ArticlePDF Available

Abstract

There has been little research into whether a persona's picture should portray a happy or unhappy individual. We report a user experiment with 235 participants, testing the effects of happy and unhappy image styles on user perceptions, engagement, and personality traits attributed to personas using a mixed-methods analysis. Results indicate that the participant's perceptions of the persona's realism and pain point severity increase with the use of unhappy pictures. In contrast, personas with happy pictures are perceived as more extroverted, agreeable, open, conscientious, and emotionally stable. The participants’ proposed design ideas for the personas scored more lexical empathy scores for happy personas. There were also significant perception changes along with the gender and ethnic lines regarding both empathy and perceptions of pain points. Implications are the facial expression in the persona profile can affect the perceptions of those employing the personas. Therefore, persona designers should align facial expressions with the task for which the personas will be employed. Generally, unhappy images emphasize realism and pain point severity, and happy images invoke positive perceptions.
Can Unhappy Pictures Enhance the Eect of Personas? A User Experiment
ere has been lile research into whether a persona’s picture should portray a happy or unhappy individual. We report a user
experiment with 235 participants, testing the eects of happy and unhappy image styles on user perceptions, engagement, and
personality traits aributed to personas using a mixed-methods analysis. Results indicate that the participant’s perceptions of the
persona’s realism and pain point severity increase with the use of unhappy pictures. In contrast, personas with happy pictures are
perceived as more extroverted, agreeable, open, conscientious, and emotionally stable. e participants’ proposed design ideas for the
personas scored more lexical empathy scores for happy personas. ere were also signicant perception changes along with the
gender and ethnic lines regarding both empathy and perceptions of pain points. Implications are the facial expression in the persona
prole can aect the perceptions of those employing the personas. erefore, persona designers should align facial expressions with
the task for which the personas will be employed. Generally, unhappy images emphasize realism and pain point severity, and happy
images invoke positive perceptions.
Keywords: Personas, Design, Pictures, Sentiment, User Experiment
1 INTRODUCTION
A persona is a ctitious person representing a real user group of particular interest [15]. Personas are a well-
established, user-centered design technique [50] for (soware) supporting developers, designers, and other
stakeholders engaged in product development, requirements engineering, UX/UI design, marketing, user support, and
other tasks requiring an understanding of users or customers [6,13,47,56]. Personas are presented in proles that
display information, such as the goals, needs, and wants of the user segment, and, thus, help designers contextualize
users [88]. An essential part of the design of these user proles is the persona picture [48,60]. Persona pictures
constitute a non-verbal form of communicating details of the persona to the users [26]. e picture inuences how the
persona is perceived and what connotations and stereotypes are associated with the persona by the stakeholders
[27,67]. As such, choosing the persona picture is a crucial step in the design of a persona.
e current study investigates the eect of happy versus unhappy pictures on the perceptions of those using the
persona. Research on person perception in psychology lends support to the general idea that people perceive others
dierently based on their observed mood [41,45]. Specically, “happy people” may be perceived dierently than
“unhappy people” [8]. However, there is no evidence whether using happy pictures is beer than unhappy pictures for
personas. In this study, we specically test whether using “unhappy” pictures enhances central persona perceptions
[78], such as empathy, pain point, usefulness, realism, and completeness. ese persona perceptions are key to the
eective employment of personas by end-users. We also perform quantitative and qualitative analyses on the wrien
outputs of the users’ design task to investigate if the design ideas generated using happy versus unhappy personas are
qualitatively dierent. This analysis is based on the participants’ proposed product ideas that address the remote-work
needs of a persona. Given that examples can be found of design personas of both types (see ), it can be stated that both
approaches (happy/unhappy pictures) are used by persona creators. However, the fact there is no study on the pros
and cons of each approach implies a research gap, which our study addresses. Given that personas in use contain both
happy/unhappy facial expressions, addressing this gap can positively aect the design of future personas.
e design scenario we use concerns developing a product that addresses the remote-work needs of a given user
segment that the persona represents. We create the persona proles that represent people with remote-work needs
[20] and vary the personas’ gender, ethnicity, and picture happiness to address our research questions (introduced in
Section 2). User needs (and personas) for remote work are extremely topical, not only because of issues such as the
COVID-19 pandemic but also because of the long-term trend of work moving toward remote (nomadic) work and
distributed teams [14,43,73], especially in knowledge-intensive industries and professions dealing with digital inputs
and outputs. erefore, the design issue itself is of importance and is of practical relevance.
(a) (b)
Figure 1: (a) Unhappy persona (picture credit: hps://dribbble.com/shots/3338369-Universe-User-Personas/aachments/723326). (b)
Happy persona (picture credit: hps://evolt.io/platform-tour/user-persona/). Whether persona creators should use pictures of happy
or unhappy individuals is unclear. Both practices are used.
2 RELATED LITERATURE
2.1 Persona Pictures as Critical Design Cues
Researchers have found persona pictures inuential for persona users’ perceptions of the persona [27,34,51,64,65,67].
Pictures aract the user’s focus of aention to personas, guide the processing of persona information, and convey
emotional signals about the persona. Perception of the personas, in turn, aects the aitudes and deployment of
personas in design tasks [38,39,57,58]. erefore, the choice of a persona picture maers for the outcome of persona
adoption and use in organizations engaged in user-centered design activities.
