“Voice-First Interfaces in a GUI-First Design World”: Barriers and

Opportunities to Supporting VUI Designers On-the-Job

CHRISTINE MURAD, University of Toronto, Canada
HUMAIRA TASNIM, University of Toronto, Canada
COSMIN MUNTEANU∗ , University of Waterloo, Canada
Voice user interfaces (VUIs) are currently experiencing rapid growth as voice interaction into existing applications or services (banking apps,
commercial devices like Google Home, Amazon Echo, and Apple Homepod travel booking, etc.). However, industry designers have often been
are adopted by users. However, due to the pace of this growth, the tech trained primarily in general UI design - which is heavily comprised
industry has had to adapt quickly and vigorously to keep up with demand. of GUI design - since most of the training that designers receive
Due to this, we currently have limited understanding of the environment (particularly in academic programs) teaches established methods
of VUI design in industry, including the various multitude of practices and
and practices for general UI or GUI design [21, 22, 26]. The problem
tools that are used. We also have a limited understanding of the barriers VUI
designers currently still face. To address such knowledge gaps, we conducted is that designing GUIs is markedly different than designing for VUIs
a large-scale online survey to explore the design practices employed by VUI [37, 44]. Therefore, current design training may not directly transfer
industry designers on-the-job, and the barriers and needs of VUI designers. over when designing for voice interaction. VUI design training in
We found that despite the availability of a wide range of guidelines, textbooks, academia is currently limited, particularly in HCI programs [22].
tools, etc, there are significant gaps in the adoption of these tools within While courses are beginning to appear in industry that aim to teach
VUI industry design, and that designers rely on their previous experience in VUI design, they are still in their infancy [21]. This leaves many
developing GUIs when designing VUIs. Based on our survey findings, we companies to employ their own internal training resources – which
provide recommendations for how the HCI community may direct research may not be sustainable or feasible for smaller companies, indepen-
efforts in developing tools to assist designers in overcoming existing barriers dent design shops, or freelance designers. The lack of established
and build usable and adoptable VUIs.
tools and resources for VUI designers also adds to the difficulty of
CCS Concepts: • Human-centered computing → Human computer interac- designing for VUIs in industry. General UI design uses established
tion (HCI); HCI design and evaluation methods; Interaction design; Interaction and well-known guidelines to ensure UIs are usable, but those do
design process and methods; not map directly over to VUI design [26, 37, 44].
Additional Key Words and Phrases: Voice user interfaces, Design, User expe- Altogether, this makes it difficult for designers that are either
rience design, Speech interfaces, Design guidelines transitioning from general UI to VUI design or are entering VUI
ACM Reference Format: design in industry for the first time. Research suggests that VUI
Christine Murad, Humaira Tasnim, and Cosmin Munteanu. 2022. “Voice- designers - that have largely been GUI-trained - use their experience
First Interfaces in a GUI-First Design World”: Barriers and Opportunities to from GUI design and training when transitioning to VUI design
Supporting VUI Designers On-the-Job. In 4th Conference on Conversational [26]. However, there is currently a limited understanding of the
User Interfaces (CUI 2022), July 26–28, 2022, Glasgow, United Kingdom. ACM, experiences and practices of current VUI industry designers – in
New York, NY, USA, 10 pages. https://doi.org/10.1145/3543829.3543842 particular, how previous training, experience, and available design
tools affect how industry designers conduct VUI design. There is
1 INTRODUCTION also a limited understanding of the current barriers and needs that
Voice user interfaces (VUIs) are becoming more prevalent in people’s industry designers face when ideating, designing, and testing new
daily lives, with an increasing pace of adoption through devices VUIs.
such as phones, smart speakers, cars, and even household appliances.
Research has shown that users’ perceptions of VUIs have become
1.1 Research Questions and Contributions
more favourable within the past five years, and have noted the
growing use cases for using voice as a form of interaction – from One of the main purposes of conducting this survey was to gather
allowing for hands-free interaction [11], communicating naturally an understanding of experience and practices in VUI industry. There
with technology [8], or even having a robot AI friend [18]. are little to no studies that aim to create a picture of the experience
As the pace of this adoption grows, industry designers are often of VUI designers in industry. We believe that having this picture
tasked to design voice user interfaces, either by creating a novel will allow HCI experts and designers to identify the main barriers
interface (e.g., the creation of a new smart speaker) or incorporating to focus on. This leads us to the first research question:

∗ Work partially conducted while at: University of Toronto Mississauga, Canada RQ1. What are the current experiences and practices that
VUI industry designers employ on the job when ideating, de-
CUI 2022, July 26–28, 2022, Glasgow, United Kingdom
© 2022 Association for Computing Machinery.
signing, and testing VUIs for usability?
This is the author’s version of the work. It is posted here for your personal use. Not
for redistribution. The definitive Version of Record was published in 4th Conference on Along with this, we wanted to understand how designers map
Conversational User Interfaces (CUI 2022), July 26–28, 2022, Glasgow, United Kingdom,
https://doi.org/10.1145/3543829.3543842. their experience and practices from general UI to VUI design. Previ-
ous papers show that, while GUI heuristics and patterns cannot be

