E., Kamruzzaman, M., & Ioppolo, G.
Communication and Ethics in Society, 19(1), 67–86. (2019). Stimulating technological innovation through incentives: Perceptions of
Santoni de Sio, F., & Mecacci, G. (2021). Four responsibility gaps with artificial Australian and Brazilian firms. Technological Forecasting and Social Change, 146,
intelligence: Why they matter and how to address them. Philosophy & Technology, 34 403–412.
(4), 1057–1084. Yigitcanlar, T., Corchado, J., Mehmood, R., Li, R., Mossberger, K., & Desouza, K. (2021).
Scherer, A. G., & Voegtlin, C. (2020). Corporate governance for responsible innovation: Responsible urban innovation with local government artificial intelligence (AI): A
Approaches to corporate governance and their implications for sustainable conceptual framework and research agenda. Journal of Open Innovation: Technology,
development. Academy of Management Perspectives, 34(2), 182–208. Market, and Complexity, 7(1), 71.
Shanghai Municipality. (2022). Regulations for the promotion of the development of the Yigitcanlar, T., Mehmood, R., & Corchado, J. (2021). Green artificial intelligence:
artificial intelligence industry in Shanghai Municipality. https://cset.georgetown. Towards an efficient, sustainable and equitable technology for smart cities and
edu/publication/regulations-for-the-promotion-of-the-development-of-the-artificial futures. Sustainability, 13(16), 8952.
-intelligence-industry-in-shanghai-municipality/. Yigitcanlar, T., Degirmenci, K., Butler, L., & Desouza, K. C. (2022). What are the key
Shenzhen Municipality. (2019). Shenzhen new generation artificial intelligence factors affecting smart city transformation readiness? Evidence from Australian
development action plan (2019–2023). https://chinai.substack.com/p/chinai cities. Cities, 120, Article 103434.
-193-all-chinai-policy-is-local. Zhengzhou Municipality. (2021). Zhengzhou new generation artificial intelligence
Shneiderman, B. (2020). Bridging the gap between ethics and practice. ACM Transactions industry development plan (2021–2025). https://www.ahchanye.com/zc/4357.
on Interactive Intelligent Systems, 10(4), 1–31. html.
Silva, L.d., Bitencourt, C., Faccin, K., & Iakovleva, T. (2019). The role of stakeholders in
the context of responsible innovation: A meta-synthesis. Sustainability, 11(6), 1766.
18

Understanding Local Government Responsible AI Strategy 1729374120

Uploaded by

Copyright:

Available Formats

Understanding Local Government Responsible AI Strategy 1729374120

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Understanding Local Government Responsible AI Strategy 1729374120

Uploaded by

Copyright:

Available Formats

Cities 155 (2024) 105502

Contents lists available at ScienceDirect

Understanding local government responsible AI strategy: An international

Table 1 Table 1 (continued )

Table 1 (continued ) (Fig. 1).

Fig. 1. PRISMA selection procedure of the strategy documents.

4. Results and analysis

Fig. 2. Hierarchy of nodes and sub-nodes.

Fig. 3. Responsible AI strategy concept map.

Lebanon City report stated that: 4.5. AI strategy on accessibility

You might also like