Despite the broadly acknowledged importance of pictures for the design of personas, it is unclear what kind of
pictures persona developers should choose when creating persona proles. e typical choice, thus far, has been to use
online photobank pictures with smiling, good-looking people [63,64] who appear happy and content in their lives.
However, there is an issue with this: Is this optimal for persona design, or are there beer alternatives for persona
imagery?
is issue is exacerbated by the fact that there is scarce research on persona pictures in the literature. Also, design
studies employing personas do not typically explain the rationale behind choosing particular photos [32], which is
interesting given the amount of eort taken to generate the other persona information.
Out of the few studies focusing on persona pictures, Long [37] analyzed the use of illustration-style pictures versus
photographs and found that photographs were more optimal for user recall, with the participants recalling more
details of the persona when using real photographs. Long also found that illustrations resulted in more vague answers
about the persona in post-session interviews and risked the participants superimposing self-referential information
about themselves onto the personas (a general challenge of the persona technique [84]).
Nieters et al. [59], in turn, found that using the pictorial style of action gures enhanced the memorability of the
personas. e downside was that these personas were taken less seriously by corporate stakeholders, who found the
action personas amusing and joked about them. Nevertheless, the ndings of Nieters and colleagues raise the
important question of whether “seriousness” is an essential quality for the persona to be remembered by the
stakeholders.
2
Oen, design studies employing personas do not explain the rationale behind choosing particular pictures [32].
From our review of persona proles in the literature, most appear to be stock photos (i.e., photos from commercial or
open-source photo archives). ere is some evidence that using stock photos of professional models is not always
optimal for persona design. Salminen et al. [72] conducted a large-scale survey study with 2,400 participants where
they compared the eect of photographs taken of “real people” against those taken of professional models on various
user perceptions of personas, as well as investigating the eect of a smiling picture on the perceptions. eir ndings
indicated that a smile increased the perceived similarity with the persona, similar personas had a higher likability, and
likability increased the willingness to use a persona [72]. Furthermore, the use of stock photos decreased the perceived
similarity with the persona and their credibility, both of which were signicant predictors of one’s willingness to use a
persona.
e collective evidence from these prior studies suggests that there is room to investigate the optimal pictures for
persona design. In particular, this research focuses on the question of “picture happiness” (i.e., the person in the image
is feeling or showing pleasure or contentment). We pose the following research questions (RQs):
RQ1: Does the use of happy/unhappy pictures alter the persona users’ perceptions of personas?
RQ2: Does (a) gender and/or (b) ethnicity inuence the persona users’ perceptions of personas when using
happy/unhappy pictures?
RQ3: Does the use of happy/unhappy pictures alter the users’ engagement (time participants spend viewing) with the
persona?
RQ4: How does the use of happy/unhappy persona pictures aect initial solutions designed for personas’ pain points?
For the rst and third RQs, we formulate specic hypotheses in the following subsection. For the second RQ, we
conduct a statistical analysis investigating these variables. For the fourth RQ, we conduct a qualitative analysis on the
initial task outputs obtained from persona users, reported in Section 5, with suggestions of future research pursuing
the eect of these personas further in the design process.
2.2 Hypotheses
For our hypothesis formulation, we draw from previous persona research investigating persona perception [42,74],
referring to persona users perceiving the personas as any other human being. is notion is consistent with Cooper’s
initial idea of personas being ctitious but realistic in their portrayal of user groups [15], but the concept of persona
perception specically makes the connection between persona research and social psychology studies, in which person
perception (i.e., the views and aitudes held by a person about others [34]) is a well-studied eld of inquiry. rough
this conceptual linkage, persona studies in HCI can borrow terms, concepts, and theories from social psychology to
beer understand how persona users perceive personas in various design tasks [42]. Particularly, our hypotheses deal
with the dualism of perceiving others as happy (or not) and how this dualism aects the perception of personas for a
design task.
To this end, we formulate the following hypotheses:
H1: Picture happiness increases users’ empathy toward the persona. [RQ1]
H2: Picture happiness decreases users’ perceptions of the persona’s pain points. [RQ1]
H3: Picture happiness increases users’ perceptions that the persona is useful for a design task. [RQ1]
H4: Picture happiness decreases users’ perceptions of the persona as being realistic. [RQ1]
H5: Picture happiness increases users’ perceptions of the persona having complete information. [RQ1]
H6: Picture happiness increases users’ perceptions of the persona’s personality. [RQ1]
H7: Picture happiness increases users’ engagement with the persona. [RQ3]
3
ese are predominant traits of persona perceptions by end-users. On the one hand, previous research suggests
that “persona creators should use smiling pictures of real people to evoke positive perceptions toward the personas”
[72] (p. 1). On the other hand, stock photos (of mostly happy people) were interpreted, in the same study, to diminish
the users’ sense of identication with the persona [72]. e “real” pictures deployed in this previous study did not
portray professional models but “everyday people.” However, acquiring such photographs, even if they would be
optimal, is dicult due to constraints of time, cost, and usage rights (privacy). erefore, persona creators typically
resort to photobank pictures largely containing professional models. Given this bounded realism of persona design, the
interesting question is whether “unhappiness” in the picture can curb some of the previously observed challenges of
stock photos by making the persona appear more “real” [H1].