1
CUI 2022, July 26–28, 2022, Glasgow, United Kingdom Murad and Munteanu

easily mapped to VUI design [37, 44] existing GUI heuristics may Furthermore, books such as “Designing Voice User Interfaces: Prin-
prove to be a good base for developing new VUI-specific heuris- ciples of Conversational Experiences” [31] by Cathy Pearl have
tics [23, 26, 42]. Some research encourages using heuristics like worked to develop of set of industry-adoptable guidelines and prac-
Nielsen’s as a framework to base new VUI heuristics around [36]. tices for Voice User Interfaces, along with guidelines from companies
Even further, previous research has suggested that using experience like Amazon [6], Google [5], and Apple [7]. Work on tools for proto-
and concepts from GUI heuristics can prove useful in the adoption typing voice user interfaces and testing said prototypes have been
of new VUI-specific heuristics [26]. This leads to our next question: growing both in academic research and in industry. Industry tools
such as Google’s Dialogflow [1] and Voiceflow [4] allow designers to
RQ2. Do designers use previous experience and design prac- easily create conversational flows for a number of different activities
tices from GUI design when they are designing for VUIs? in a user-friendly manner. Recently, Porcheron et al. [33] developed
a python tool to allow for Wizard of Oz prototype testing with Voice
We anticipate that designers map some their practices and their User Interface.
experience in general UI design to VUI design, as previous research Among the many proposals of tools and resources across both
suggests [26]. However, due to the different nature of VUI design academia and industry, many have little to no validation, and very
compared to GUI design, we also wanted to identify the current few have been adopted in industry [25]. Tools are still being devel-
barriers and potential needs that VUI industry designers have. Con- oped and are in their infancy. We therefore chose not to ask about
tinuing issues with current commercial VUIs [11, 17] suggest this. specific tools and books developed for VUI design due to the lack
This leads to our final research question: of validation of the efficacy of these tools and would require an
extensive scoping of all existing tools and books about VUI design,
RQ3. What are the barriers and needs of current VUI indus- which is outside the scope of this paper. We focus on exploring
try designers? general principles and prototyping methods in VUI design indus-
try, bringing to light where sources of struggle may be, and where
further research needs to be directed. We have chosen only to ask
2 LITERATURE REVIEW
about guidelines for VUIs from the primary companies that develop
This section provides a background of the research currently con- commercial voice-first devices (Amazon, Google, Apple, IBM, and
ducted in the areas of Voice User Interface usability, design, and Microsoft), as these companies claim their guidelines are derived
training, in order to provide context on the problems we explore in from and implemented within the devices that they develop.
this paper.
2.3 Voice User Interface Design Training
2.1 Usability of Voice User Interfaces Currently, there is a limited amount of research that explores how
Previous research has explored the usability of VUIs, particularly to train designers in VUI design. Research has begun to explore
in the last 5-10 years. Infrequent users of personal voice assistants VUI design training in HCI Education [22]. Even so, previous re-
encounter numerous usability challenges [11], including difficul- search shows that there is very little discussion and training on
ties recalling information [38], system feedback [11, 17], recogni- VUIs in current HCI curricula [22], making it difficult for designers
tion errors [11] and learnability [13, 27, 44]. Many difficulties are a to transition from the familiar space of designing general UIs to
consequence of users’ expectations of VUI interaction versus the designing VUIs. We can see this in present VUI designs and their
reality of using existing commercial VUIs [17]. Nielsen and Shnei- serious usability issues [10, 11, 23].
derman also commented on the usability issues in speech interfaces. In comparison, much effort has been dedicated to developing
Shneiderman [38] highlighted key issues such as the cognitive ef- better methods for teaching general UI design in post-secondary
fort required for interaction, problems with presenting information education – particularly in Human-Computer Interaction [39]. HCI
through speech, and resolving errors due to speech recognition. conferences, workshops and symposiums have been held to discuss
Nielsen [29] and colleagues [9] note that the usability of VUIs limits the methods to better teach HCI methods in their classrooms [2, 3].
their usefulness, and that visual interfaces surpass voice interfaces Topics range from gaps in current HCI teaching [43] to teaching
in usability for most cases (besides hands-free interaction or for HCI to non-computing disciplines [19]. Several papers also discuss
people with physical disabilities). Even with research and techno- educators’ practical efforts to develop a curriculum for teaching HCI
logical advances over the past decade, users continue to experience design methods, from design studio courses [14, 34] to experiential
usability issues with VUIs, often leading to a lack of adoption or learning [30].
abandonment [11, 17, 20].
3 METHODS
2.2 Resources and Tools for Voice User Interface Design To answer our research questions, we collected and analyzed data
Many efforts have been made to develop resources for designing through a large-scale online survey about the practices currently
Voice User Interfaces. One that has increased over the past 5 years employed by VUI designers in industry, along with the needs and
is the develop of design guidelines or heuristics for VUIs. Several barriers they face when ideating, designing, and testing new VUIs.
sets of VUI heuristics have been developed in previous research The purpose of large-scale survey research has been described as “to
[15, 23, 35, 40, 41], and have received varying levels of validation. obtain information describing the characteristics of a large sample

2
“Voice-First Interfaces in a GUI-First Design World”: Barriers and Opportunities to Supporting VUI Designers On-the-Job CUI 2022, July 26–28, 2022, Glasgow, United Kingdom

of individuals of interest relatively quickly” [32]. It has furthermore boards or groups on platforms like Reddit and Facebook. Participants
been described as a way to “gather information about people’s habits, were asked to complete a short online enrollment form identifying
interaction with technology, or behaviour” when applied in the HCI their interest, with a CAPTCHA to help prevent bot responses, and
space [28]. We aimed with our survey to identify characteristics of screening questions (not included in the analysis) asking respon-
VUI industry through their experiences, practices, and perceptions dents to describe their experience in UX design. Each response was
of VUI design. We believe conducting a survey was the most effective verified by one of the authors for duplicate or invalid responses (e.g.,
way to gather this information to understand these areas. copied from Googleable job design descriptions) before the survey
link was emailed to them. Survey data was anonymized by partici-
3.1 Survey Design pant ID, and not associated to initial enrollment or compensation
We designed a multiple-choice questionnaire, distributed as an on- forms.
line survey hosted on Alchemer. The questionnaire was derived
from priorly validated research instruments on digital accessibil- 3.3 Survey Analysis
ity awareness and practices, by Lazar et al. [16] and Freire et al. Responses were collated and reviewed to ensure completion and
[12], adjusting the application space to “VUI design”, instead of consistency and to identify duplicates or outliers. Responses to
“accessible design”. We opted for this approach on consideration open-ended questions were reviewed for quality by checking if the
that designing for VUIs may be at the same stage of awareness and answers provided were relevant to the questions asked, in order to
practice as designing for accessibility was when Lazar et al. [16] avoid fraudulent responses [1].
conducted their seminal research on this topic. We ground the VUI- Data was then processed and coded through SPSS. Descriptive
specific questions in papers such as [22, 24, 26, 31], and draw from statistics were employed to analyze the quantitative data provided
the authors’ combined 30+ years of research and industry-based through the multiple-choice questions. Frequency analysis was per-
VUI design experience. formed on all non-open-ended questions, and percentage was re-
The questionnaire consisted of 54 questions, grouped into seven ported - with both relative percentage and percentage of cases
sections: (responses) being reported for multiple-response questions. We dis-
A General Demographics: nine questions about participants’ cuss the percentage of responses where relevant, rather than the
gender, age, education level, work experience, etc. relative percentage (adding up to 100%), as it best illustrates the
B General UI Design Experience: five questions about par- percentage of times a particular response was chosen and assists
ticipants’ general design experience and interfaces they are our analysis where we aim to identify the most common practices
currently or have previously designed. and barriers experienced in VUI design (participants could choose
C Voice User Interface Design Experience: seven ques- several options).
tions which asked pa experience with VUIs in particular. Spearman’s correlation (with two-tailed significance testing) was
D General UI Design Practices: eleven questions (one open- also calculated across responses from equivalent questions from
ended) which asked about the ideating, prototyping, and us- the “General User Interface Design” sections and the “Voice User
ability practices used in general UI design, along with their Interface Design” sections, to explore correlations that exist across
familiarity and usage and design tools/resources. the two different design spaces. For example, correlation would be
E VUI Design Practices: eleven questions, a duplicate of calculated across responses to “What do you do when you are first
Section D except for asking practices and tools used for VUI starting the development of a new interface?” and “What do you do
design when you are first starting the development of a new voice inter-
F Perceptions and Views of General UI Design: n: seven face?”. Spearman’s correlation was used due to the non-parametric
questions asking how general usability is perceived in their nature of the multiple-choice questions in the survey (categorical
companies and what influences usability. It also asks about values with non-normal distribution). These correlations are primar-
the challenges, barriers, and needs (tools-wise) of designers ily used to analyze how designers potentially map their experience
in general UI design. and practices from UI design over to VUI design, and whether par-
G Perceptions and Views of VUI Design: seven questions, ticular barriers are unique to VUI design or are also present in UI
a duplicate of Section F except for VUI design design.
The full survey is included in the supplementary materials. As
most (excluding demographics) were multiple-response questions, 3.4 Participants
response percentages do not add to 100. Only three open ended ques- 105 participants completed the survey. Participants were required
tions were included in the survey – one verifying human responses, to have experience in UX design – this was based on a self-reported
and two optional questions on why general or VUI usability was open-ended question at the beginning of the survey which asked
not considered by participants, which were rarely filled in. participants to briefly describe the past one or two design projects
they had taken part in. Participants were compensated with an Ama-
3.2 Recruitment zon gift card ($15) on completion of the survey, that was delivered
Participants were recruited through snowball sampling through the to them through email.
author’s professional network within industry, as well as through ad- Participants consisted of both VUI-specific designers and gen-
vertising the survey by making posts on online designer discussion eral UI designers, in order to compare the practices, experiences,