Research in social psychology suggests that negative people may be taken more seriously than happy people,
especially when negative messages are communicated with unhappy facial expressions, as these expressions enhance
the eect of the message [22]. In turn, happy people may, in some circumstances, be perceived as “fake” or non-serious
[24] [H4], especially when dealing with understanding a person’s needs [28] [H2].
Based on Gestalt theory, which suggests that the whole of a design is more than the sum of its parts [46], we
hypothesize that personas with unhappy pictures are perceived as more useful [H3] as they emphasize and support
the pain-point information (i.e., relevant information for the task), whereas happy pictures could be seen as redundant
information. Similar to the holistic processing of faces [30], the information in persona proles is processed in a
holistic manner [6]. is same eect may lead to personas with unhappy pictures being perceived as more complete
than those with happy pictures [H5], as the laer may conict with the other information the users are focused on
(i.e., the pain points themselves).
Furthermore, users can make inferences about the persona’s personality based on the persona picture [75]. ese
personality ratings can then aect what kind of solutions are developed for the persona [3,4]. For example, extroverted
personas may be considered to require dierent kinds of products than introverts. e underlying notion here is that
people tend to extrapolate someone’s personality from a single picture (the rst impression eect [40]): Hence, the
momentary “mood” of the persona aects how its personality is judged [H6].
Finally, happy people may be perceived as more pleasant, which may aect how receptive users are toward them
[35]. e underlying notion here is positivity bias [87]—people are more drawn toward happy people and therefore
spend more time geing to know such individuals. erefore, we expect that users would be more interested in
personas with happy pictures and would consequently spend more time reviewing their information [H7]. Overall,
these reasons suggest that the choice of picture based on happiness can have an impact on persona design, especially
when considering personas as individual human beings. e impacts are far-reaching, as the ndings can be leveraged
for a range of systems that employ pictures of people and as the facial expressions for happy and unhappy tend to be
universal [16,17].
2.3 Measurement Items
Table 1 reports the constructs and measurement items (indicators). e indicators for empathy, usefulness, realism,
completeness, and likability were adopted from the Persona Perception Scale (PPS) [78], an instrument specically
developed for measuring user perceptions of personas and applied in several previous persona user studies
[61,62,64,69]. e indicator for physical aractiveness was adopted from Berscheid and Walster [10]. e pain point
intensity indicator was specically developed for this study—for this, we used Chaopadhyay et al. [14] as inspiration,
as the researchers in that study operationalized user needs statements in a contextual manner.
Personality rating items were obtained from the Ten-Item Personality Inventory (TIPI) scale by Gosling [23], which
measures the Big Five personality traits: Extroversion (EX), Agreeableness (AG), Conscientiousness (CN),
4
Emotional Stability (ES), and Openness (OP). e constructs were measured with Likert scale (7-point) response
options, ranging from “Strongly agree” to “Strongly disagree.” For Picture happiness, the options ranged from
“Extremely happy” to “Extremely unhappy.” Finally, the indicator for Engagement was obtained from altrics by
recording the duration (in seconds) the participants spent perusing the persona they were presented.
Table 1: Constructs and measurement items used in the study. The last column shows their connection to the hypotheses.
ID Construct Measurement item Hypothesis
EM Empathy I felt I could understand the persona as a human being. H1
PP Pain point intensity e persona struggles with remote work. H2
US Usefulness e persona contained useful information for my task of creating
a remote work product.
H3
RE Realism e persona seemed realistic. H4
CO Completeness e persona prole was complete, so that it contained all the
necessary information to understand the users it represents.
H5
PE Personality (Big Five) I see the persona as:
Extraverted, enthusiastic.
Critical, quarrelsome.
Dependable, self-disciplined.
Anxious, easily upset.
Open to new experiences, complex.
Reserved, quiet.
Sympathetic, warm.
Disorganized, careless.
Calm, emotionally stable.
Conventional, uncreative.
H6
EN Engagement Dwell time (duration) H7
PH Picture happiness1How (un)happy did the persona look like? Independent variable
for H1-06
We used the Big Five for two reasons: (1) its commonness in psychological studies and in HCI [12,31] and (2) the
fact that it has been deployed in previous persona studies, with ndings indicating that the persona’s perceived
personality traits can aect design outcomes [4,5,76]. We chose the TIPI scale for the operationalization of the Big Five
traits because this scale aords ease of completion by non-psychology experts while simultaneously providing valid
personality assessment [23]. e denitions of the Big Five traits are as follows [1]:
Extrovert (EX): Active, amicable, assertive, energetic, enthusiastic, outgoing, talkative. ese individuals are
friendly and draw inspiration from social situations.
Agreeable (AG): Compassionate, cooperative, generous, helpful, kind, nurturing, sympathetic. ese
individuals are generally optimistic and trusting of others.
Conscientious (CN): Ecient, hardworking, organized, persevering, responsible, self-disciplined. ese
individuals tend to be reliable and focused on achieving and planning for the future.
Emotionally stable (ES): Calm, relaxed, self-condent. Emotionally stable individuals are not moody or
tense, and they are not easily tipped into experiencing negative emotions.