3
CUI 2022, July 26–28, 2022, Glasgow, United Kingdom Murad and Munteanu

Table 1. Demographics Table of Participants (n=105) 4 RESULTS: VUI DESIGNERS’ PRACTICES AND
BARRIERS
Variables Responses n Rel. % We synthesize here the survey findings, to answer RQ1 and RQ3. We
Female 48 45.7 report both the descriptive statistics and frequency analysis of sur-
Male 54 51.4 vey responses. As mentioned in Section 3.3, we discuss percentage
Gender of responses vs. relative percentage, due to questions being primar-
Non-Binary/Genderqueer 1 0.9
Choose Not To Answer 2 2 ily multiple-choice. We group the findings along the two aspects of
VUI design we aimed to explore – practices employed in VUI design
Canada 26 24.7
(and how they may map from GUI design), and barriers and needs
USA 56 53.3
in VUI design.
United Kingdom 11 10.4
Spain 1 0.9
Geographic Location
Italy 1 0.9 4.1 Practices Employed by VUI Designers in Industry
Germany 8 7
First, we provide a synthesis of the current practices employed
France 1 0.9
by VUI industry designers – particularly in terms of the ideation,
Australia 2 1.9
prototyping, and usability testing phases of VUI design, along with
Academia 15 14.3 the resources used during all these phases of design. In this section,
Industry Researcher 2 1.9 we first report percentage of responses for the entire population
UX/UI Designer 59 56.2 set (both VUI and general UI designers) for questions about general
Occupation Usability Professional 1 1.0 UI design (n=105). We then control for and report percentage of
Student 12 11.4 responses for the subset of designers who have conducted VUI
Other 15 14.3 design (n=54), when for questions about VUI design.
No occupation 1 1.0
High School Diploma 7 6.7 4.1.1 Design Ideation and Development of Voice User Interfaces.
College Diploma 11 10.5 When asked what designers did when developing a new UI, the top
Bachelor’s Degree 50 47.6 responses were referring to existing UIs (78.8%), referring to online
Education
Master’s Degree 26 24.8 resources (76%), and speaking to colleagues (59.6%). In comparison,
Doctor of Philosophy 6 5.7 when those who had conducted VUI design were asked when they
Other 5 4.8 did when developing a new VUI, the top responses were the same:
referring to existing VUIs (71.1%), referring to online resources
(71.1%), and speaking to colleagues (66.7%). 35% reported haven’t
being involved in the beginning of a new VUI, while no one reported
to not have been involved with the beginning of development for a
general UI.
and challenges of both VUI-specific and general design. While we
When asked how designers ensure general UI usability, the top
consider a VUI designer someone who has been involved in any
responses were interviews (78.8%) and user tests (71.2%) – however,
phase of the design & creation process (e.g. ideation, prototyping,
questionnaire/surveys (63.5%), use cases (62.5%), and usability tests
research, evaluation), we did not place artificial constraints on this
(67.3%) were close behind. In comparison, when designers who had
definition given the broad range of designation in industry. Instead,
conducted VUI design were asked how they ensure VUI usability,
we allowed for respondents’ self-identification through the survey.
the top responses remained the same, but at a lower frequency:
We also understand that many VUI designers may have been pri-
interviews (56.9%), questionnaires/surveys (47.1%) and user tests
marily GUI-trained, who transitioned to VUI design or continued in
(51%). 13.7% of respondents who conduct VUI design selected they
both spaces, due to increasing market demand for VUIs. Therefore,
do not use any research methods in initial VUI design, in comparison
we wanted to explore how established UI practices may have/have
to the 2.9% of all participants who selected that they do not use any
not transferred over to VUI design.
for initial UI design. This suggests that designers are more likely to
conduct usability research during the ideation phase for general UIs
3.4.1 Demographics and Design Experience. We present partici- vs. for VUIs.
pants’ demographics in Table 1, and design experience in Table When asked how usability is perceived in one’s company or
2. Ages of respondents ranged between 19 and 54, and the mean practice, on a 5-point Likert scale from “Not Important” to “Highly
age was 31.44. 51.4% (n=54) of participants had stated that they had Important”, 78.1% of all participants selected general UI usability
previously worked on designing a voice application or interface. being perceived as “Fairly important” or higher by their company. In
comparison, 52.9% of respondents who conduct VUI design selected
3.4.2 VUI Usage. We illustrate the familiarity and usage of VUIs VUI usability being perceived as “Fairly Important” or higher by
by participants in Table 3, and the types of interfaces currently and their company. This suggests that usability is more highly valued
previously designed in Table 4. by companies for general UI design than it is for VUI design.