Open (OP): Artistic, creative, curious, deep, intelligent, imaginative, open-minded, reective. Open
individuals tend to appreciate diverse views, ideas, and experiences.
ese denitions were not shown to the participants, as they are not required for assessing the personality traits,
which is done by calculating the score of each trait from the statements posed to the participants. TIPI is usually
administered as a form of self-evaluation of personality [23]. However, as we wanted the participants to evaluate the
1 Picture happiness is inversely correlated to the “Picture unhappiness” in the hypotheses. In other words, a high
Picture happiness is low Picture unhappiness.
5
personality of the personas, we transformed the original statement ofI see myself as intoe persona seemed
like …” e scores were processed using the instructions given by Gosling [23] to obtain the nal personality ratings.
3 METHODOLOGY
3.1 Overview
ree separate within-subject experiments were conducted, each with a dierent set of personas (see Appendix A for
the full persona proles). Given that our primary focus of the study was the eect of happy and unhappy pictures on
perceptions of the personas, an experimental seing is an appropriate methodological approach as it maximizes
control and precision [47], albeit at the cost of realism. e unhappy/happy conditions and the order were mixed
within each 2 × 2 sequence by randomized assignment and counterbalancing. In other words, each participant sees two
dierent personas, one with a “happy” picture and another one with an “unhappy” picture. We used the conguration
options in the survey soware (altrics) to group the dierent persona pairs, and we then applied randomization and
counterbalancing to the groups. is removes the possibility of a participant seeing the same persona or the same pain
point prole twice.
3.2 Persona Creation
Various methods can be applied to create personas with varying degrees of manual work and automation [33,51]. In
this work, we opted for manual persona creation, as this enables a greater degree of control in the output personas
than using automatic methods. To address our hypotheses and to analyze the consistency of results by demographic
groups, we created personas from multiple genders (male/female) and ethnic backgrounds (see Appendix A for the full
persona proles), leaving the study of other genders and other ethnicities for future research. We selected three ethnic
backgrounds from the listing of the American Psychological Association [2]: persons of African origin, persons of
European origin, and persons of Middle Eastern origin. As far as we know, this is the rst study to investigate persona
perceptions of persons of Middle Eastern origin (African and European origins have been investigated previously,
along with Asian origins [72]). e inclusion of personas with dierent ethnic backgrounds enables us to investigate
possible stereotyping [84].
Male-AO-happy Male-AO-unhappy Female-AO-happy Female-AO-unhappy
Male-EO-happy Male-EO-unhappy Female-EO-happy Female-EO-unhappy
6
Male-MO-happy Male-MO-unhappy Female-MO-happy Female-MO-unhappy
Figure 2: Final persona picture pairs (happy/unhappy versions). The pictures were manually curated from an online photobank
service (hps://www.123rf.com), with a license for research use purchased. AO = African origin; EO = European origin; MO = Middle
Eastern origin. The source for ethnical terms in the study is American Psychological Association [2].
e persona pictures (see Figure 2) were selected by browsing online photo banks and identifying suitable picture
pairs of happy/unhappy individuals. We dened the following selection criteria for the pictures:
(1) the pictures are taken by the same photographer (to ensure a consistent technical quality),
(2) the environment (“scene”) of the picture is at home (consistent with the remote-work scenario),
(3) the human model is the same in each picture pair (to avoid possible inconsistencies from the use of dierent
people for the same persona demographics),
(4) each picture has a laptop that the person is using (to signify remote work scenario), and
(5) in all pictures, the gaze is indirect (i.e., not looking at the camera) to mitigate the possibility that the persona-
user rapport is aected by the persona’s pose in the picture.
In other words, we kept as many aspects in the pictures as constant as possible, only varying the sentiment of
happiness. We focused on the broad categories of happy or unhappy facial expressions, keeping the general facial
expression in each category for each treatment approximately the same. Multiple iterations of identifying candidate
pictures were conducted, with one of the authors identifying the pictures and the others commenting on observed
inconsistencies. Once an internal agreement on the pictures was reached, the treatments incorporating these pictures
were created, with examples provided in Figure 3. e process for this image selection was quite lengthy,
demonstrating the issues of using photo banks for persona proles. All treatments are available in Appendix A.
e above choices resulted in the creation of 12 baseline persona treatments (two happiness levels × two genders ×
three ethnic backgrounds). Because each participant would see two personas, it was required to vary the personas’ text
content as well. For this purpose, we created two dierent “pain point proles” that reected dierent user needs for
remote working. We chose the research context of remote work, because of the topicality of remote work, especially
with the global COVID-19 pandemic that is strongly aecting the design profession (and other professions) at the time
of study. As the research context was remote work, the personas were created to contain pain points and other
information related to remote work user types. We created two “pain point proles” (see Figure 3) to avoid showing
one participant two identical persona descriptions; the personas need to dier in their content. e pain point proles
are based on a previous study on challenges observed by remote workers [20].
7
Pain point prole 1 Pain point prole 2
Figure 3: Two persona profile treatments. The le picture shows Pain Point Profile 1 (lack of computer skills + slow Internet) with a
happy persona picture. The right picture shows Pain Point Profile 2 (inability to separate work from leisure + lack of focus and
productivity) with an unhappy persona picture. Pain point profiles and happy/unhappy pictures were combined for each persona,
resulting in four treatments per persona and 24 treatments overall.