4
“Voice-First Interfaces in a GUI-First Design World”: Barriers and Opportunities to Supporting VUI Designers On-the-Job CUI 2022, July 26–28, 2022, Glasgow, United Kingdom

Table 2. Design Experience (n=105)

Variables Responses n Rel. % % of responses

HCI/Human Factors 105 43.2 100.0
Computer Science 38 15.6 36.2
Psychology, Behavioural Science, or Cognitive Science 20 8.2 19.0
Type of design training Science and Engineering 16 6.6 5.2
Media or Communication 33 13.6 31.4
Other 20 8.2 19.0
No design training 11 4.5 10.5
Academia 77 32.1 73.3
Professional Industry training 38 15.8 36.2
Self-trained (textbooks/online resources) 58 24.2 55.2
Medium of design training
Learn through experience 58 24.2 55.2
Other 6 2.5 5.7
Not applicable 3 1.3 2.9
Undergraduate courses 10 7.6 9.6
Graduate Courses 13 9.8 12.5
Online Courses 19 14.4 18.3
Method of VUI design training
Training in the workplace 29 22.0 27.9
Other 9 6.8 8.7
None 52 39.4 50.0

Table 3. VUI Usage (n=105)

Variables Responses n Rel. % % of responses

Smart speakers 91 22.3 87.5
Intelligent Personal Assistants in phones 98 24.0 94.2
Automotive voice interfaces 58 14.2 55.8
Voice-controlled appliances/internet-of-things 46 11.3 44.2
VUI application familiarity
Speech-to-text applications 68 16.7 65.4
Voice accessibility control 41 10.0 39.4
Other 5 1.2 4.8
None 1 0.2 1.0
Amazon Alexa/Echo 42 23.9 40.0
Google Assistant/Home 50 28.4 47.6
Siri/Apple Homepod 47 26.7 44.8
VUIs used in daily life Microsoft Cortana 8 4.5 7.6
Samsung Bixby 6 3.4 5.7
Other 4 2.3 3.8
None 19 10.8 18.1
Never 39 37.1 37.1
1-2 times a day 27 25.7 25.7
Amount of VUI usage 3-5 times a day 10 9.5 9.5
6-10 times a day 9 8.6 8.6
+10 times a day 20 19.0 19.0

4.1.2 Tools and Resources Used for Voice User Interface Design. they refer to when developing VUIs, the top responses were the
When asked what resources designers refer to when developing gen- same, but at a lower rate: personal professional experience (59.3%)
eral UIs, the top responses were personal professional experience and general standards/norms (59.3%). This may suggest that design-
(80%) and standards/norms (81.9%). In comparison, when partici- ers feel more confident in their personal experience and established
pants who had conducted VUI design were asked what resources standards for general UI design than for VUI design.

5
CUI 2022, July 26–28, 2022, Glasgow, United Kingdom Murad and Munteanu

Table 4. Types of Interfaces Designed (n=105)

Previously Rel. % of Currently Rel. % of

Variables Responses
Designed % Responses Designed % Responses
Web-based applications 87 21.4 83.7 85 22.4 82.5
Business applications 46 11.3 44.2 43 11.3 41.7
Forms and documents 56 13.8 53.8 36 9.5 35.0
Big complex systems 19 4.7 18.3 36 6.8 25.2
Mobile ICT 33 8.1 31.7 34 8.9 33.0
Types of interfaces
Kiosk displays 21 5.2 20.2 15 3.9 14.6
designed at company/
Consumer software 44 10.8 42.3 35 9.2 34.0
independent practice
Video game software 19 4.7 18.3 14 3.7 13.6
Physical products w/ digital capabilities 21 5.2 20.2 18 4.7 17.5
Voice user interfaces 32 7.9 30.8 35 9.2 34.0
Chatbots 22 5.4 21.2 29 7.6 28.2
Other 7 1.7 6.7 10 2.6 9.7
Voice dictation/speech-to-text/voice keyboard 16 11.9 15.5 19 13.8 18.1
Incorporating voice interaction into existing 34 25.2 33.0 38 27.5 36.2
virtual application (website, phone app, etc.)
Types of VUI interfaces
Incorporating voice interaction into existing 15 11.1 14.6 15 10.9 14.3
currently designed at
physical application (microwave, thermostat, etc)
company/independent
Voice in automobiles 6 4.4 5.8 8 5.8 7.6
Other 15 11.1 14.6 12 8.7 11.4
None 49 36.3 47.6 46 33.3 43.8

In terms of established written guidelines and patterns, 68.8% of guidelines will tend to use them, for both general UI design and VUI
all respondents selected “heuristics and guidelines” as a tool used for design
general UI design, compared to the 44.4% of respondents who con-
duct VUI design selected it as a tool used for VUI design. Similarly, 4.1.3 Prototyping and Evaluation of Voice User Interfaces. When
76.2% of all respondents selected “design patterns” as a method used asked about the most commonly used evaluation techniques for
for general UI design, compared to the 40.7% of respondents who general UIs, the top responses were conducting usability tests with
conduct VUI design that selected it for VUI design. When evaluating experienced UI users (65.4%) and with novice UI users (79.8%). In
the usability of interfaces, 59.6% of all respondents selected that comparison, when respondents who conduct VUI design were asked
they check compliance with design guidelines for general UI design, about the most commonly used evaluation techniques for VUIs, the
compared to the 32.1% of respondents who conduct VUI design who top responses remained the same, but at a lower rate: conducting
stated that they check compliance with design guidelines for VUI usability tests with experienced VUI users (56.6%) and with novice
design. Across the board, the percentage of responses noting the VUI users (75.5%). However, 13.2% of respondents who conduct VUI
usage of general UI design guidelines is higher than the percentage design responded that they do not evaluate their VUIs for usability,
of responses for VUI guidelines. This may suggest a level of confi- while only 2.9% of all respondents stated they do not evaluate their
dence in the development and validation of design guidelines and GUIs for usability.
patterns for general UI design vs. for VUI design. When asked what prototyping methods were used for general UI
When asked about industry-established guideline familiarity and design, the top responses were mock-ups (76%), sketches (68.3%),
awareness, 24.3% claim to not use industry-established guidelines and wireframes (75%). In comparison, when respondents who have
(such as those from Nielsen, Norman, and Shneiderman) while de- conducted VUI design were asked what prototyping methods were
signing general UIs in practice, with 15.4% that stated that they used for VUI design, the top responses remained the same: mock-ups
were not aware of such design guidelines. In comparison, 26.9% (59.2%), sketches (38.8%), and wireframes (44.9%), but yet again at a
of respondents who conduct VUI design stated they do not use lower rate. This suggests a general trend towards not prototyping
industry-established VUI design guidelines (such as those from com- VUIs as frequently or as conventionally as general UIs.
panies like Google, Siri, Amazon, etc) in practice, with 18.5% that When asked what evaluation techniques were used for testing
stated they were not aware of such guidelines. The difference be- intermediate UI prototypes, the top responses were think-alouds
tween those who do not use industry-established guidelines and (71.8%), cognitive walkthroughs (54.4%) and A/B testing (51.5%).
those who aren’t aware of them are similar across the two design In comparison, when respondents who conduct VUI design were
spaces. This suggests that those familiar with industry-established asked what evaluation methods were used for testing intermediate
VUI prototypes, the top responses remained the same: think aloud

6
“Voice-First Interfaces in a GUI-First Design World”: Barriers and Opportunities to Supporting VUI Designers On-the-Job CUI 2022, July 26–28, 2022, Glasgow, United Kingdom