The treatments were extensively pilot tested. One member of the research team developed the personas. The set of
personas was then critically reviewed by other research team members. Changes centered on the wording of the text and
ensured minimal differences between images as possible. One member of the research team alternated the personas, and
the personas were again pilot tested by members of the research team. This process was repeated until there were no
suggested changes to the persona sets.
is [20] study was based on two surveys with a total of 3,634 responses addressing challenges remote workers
faced during the COVID-19 pandemic. e study found several challenges, of which we chose the following: problems
with connectivity (slow/shared Internet connections, unstable home Internet; captured in pain point 2: “My Internet is
frustratingly slow.”), managing work-life balance (pain point 3: Sometimes I end up working the whole day.”),
interruptions and distractions (leading to a loss of focus and productivity; pain point 4:Lately, I’ve been struggling to
keep up with my work.”). We also included one other aspect from the research on the digital divide [86], namely, the
fact that not all remote workers feel comfortable with or are procient at using remote work hardware/soware (Pain
point 1: “I wish I was beer with computers”). Overall, with these personas, we wanted to represent “regular people”
with these pain points – not only those that nd remote work “easy” and “natural.”
8
3.3 Data Collection
For data collection, we used Prolic2, an online survey platform. is platform has been deployed in several previous
persona user studies [63–65,69,70] and survey-based research in other domains [52,76], with evaluation studies
showing high data integrity relative to other platforms [53,54]. e survey was pilot tested with three participants, two
from the research team and one external reviewer. Based on the comments by the test participants, several changes
were made to the study introduction. Some minor wording changes were also made to the statements.
To reach adult professionals working in industries relevant to personas, we applied the following sampling criteria
in Prolic: student status (“no”), highest education (“at least undergraduate”), age (“25–60 (inclusive)”), industry
(“art/design; product development; college, university, and adult education; information services and data processing;
other education industry; soware; and scientic or technical services”).
e survey platform had 5,079 matching participants who had been active in the past 90 days. For each 2 × 2
persona pair, we recruited 80 participants (240 in total). ree studies were administered sequentially, and participants
were prevented from participating in more than one study by using the blacklist option on the Prolic platform. e
estimated survey completion time was 21 minutes according to altrics’ estimation tool, which our pilot testing
conrmed to be close to accurate. us, we oered the participants an hourly rate of 8.94, which exceeds the UK
minimum wage (8.72 for workers above 25 years of age as of April 20203). Based on these parameters, the total data
collection cost was 340.55 x 3 = 1,021.65, which included the VAT and the platform’s commission (30%).
3.4 Survey Protocol
Here, we explain how the survey was administered to the participants. Each step is explained in the following.
A. Survey introduction (1/2): “Welcome to this online survey on using personas for the development of remote-
work products. Your responses will be used for an academic study. Participation is voluntary, and you can stop at any
time. For any questions about the study, you can reach out to Dr. [name and email of the principal investigator].”
B. Survey introduction (2/2): “You will be shown a persona (a ctional character) that contains key information
about the remote working needs of a given user segment. Your task is to develop a product that addresses the remote
working needs of this user segment.” In this survey, you will be shown two dierent personas.” e introduction was
split into two pages to facilitate the participants’ absorption of information.
C. Persona denition: “A persona is dened as a ctitious user type and is not a real person. It is a character that
portrays many users.” Acknowledgment of understanding this denition was required (see Figure 4a).
2 hps://www.prolic.co/
3 hps://www.gov.uk/national-minimum-wage-rates, accessed December 2020.
9
(a) (b)
Figure 4: (a) Persona definition in mobile view of survey. (b) Persona profile example in mobile view (James, male of European
origin, unhappy version).
D. Persona experience: “Your experience with personas” was asked, with the following options:
oUnfamiliar – no experience of personas whatsoever
oNovice – have used personas before but not much
oProcient – have used personas several times before
E. Task introduction: “Review the information shown about the persona. Aerward, you will be asked some
questions about this persona. en you will be asked to describe the product that you want to develop for the persona.
e product can be any oering—digital or physical—that helps the persona cope with remote work.”
F. Persona prole: One of the tested persona proles was randomly loaded; see an example in Figure 4b.
G. Task: “Describe your product idea that addresses a remote working need of this persona. Please write as
detailed a concept as possible, provide information from the persona to support your product, and explain what
need(s) the product addresses (min. 300 characters).”
H. Background information: estions were asked regarding age, gender, ethnic background, and job title (if
not currently working, then the most recent occupation). Aer this, the participants were thanked, and they were
redirected to the survey platform’s conrmation page.
3.5 Participant Information
e three pilot test participants were removed from the dataset. Additionally, 5 participants who failed an aention
check included in the survey were also removed. is led to a nal sample size of 235 participants. e sample had
10
slightly more males (n = 142; 60.4%), 92 females (39.1%), and a single non-binary/third gender individual (0.4%). In
terms of age, the sample was relatively young (M = 34.51; SD = 7.858). e most represented ethnicity was European
(n = 189; 80.4%), followed by Latinx (n = 19; 8.1%), Asian (n = 17; 7.2%), African (n = 4; 1.7%), Middle Eastern (n = 3;
1.3%), and other (n = 3; 1.3%). e most represented countries were Portugal (n = 33; 14.0%), United Kingdom (n = 30;
12.8%), Poland (n = 24; 10.2%), Italy (n = 17; 7.22%), and Spain (n = 12; 5.1%), with the rest representing numerous other
countries.