(55.1%), cognitive walkthroughs (32.7%), and A/B testing (53.1%). both general UI design and VUI design. 13.5% noted that they do not
While the same types of methods are used across general UI and need new tools or resources for general UI design, while only 6%
VUI design, they are used at a much lower rate for VUI design noted the same for VUI design. These results may suggest that a lack
(besides A/B testing). Along with that, no particular evaluation of tools, resources, and training for VUI design may be contributing
method stands out for VUI design as it does for GUI design. This to existing usability issues in commercial VUIs.
suggests that the tools currently used for VUI design prototyping
and evaluation may not be adequate, and we have yet to develop 5 RESULTS: COMPARING VUI DESIGNERS’ TO
specific ones for VUI design. GENERAL UI DESIGNERS’ PRACTICES AND BARRIERS
Here, we performed correlation analysis and report significant cor-
4.2 Barriers and Needs of VUI Industry Designers relations found from conducting Spearman’s correlation across re-
Second, we describe the barriers and needs of VUI industry design- sponses from equivalent questions in the survey across the “General
ers as reported through the survey. In this section, we report on User Interface design” sections and “Voice User Interface design”
the percentages of all respondents (VUI and non-VUI designers) questions (as described in Section 3.1).
when discussing both general UI and VUI design. As many non-VUI The purpose of performing this analysis was to answer RQ2: Do
designers are quite aware of VUIs and may even have colleagues designers use previous experience and design practices from GUI design
who conduct VUI design, it is important to capture their perception when they are designing for VUIs? Performing correlation analysis
on barriers to VUI design, as well as tools and resources required for across responses for equivalent questions for general UI and VUI
VUIs, and whether they would be prepared if they were to transition design can provide insight on whether designers perform similar
to this field. practices from GUI to VUI design, and whether any experience from
general UI design is mapped over to VUI design. This analysis also
4.2.1 Challenges and Barriers to Voice User Interface Design. The helps answer RQ3. What are the barriers and needs of current VUI
largest challenge in VUI design found by respondents was speech industry designers?, as it helps to provide us insight on whether the
recognition accuracy (69.5%) and making interaction natural/ con- same barriers are seen across general UI and VUI design.
versational (64.8%). Creating appropriate voice commands (55.2%),
helping users recognize and correct errors (50.5%) and security and
privacy (54.3%) were close behind. According to participants, the
5.1 Practices Employed by VUI Designers in Industry
largest barriers to designing VUIs are budget restrictions (50.5%) and Spearman’s correlation found significant correlations between re-
project scope (54.6%). Other logistical barriers with high response spondents who refer to textual resources when designing a new UI
rates are time restrictions (44.3%) and technical constraints (43.3%). and those who do when designing a new VUI (rs =0.388, p < .001).
We were most interested in the design and knowledge barriers There was also a significant correlation between those who refer to
that VUI industry designers currently experience. 40.4% of respon- academic research papers when designing a new UI vs those who
dents who conduct VUI design selected “lack of design knowledge do when designing a new VUI (rs =0.418, p < .001). These mean
between colleagues”, and 42.3% selected “lack of tools/resources that general UI designers who use textual resources or academic
information” each as barriers to VUI design. While a similar per- papers for designing UIs are also likely to use them for VUIs. Less
centage of all respondents noted “lack of design knowledge between significant correlations were found between those who refer to on-
colleagues” as a barrier to GUI design (42.3%), only 27.9% percent of line resources when designing a new general UI vs those who do
all respondents selected “lack of tools/resources/information avail- when designing a new VUI (rs =0.240, p < .05), and those who look
able” as a barrier to GUI design. This may suggest two things: 1) at reference industry UIs vs. those who look at reference industry
The general lack of VUI design guidelines, or knowledge of existing VUIs (rs =0.218 p < .05).
ones, causes it to be a much larger barrier than for general UI design; Across the board, significant correlations were also found be-
or 2) While designers may be familiar with VUI design guidelines, a tween using the same types of methods for ensuring UI usability
lack of confidence in their validity may lead to a lower rate of usage, vs. ensuring VUI usability. Overall, these correlations suggest that
and therefore be identified as a larger barrier. practices and methods that are used in UI design are also likely to
be used in VUI design – perhaps suggesting that designers actively
4.2.2 Tools and Resources Needed for Voice User Interface Design. attempt to map the current practices they use from general UI to
As mentioned earlier, 40.2% of respondents who conduct VUI design VUI design. Spearman’s correlation analysis also showed significant
selected “lack of tools/resources/information” as a barrier for VUI correlations between general UI and VUI designers with guideline
design, while 33% of respondents selected that a lack of familiarity usage (rs =0.351, p < .01), using personal professional experience
with tools and guidelines was a barrier for VUI design. The highest (rs =0.328, p < .01), and using established personas (rs =0.403, p <
requested tool for VUI design by respondents was prototyping tools .01) and scenarios (rs =0.293, p < .01). There was also significant
(70%), with design guidelines (59%) close behind. Respondents noted correlation between those who were aware of general UI guidelines
the need for design training, both academic (56%) and industry (57%) and those who were aware of VUI guidelines (rs =0.245, p < .05). As
training. In comparison, while only 38.5% of all participants noted a in the last section, this further suggests that tools and resources that
need for more general UI design guidelines, prototyping tools (49%) are used in UI design are also likely to be used in VUI design, even
and academic (45.2%) and industry (53.8%) design training were also if at a lower rate. The lower rate may be due to a lack of confidence
requested. This suggests that design training is highly valued in in the validity of tools, as previously mentioned.