Regarding the previous experience with personas, 91 (38.7%) participants reported no previous experience, 115
(48.9%) indicated they had previous experience, and 29 (12.3%) considered themselves procient.
Due to the low number of African and Middle Eastern participants, these participants were assigned into one
group, “Middle East and Africa,” which is conceptually similar to the cultural region of the Middle East and North
Africa (MENA4). e most typical professions among the participants were educators ( n = 36, 15.3%), soware
developers (n = 34, 14.5%), managers (n = 29, 12.3%), researchers (n = 27, 11.5%), IT professionals (n = 22, 9.4%),
designers (n = 18, 7.7%), and data scientists or analysts (n = 12, 5.1%).
3.6 Statistical Analysis
3.6.1 Control Variables
We compare the consistency of the results based on personas’ gender and ethnic background and participants’ gender
and ethnic background. is is primarily because gender and ethnic background were found to be inuential variables
in previous persona research [27,64,67] [43]. erefore, four control variables were included in the model: Persona
Gender, which indicates the persona’s gender (reference category, “Male”); Persona Ethnicity, which indicates the
persona’s ethnicity (reference category, “European origin”); Participant Ethnicity (reference category, “European”);
Participant Gender (reference category, “Female”); and Age (continuous variable indicating a participant’s age).
3.6.2 Manipulation Check
To conrm the eects of the manipulation, we asked participants to indicate whether the persona looked happy,
unhappy, or neutral. It was expected that happy personas were correctly identied as happy and unhappy as unhappy.
However, aer crossing the observed responses with the expected responses, it was noted that, for four treatments (see
Table 1), the participants did not identify the persona’s happiness as expected. Because of this, we used the perceived
picture happiness as the independent variable in lieu of the happy/unhappy assignment grouping variable. Potential
same-participant bias is not expected since the participant ow was randomized and accounted for all possible
combinations of personas.
Table 2: The observed ratio values. The percentages indicate how many responses from the participants corresponded with the
expected value. For example, if nine out of 10 participants chose “Happy” for the statement “The persona looked…”—with the
options being “Happy,” “Unhappy,” or “Neither”—then the observed ratio would be 90%. Instances in which less than three-fourths
of the participants agreed with the expected value are highlighted in red.
Observed
ratio
Khaled Hind Roger Sarah James Jane
Happy 84.21% 76.19% 72.97% 85.00% 69.23% 75.00%
Unhappy 95.00% 62.16% 76.19% 48.72% 77.50% 94.74%
4 hps://ustr.gov/countries-regions/europe-middle-east/middle-east/north-africa
11
3.6.3 Procedure
is structuring of the data allowed the hypothesis to be tested through multiple linear regressions (GLM) [23,42],
where the aforementioned variables were used as independent variables and the dependent variables were the various
persona and personality traits mentioned in the hypothesis section. In line with the nature of the variables, continuous
variables (Age and Perceived Happiness) were inserted as is, whereas categorical variables were re-coded as dummies
beforehand (with the reference categories as previously noted). Before conducting the analysis, the various
assumptions for multiple linear regression were checked. We began by verifying the absence of multicollinearity, that
is, a substantial degree of inter-correlation between the independent variables. For this purpose, we used the Tolerance
and variance ination factor (VIF) indicators. Tolerance values should be close to 1, whereas VIF typically indicates
issues with multicollinearity when it exceeds the threshold of 5 [52]. No evidence of multicollinearity was found, as
shown in Table 3: Multicollinearity diagnostics.
Table 3: Multicollinearity diagnostics.
Variable Tolerance VIF
Participant Age 0.967 1.035
Persona Gender (Female) 0.995 1.005
Persona Ethnicity (Arab) 0.736 1.359
Persona Ethnicity (Black) 0.732 1.365
Participant Ethnicity (Middle East and Africa) 0.987 1.013
Participant Ethnicity (African) 0.979 1.021
Participant Ethnicity (Latinx) 0.965 1.036
Participant Gender (Non-binary / third gender) 0.990 1.010
Participant Gender (Male) 0.975 1.025
Perceived Happiness 0.987 1.013
We proceed by evaluating the remaining assumptions for each multiple regression, notably linearity,
homoscedasticity, independence of the residuals, and normality [25]. is was done through visual inspection of the
residual plots, using the criteria established by Hair et al. (2014). A regression plot that meets the assumptions
manifests itself as a random distribution of residuals concentrated around the point of origin. None of the reported
regressions exhibited indication of assumption violation; as such, the analysis proceeded as normal. e residual plots
are included in Appendix B for reference. e subsequent analysis reported both unstandardized (B) and standardized
(Beta) coecients.
4 RESULTS
4.1 Summary of Results
Table 4 and show the main results for RQ1 (Does the use of happy/unhappy pictures alter persona users’ perceptions of
personas?). Support for the hypotheses is discussed in the following subsections.