7
CUI 2022, July 26–28, 2022, Glasgow, United Kingdom Murad and Munteanu

Furthermore, across the board, Spearman’s correlation calculation as a “lack of design knowledge between colleagues” was also noted
showed that there was a positive correlation (p < 0.01) between one as a large VUI design barrier – so referring to colleague experience
practice used for general UI design and the same practice being may not be a helpful resource, even though designers may be opted
used for VUI design. Significant correlations also appeared (rs > to turn to it when lacking their own VUI design experience. In turn,
0.4, p < 0.01) between prototyping techniques being used for both established tools such as design guidelines or heuristics may be
VUI and general UI design, as well as between prototype evaluation more useful for VUI design at the current time, due to the lack of
techniques being used for both VUI and general UI design (rs > 0.3, personal experience on proper VUI design practices.
p < 0.01). This which suggests that designers use the same methods Validated tools and resources seem to be used less often in VUI
for VUI prototypes that they use for general UI prototypes as well. design than in general UI design. As noted in Section 4.1, across
the board, tools and resources used for ideation, prototyping, and
5.2 Barriers and Needs of VUI Industry Designers evaluation were used at a lower rate for VUI design than for gen-
eral UI design. It’s not clear from the survey responses the reason
There were significant positive correlations (rs > 0.3, p < 0.01) be-
for this, but research shows that VUI design is still relatively new
tween those who found different company logistical constraints
in industry, and tools and resources are either underdeveloped or
(budget restrictions, time restrictions, project scope, lack of client/
not heavily known. As mentioned earlier, previous design experi-
managerial support, and technical constraints) as barriers for general
ence is not always helpful, as general UI and GUI design is very
UIs and as similar barriers for VUIs – with most of the significant
different then VUI design. As well, while research methods such as
correlations being with the same type of constraint across both
questionnaires/surveys, interviews, and user tests are shown to be
design spaces. While this may not be something that the HCI com-
commonly used in VUI design, they are used at a much lower rate
munity can help solve through research, it is important to note, as
than they are for general UIs, as noted in Section 4.1.1. The 13.7% of
it may help companies fuel more resources into their VUI design
respondents who conduct VUI design who do not using research
teams.
methods for VUI design may be due to the lack of development and
As well, across the board, Spearman’s correlation found signifi-
validation of practices for VUI design [26].
cant positive correlations between a tool that designers wish they
In particular, we were interested on how design guidelines are
had for general UIs vs. for VUIs (rs > 0.3, p < 0.01). Strong significant
used by VUI industry designers. We found that design guidelines
correlations were found between those who wanted academic and
were used much more heavily in general UI design (68.6%) than they
industry training for general UIs, vs. those who wanted the same for
are in VUI design (44.4%), as noted in Section 4.1.2. Only 32.1% of
VUIs, and for prototyping tools among the two design spaces. This
VUI design participants noted checking compliance according to
heavily supports the argument that designers currently map the
VUI design guidelines, compared to the 59.6% who do for GUIs. This
practices and resources they use in UI design over to VUI design.
makes sense, there is little research working on validating design
guidelines and heuristics for VUIs [15, 20, 23, 24, 41]. As mentioned
6 DISCUSSION AND TAKEAWAYS - ANSWERING THE earlier, there may be a lack of confidence in the validity of design
RESEARCH QUESTIONS guidelines currently developed for VUIs. This is in addition to previ-
Grounded in the analysis of data from our survey responses, we ous research finding that GUI guidelines can’t be mapped directly to
return to answer the research questions we formulated at the begin- VUI design [24, 37, 44]. However, there is a significant correlation
ning of this paper (Section 1.1). between those who use GUI guidelines and those who also use VUI
guidelines – this may mean that designers are willing to use existing
RQ1. What are the current practices that VUI industry designers design guidelines as a design resource but have a barrier to doing
employ on the job when ideating, designing, and testing VUIs for us- so due to the uniqueness of voice as an interaction technique. This
ability? suggests that more work needs to be done to develop and validate
design heuristics for VUIs and get them into the hands of industry
We found many similarities between the practices used in VUI designers, so that these established practices may be conducted in
design and those used in general UI design. Previous experience VUI design. This leads to our first takeaway:
seems to be relied on more than other practices when designing for
VUIs, particularly in the ideation and development phase. We see Takeaway 1: VUI designers want to use guidelines – but
this through participants referring to existing VUIs (71.1%) and their may not find the available ones adequate, making guideline
fellow colleague’s experience (66.7%) most when in the VUI ideation development an immediate necessity
phase. This is also shown when respondents noted using profes-
sional experience (59.3%) as a design resource more than validated RQ2. Do designers use previous experience and design practices from
tools such as heuristics and guidelines (44.4%). Interestingly, the GUI design when they are designing for VUIs?
difference between those who use professional experience vs heuris-
tics and guidelines is much wider for general UI designers than for Our survey results show that, yes, much of the practices that respon-
VUI designers – this may speak to the lack of personal experience dents noted using for GUI design are the ones that are also used when
that designers have in VUI design. This coincides with designers conducting VUI design. We found significant correlations across the
referring to VUIs that already exist as inspiration, or experience board for many the practices used for GUI design and the same ones
from their own colleagues. However, this also may be problematic,

8
“Voice-First Interfaces in a GUI-First Design World”: Barriers and Opportunities to Supporting VUI Designers On-the-Job CUI 2022, July 26–28, 2022, Glasgow, United Kingdom

used for VUI design, such as methods for ensuring usability, proto- the immediate future.
typing methods, and usability testing methods. This matches our
initial expectations, in that general UI designers use their previous There was a high response rate of “lack of established prototyping
experience when employing new practices for VUI design. However, tools” as a barrier to VUI design, which was a finding that we did not
even so, we found that existing practices are still used much less initially anticipate. Designers said that they currently employ wire-
often for VUI design than they are for general UI design, as noted frames, sketches, and mockups for both GUIs and VUIs, so (digital)
in Section 4.1. This may also coincide with respondents noting that tools that allow them to do this efficiently for voice would be very
usability is perceived by companies as more important in GUI design useful. More interesting is that there are several prototyping tools
than it is for VUI design, along with the 34.7% of participants who that are well-known and largely advertised as useful for industry
do test their VUI prototypes for usability. Our findings show a clear VUI designers, such as Dialogflow [1] and Voiceflow [4], among oth-
focus on ensuring usability for GUIs vs. ensuring usability for VUIs. ers. This may suggest two things: either 1) existing VUI prototyping
This may be, again, due to the lack of validated practices for VUI de- tools are still relatively unknown to current industry designers, or 2)
sign, since it is clear that designers do try to map practices they use they do not satisfy the prototyping requirements that VUI designers
for general UI design to VUI design – just at a much lower frequency. have.

Takeaway 2: Current industry designers rely on their exist- Takeaway 4: VUI industry designers want better digital
ing GUI experience when designing for VUIs, and we should prototyping tools, and current ones are not being adopted,
leverage this experience when in transitioning them from necessitating immediate development and adoption efforts
GUI to VUI design, to improve adoptability. for VUI prototyping tools.

RQ3. What are the barriers and needs of current VUI industry de- 7 CONCLUSION
signers?
In this paper, we conducted a large-scale online survey with 105
designers, exploring the design practices of current VUI industry
Much of the barriers noted by respondents were often logistical
designers, and how existing experience, practices, and tools used in
barriers imposed by their workplace, such as time constraints, bud-
GUI design are mapped or carried over to VUI design. Our study also
get constraints, project scope, and technical constraints, as noted
explored the barriers and needs of current VUI industry designers.
in Section 4.2.1. These are not barriers that the HCI community
We found that VUI designers do use their previous experience in
can directly address now and will vary from company to company.
GUI design and employ it when working on VUI design, however
However, it was noted that a lack of design knowledge and lack of
several barriers are experienced while doing so. Our survey results
tools/resources/information were in fact large barriers to designing
showed a large gap in the adoption of existing guidelines, tools,
VUIs. This likely coincides with why designers noted using existing
and patterns in current VUI design industry, and that designers rely
examples of VUIs or their own personal knowledge for VUI design
on their previous experience in developing GUIs when designing
vs. other types of resources. Due to the nature of VUI design re-
VUIs. We believe immediate future research should be focused on
search at the moment, it makes sense that this is an existing barrier,
guideline and prototyping development and adoption, and introduc-
and supports the immediate need to develop and validate tools such
ing academic and industry training curriculums, in order to better
as design guidelines, evaluation tools, prototyping methods, etc.
support designers in the transition to voice interaction design.
As far as current needs for VUI design, most noted were prototyp-
ing tools, and academic and industry training, as shown in Section
4.2.2. Interestingly, these were noted as being equal or even more ACKNOWLEDGMENTS
important than design guidelines. Previous research has shown that This work is supported by the Natural Sciences and Engineering
there is a lack of training for designers[21, 22] – though the amount Research Council of Canada (NSERC). This work is also supported
of industry courses that are available for VUI designers to take is by AGE-WELL NCE Inc., a member of the Networks of Centres of
growing [20], as is evidenced by “training in the workplace” and Excellence (NCE), a Government of Canada program supporting
“online courses” being the largest sources of professional VUI design research, networking, commercialization, knowledge mobilization
training according to our survey. This suggests that there should be and capacity building activities in technology and ageing to improve
much more focus for developing a training curriculum for VUI design- the quality of lives of Canadians.
ers. Given that each one of our participants has had professional
academic training in Human-Computer Interaction, this suggests
REFERENCES
that developing a curriculum for VUI design in academic HCI pro-
[1] 2017. Dialogflow. https://dialogflow.com/
grams may be a great way to give designers VUI design training [2] 2018. HCI Education Workshop | CHI 2018. https://chi2018.hcilivingcurriculum.
that may take into industry, as this is a primary way to transfer org/
[3] 2019. EduCHI 2019. https://educhi2019.hcilivingcurriculum.org/
knowledge of guidelines and tools to new designers. [4] 2021. Voiceflow | Design, prototype & launch voice & chat apps. https://www.
voiceflow.com/
Takeaway 3: VUI industry designers want industry and aca- [5] 2022. Conversation Design. https://developers.google.com/assistant/
conversation-design/welcome
demic training for VUI Design, and this should be a focus for [6] 2022. Get Started with the Guide | Alexa Design Guide. https://developer.amazon.
com/en-US/docs/alexa/alexa-design/get-started.html