4.2 RQ1: Eects on Users’ Perceptions of Personas
Figure 5 illustrates the scores obtained using happy and unhappy personas.
12
Empathy
PP
Usefulness
Realism
Completeness
0 1 2 3 4 5 6 7
Unhappy Happy
Extraversion
Agreeableness
Conscien%ousness
Stability
Openness
0 1 2 3 4 5 6
Unhappy Happy
(a) PPS scores (b) TIPI scores
Figure 5: Perception scores for happy and unhappy personas. The results indicate more positive personality traits (TIPI scores) are
associated with personas with happy pictures, but perceptions important for design tasks (PPS scores) are slightly higher when
using unhappy pictures.
H1: Picture happiness increases users’ empathy toward the persona. Aer taking into account the eects of
the control variables, perceived happiness was found to have no relation to perceived empathy ( B = -0.024, p = 0.252).
erefore, H1 is not supported; picture happiness does not aect the users’ perceived empathy toward the persona.
H2: Picture happiness decreases users’ perceptions of the personas’ pain points. Aer considering the
eects of the controls, perceived happiness was found to signicantly decrease the perceived pain points ( B = -0.251, p
< 0.001). erefore, H2 is supported; picture happiness decreases users’ perception of personas’ pain points.
H3: Picture happiness increases users’ perceptions that the persona is useful for a design task. Aer
taking into account the eects of the control variables, perceived happiness did not exhibit any signicant eect
regarding the perceived usefulness of a persona (B = -0.033, p = 0.274). erefore, H3 is not supported; picture
happiness does not aect the users’ perceptions of the persona as being useful for their task.
H4: Picture happiness decreases the users’ perceptions of the persona as being realistic. Aer taking into
account the eect of the control variables, increases in perceived happiness led to decreases in perceived realism ( B = -
0.063, p = 0.013). erefore, H4 is supported; picture happiness decreases users’ perceptions of a persona as being
realistic.
H5: Picture happiness increases the users’ perceptions of the persona having complete information.
Aer considering the eect of the control variables, no signicant eect could be found regarding the perceived
happiness of a persona and the degree of perceived completeness (B = -0.029, p = 0.357). erefore, H5 is not
supported; picture happiness does not aect the users’ perceptions of the persona as having complete information.
H6: Picture happiness increases the users’ perceptions of the persona’s personality. Aer considering the
control eects, it was determined that increased perception of a persona’s happiness led to increased perceptions of all
personality traits, notably, Extraversion (B = 0.245, p < 0.001), Agreeableness (B = 0.180, p < 0.180), Conscientiousness
(B = 0.178, p < 0.001), Emotional Stability (B = 0.320, p < 0.001), and Openness to Experience (B = 0.094, p < 0.001).
erefore, H6 is supported; picture happiness aects how users rate the personality of the personas. Personas with
happier pictures are perceived as more extroverted, agreeable, conscientious, emotionally stable, and open.
13
Table 4: Regression model coeicients for persona perceptions. A negative coeicient between the picture happiness variable (last row) and other variables indicates that
as picture happiness increases, persona perceptions generally decrease. See Table 1 for variable names and definitions.
EM PP US RE CO
Variable B / SE Beta p B / SE Beta p B / SE Beta p B / SE Beta p B / SE Beta p
Participant Age 0.009
(0.005) 0.082 0.078 0.004
(0.010) 0.018 0.694 0.014
(0.007) 0.088 0.060 0.007
(0.006) 0.053 0.248
0.018*
(0.008) 0.111 0.018
Persona Gender (Female) 0.100
(0.077) 0.059 0.195 0.081
(0.151) 0.024 0.591 0.190
(0.113) 0.078 0.092 0.110
(0.094) 0.054 0.240 0.119
(0.118) 0.046 0.315
Persona Ethnicity (Middle
Eastern origin)
-0.059
(0.095) -0.033 0.533 0.458*
(0.186) 0.127 0.014 0.064
(0.139) 0.025 0.644 -0.023
(0.115) -0.011 0.842 -0.009
(0.145) -0.003 0.950
Persona Ethnicity (African
origin)
0.090
(0.095) 0.050 0.347 0.423*
(0.187) 0.117 0.024 0.055
(0.139) 0.021 0.693 0.076
(0.116) 0.035 0.510 0.117
(0.145) 0.043 0.421
Participant Ethnicity (Middle
East and Africa)
0.341
(0.192) 0.081 0.077 0.089
(0.376) 0.011 0.812 0.467
(0.281) 0.077 0.097 0.150
(0.233) 0.029 0.520 0.556
(0.293) 0.088 0.059
Participant Ethnicity (Asian
origin)
-0.388**
(0.150) -0.119 0.010 -0.308
(0.294) -0.047 0.295 -0.137
(0.220) -0.029 0.534 -0.498**
(0.182) -0.125 0.007 -0.117
(0.230) -0.024 0.609
Participant Ethnicity (Latinx) 0.156
(0.144) 0.050 0.280 0.622*
(0.282) 0.099 0.028 0.319
(0.210) 0.071 0.130 -0.010
(0.175) -0.003 0.954 0.284
(0.220) 0.060 0.197
Participant Gender (Non-
binary / third gender)
-0.540
(0.595) -0.041 0.365 -0.677
(1.165) -0.026 0.561 -1.320
(0.870) -0.070 0.130 -1.611*
(0.722) -0.102 0.026 -0.505
(0.908) -0.026 0.579
Participant Gender (Male) -0.112
(0.080) -0.065 0.159 -0.131
(0.156) -0.037 0.402 -0.100
(0.117) -0.040 0.392 -0.187
(0.097) -0.089 0.053 -0.082
(0.122) -0.031 0.502
Perceived Happiness -0.024
(0.021) -0.052 0.252 -0.251***
(0.040) -0.276 0.000 -0.033
(0.030) -0.051 0.274 -0.063*
(0.025) -0.114 0.013 -0.029
(0.031) -0.043 0.357
Notes: *** p < 0.001; ** p < 0.01; * p < 0.05. Standard errors are in parenthesis. Persona Gender reference category is “Male”. Persona Ethnicity reference category is “White”. Participant
Ethnicity reference category is “European origin”. Participant Gender reference category is “Female”.