9
CUI 2022, July 26–28, 2022, Glasgow, United Kingdom Murad and Munteanu

[7] 2022. Introduction - Siri - Human Interface Guidelines - Apple Devel- Robert J. Moore. 2021. Let’s Talk About CUIs: Putting Conversational User
oper. https://developer.apple.com/design/human-interface-guidelines/siri/ Interface Design Into Practice. In Extended Abstracts of the 2021 CHI Conference
overview/introduction/ on Human Factors in Computing Systems. 1–6.
[8] Tawfiq Ammari, Jofish Kaye, Janice Y. Tsai, and Frank Bentley. 2019. Music, [26] Christine Murad, Cosmin Munteanu, Benjamin R. Cowan, and Leigh Clark. 2021.
Search, and IoT: How people (really) use voice assistants. ACM Transactions on Finding a New Voice: Transitioning Designers from GUI to VUI Design. In CUI
Computer-Human Interaction 26, 3 (2019). https://doi.org/10.1145/3311956 2021 - 3rd Conference on Conversational User Interfaces (CUI ’21). Association
[9] Raluca Budiu and Page Laubheimer. 2018. Intelligent Assistants Have Poor Us- for Computing Machinery, New York, NY, USA, 1–12. https://doi.org/10.1145/
ability: A User Study of Alexa, Google Assistant, and Siri. https://www.nngroup. 3469595.3469617
com/articles/intelligent-assistant-usability/ [27] Chelsea M. Myers, Anushay Furqan, and Jichen Zhu. 2019. The Impact of User
[10] Eric Corbett and Astrid Weber. 2016. What Can I Say? Addressing User Experience Characteristics and Preferences on Performance with an Unfamiliar Voice User
Challenges of a Mobile Voice User Interface for Accessibility. In Proceedings of the Interface. In Proceedings of the 2019 CHI Conference on Human Factors in Computing
18th International Conference on Human-Computer Interaction with Mobile Devices Systems (CHI ’19). Association for Computing Machinery, New York, NY, USA,
and Services (MobileHCI ’16). Association for Computing Machinery, New York, 1–9. https://doi.org/10.1145/3290605.3300277 event-place: Glasgow, Scotland Uk.
NY, USA, 72–82. https://doi.org/10.1145/2935334.2935386 event-place: Florence, [28] Hendrik Müller, Aaron Sedley, and Elizabeth Ferrall-Nunge. 2014. Survey Research
Italy. in HCI. In Ways of Knowing in HCI, Judith S. Olson and Wendy A. Kellogg (Eds.).
[11] Benjamin R Cowan, Nadia Pantidi, David Coyle, Kellie Morrissey, Peter Clarke, Springer New York, New York, NY, 229–266. https://doi.org/10.1007/978-1-4939-
Sara Al-Shehri, David Earley, and Natasha Bandeira. 2017. "What Can I Help You 0378-8_10
With?": Infrequent Users’ Experiences of Intelligent Personal Assistants. In Proc. [29] Jakob Nielsen. 2003. Voice Interfaces: Assessing the Potential. https://www.
of MobileHCI ’17. 1–12. https://doi.org/10.1145/3098279.3098539 nngroup.com/articles/voice-interfaces-assessing-the-potential/
[12] Andre P. Freire, Cibele M. Russo, and Renata P. M. Fortes. 2008. A survey on [30] Željko Obrenović. 2012. Rethinking HCI education: teaching interactive comput-
the accessibility awareness of people involved in web development projects in ing concepts based on the experiential learning paradigm. interactions 19, 3 (May
Brazil. In Proceedings of the 2008 international cross-disciplinary conference on Web 2012), 66. https://doi.org/10.1145/2168931.2168945
accessibility (W4A). Association for Computing Machinery, New York, NY, USA, [31] Cathy Pearl. 2016. Designing Voice User Interfaces: Principles of Conversational
87–96. https://doi.org/10.1145/1368044.1368064 Experiences. "O’Reilly Media, Inc.". Google-Books-ID: MmnEDQAAQBAJ.
[13] Anushay Furqan, Chelsea Myers, and Jichen Zhu. 2017. Learnability through [32] Julie Ponto. 2015. Understanding and Evaluating Survey Research. Journal of the
Adaptive Discovery Tools in Voice User Interfaces. In Proceedings of the 2017 Advanced Practitioner in Oncology 6, 2 (2015), 168–171. https://www.ncbi.nlm.
CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI nih.gov/pmc/articles/PMC4601897/
EA ’17). Association for Computing Machinery, New York, NY, USA, 1617–1623. [33] Martin Porcheron, Joel E. Fischer, and Michel Valstar. 2020. NottReal: A Tool
https://doi.org/10.1145/3027063.3053166 event-place: Denver, Colorado, USA. for Voice-Based Wizard of Oz Studies. In Proceedings of the 2nd Conference on
[14] Panayiotis Koutsabasis and Spyros Vosinakis. 2012. Rethinking HCI Education Conversational User Interfaces (CUI ’20). Association for Computing Machinery,
for Design: Problem-Based Learning and Virtual Worlds at an HCI Design Studio. New York, NY, USA. https://doi.org/10.1145/3405755.3406168 event-place: Bilbao,
International Journal of Human-Computer Interaction 28, 8 (Aug. 2012), 485–499. Spain.
https://doi.org/10.1080/10447318.2012.687664 [34] Yolanda Jacobs Reimer and Sarah A. Douglas. 2003. Teaching HCI Design With
[15] R. Langevin, R. Lordon, and T. Avrahami. 2021. Heuristic evaluation of conversa- the Studio Approach. Computer Science Education 13, 3 (Sept. 2003), 191–205.
tional agents. In Conference on Human Factors in Computing Systems - Proceedings. https://doi.org/10.1076/csed.13.3.191.14945
https://doi.org/10.1145/3411764.3445312 [35] Alexander I Rudnicky. 1996. Speech Interface Guidelines. (1996). www.speech.cs.
[16] Jonathan Lazar, Alfreda Dudley-Sponaugle, and Kisha-Dawn Greenidge. 2004. cmu.eduu
Improving web accessibility: a study of webmaster perceptions. Computers in [36] C. Rusu, S. Roncagliolo, V. Rusu, and C. Collazos. 2011. A Method-
Human Behavior 20, 2 (March 2004), 269–288. https://doi.org/10.1016/j.chb.2003. ology to Establish Usability Heuristics. undefined (2011). https:
10.018 //www.semanticscholar.org/paper/A-Methodology-to-Establish-Usability-
[17] Ewa Luger and Abigail Sellen. 2016. "Like Having a Really Bad PA": The Gulf Heuristics-Rusu-Roncagliolo/56da59df101cc3b12324665ab3a647c25bbee932
between User Expectation and Experience of Conversational Agents. In Proceed- [37] J Sherwani, Dong Yu, and Tim Paek. 2007. Voicepedia: towards speech-based
ings of the 2016 CHI Conference on Human Factors in Computing Systems - CHI ’16. access to unstructured information. Interspeech (2007), 2–5. http://research.
5286–5297. https://doi.org/10.1145/2858036.2858288 microsoft.com/pubs/78835/VoicePedia-Interspeech2007.pdf
[18] Sarah Mennicken, Oliver Zihler, Frida Juldaschewa, Veronika Molnar, David [38] Ben Shneiderman. 2000. The limits of speech recognition. Commun. ACM 43, 9
Aggeler, and Elaine May Huang. 2016. "It’s like living with a friendly stranger": (2000), 63–65. https://doi.org/10.1145/348941.348990
Perceptions of personality traits in a smart home. In UbiComp 2016 - proceedings of [39] Olivier St-Cyr, Elizabeth F Churchill, and Craig M MacDonald. 2019. EduCHI
the 2016 ACM international joint conference on pervasive and ubiquitous computing. 2019 Symposium: Global Perspectives on HCI Education. 2019 (2019), 7.
120–131. https://doi.org/10.1145/2971648.2971757 [40] Bernhard Suhm. 2003. Towards Best Practices for Speech User Interface Design.
[19] Cosmin Munteanu. 2018. HCI Curriculum Across Disciplinary Borders: a Case In Proc. of EuroSpeech ’03. 2217–2220.
Study of Teaching UX outside the Computing Sciences. (2018), 4. [41] Z. Wei and J. A. Landay. 2018. Evaluating Speech-Based Smart Devices Using
[20] Christine Murad and Cosmin Munteanu. 2019. "I Don’t Know What You’re Talking New Usability Heuristics. IEEE Pervasive Computing 17, 2 (2018), 84–96. www.
about, HALexa": The Case for Voice User Interface Guidelines. In Proceedings scopus.com
of the 1st International Conference on Conversational User Interfaces (CUI ’19). [42] Kathryn Whitenton. 2016. Voice Interaction UX: Brave New World...Same Old
Association for Computing Machinery, New York, NY, USA. https://doi.org/10. Story. https://www.nngroup.com/articles/voice-interaction-ux/
1145/3342775.3342795 event-place: Dublin, Ireland. [43] Lauren Wilcox, Betsy DiSalvo, Richard Henneman, and Lindsay Kelly. 2018. Design
[21] Christine Murad and Cosmin Munteanu. 2020. "Alexa, How do I Build a VUI and the Future of the HCI Classroom: Lessons Learned from an International
Curriculum?". In Proc. of CUI 2020. Survey on HCI Education. (2018), 4.
[22] Christine Murad and Cosmin Munteanu. 2020. Designing Voice Interfaces: Back [44] Nicole Yankelovich, Gina-Anne Levow, and Matt Marx. 1995. Designing
to the (Curriculum) Basics. In Proceedings of the 2020 CHI Conference on Human SpeechActs: Issues in Speech User Interfaces. In Proc. of CHI ’95. 369–376.
Factors in Computing Systems (CHI ’20). Association for Computing Machinery, https://doi.org/10.1145/223904.223952
New York, NY, USA, 1–12. https://doi.org/10.1145/3313831.3376522 tex.ids=
Murad2020 tex.art_number: 3376522 tex.author_keywords: conversational in-
terface; hci curriculum; hci education; speech; voice user interface; vui design
tex.document_type: Conference Paper tex.source: Scopus event-place: Honolulu,
HI, USA.
[23] Christine Murad, Cosmin Munteanu, Leigh Clark, and Benjamin R. Cowan. 2018.
Design guidelines for hands-free speech interaction. In Proc. of MobileHCI ’18.
ACM Press, New York, New York, USA, 269–276. https://doi.org/10.1145/3236112.
3236149
[24] Christine Murad, Cosmin Munteanu, Benjamin R. Cowan, and Leigh Clark. 2019.
Revolution or Evolution? Speech Interaction and HCI Design Guidelines. IEEE
PERVASIVE COMPUTING 18, 2 (June 2019), 33–45. https://doi.org/10.1109/MPRV.
2019.2906991
[25] Christine Murad, Cosmin Munteanu, Benjamin R. Cowan, Leigh Clark, Martin
Porcheron, Heloisa Candello, Stephan Schlögl, Matthew P. Aylett, Jaisie Sin, and

10