Table 5: Regression model coeicients for personality traits. A positive coeicient between the picture happiness variable (last row) and other variables indicates that as
picture happiness increases, personality traits generally increase. See Table 1 for variable names and definitions.
EX AG CN ES OP
Variable B / SE Beta p B / SE Beta p B / SE Beta p B / SE Beta p B / SE Beta p
Participant Age 0.001
(0.006) 0.006 0.892 0.014*
(0.006) 0.111 0.011 0.007
(0.008) 0.041 0.366 0.011
(0.007) 0.070 0.090 0.002
(0.006) 0.018 0.698
Persona Gender (Female) 0.062
(0.089) 0.028 0.490 0.218*
(0.087) 0.108 0.012 0.373**
(0.124) 0.134 0.003 0.088
(0.102) 0.035 0.388 0.138
(0.091) 0.069 0.128
Persona Ethnicity (Middle
Eastern origin)
-0.434***
(0.110) -0.189 0.000 -0.109
(0.107) -0.051 0.309 -0.017
(0.153) -0.006 0.912 -0.073
(0.126) -0.027 0.566 -0.175
(0.112) -0.082 0.118
Persona Ethnicity (Black) -0.029
(0.110) -0.012 0.796 0.025
(0.107) 0.012 0.818 -0.179
(0.153) -0.061 0.243 0.221
(0.126) 0.083 0.081 0.022
(0.112) 0.010 0.845
Participant Ethnicity (Middle
East and Africa)
0.122
(0.222) 0.023 0.582 -0.121
(0.216) -0.024 0.576 -0.074
(0.309) -0.011 0.810 -0.226
(0.255) -0.036 0.375 0.127
(0.226) 0.026 0.573
Participant Ethnicity (Asian
origin)
-0.168
(0.174) -0.040 0.333 -0.174
(0.169) -0.044 0.305 -0.270
(0.241) -0.050 0.264 -0.242
(0.199) -0.050 0.225 -0.420*
(0.177) -0.108 0.018
Participant Ethnicity (Latinx) 0.255
(0.166) 0.064 0.126 -0.019
(0.162) -0.005 0.906 0.074
(0.231) 0.015 0.748 0.025
(0.191) 0.005 0.897 0.235
(0.169) 0.063 0.165
Participant Gender (Non-
binary / third gender)
0.600
(0.688) 0.036 0.384 -0.114
(0.670) -0.007 0.864 -0.266
(0.956) -0.012 0.781 0.358
(0.789) 0.018 0.650 -0.479
(0.699) -0.031 0.494
Participant Gender (Male) -0.014
(0.092) -0.006 0.883 -0.235**
(0.090) -0.114 0.009 -0.123
(0.128) -0.043 0.338 -0.079
(0.106) -0.030 0.456 -0.208*
(0.094) -0.101 0.027
Perceived Happiness 0.245***
(0.024) 0.426 0.000 0.180***
(0.023) 0.334 0.000 0.178***
(0.033) 0.242 0.000 0.320***
(0.027) 0.476 0.000 0.094***
(0.024) 0.176 0.000
Notes: *** p < 0.001; ** p < 0.01; * p < 0.05. Standard errors are in parenthesis. Persona Gender reference category is “Male”. Persona Ethnicity reference category is “White”. Participant
Ethnicity reference category is “European origin”. Participant Gender reference category is “Female”.
4.3 RQ2 Eect of Persona Gender and Ethnic Background
Two signicant eects were found regarding Persona Gender: the participants perceived female personas as more
agreeable (B = 0.218, p < .05) and conscientious (B = 0.373, p < .01) than male personas (see Figure 6). For ethnicity, the
pain points of personas of Middle Eastern origin were perceived more strongly (B = 0.458, p < .05) than those of
personas of European origin (see Figure 7). Interestingly, these personas were also perceived as less extroverted (B = -
0.434, p < .001) when compared to personas of European origin. e pain points of personas of African origin were also
perceived more strongly (B = 0.423, p < .05) than those of personas of European origin. Apart from these ndings, the
observed eects were consistent among persona genders and ethnic backgrounds.
Empathy
PP
Usefulness
Realism
Completeness
Extraversion
Agreeableness
Conscien%ousness
Stability
Openness
0
1
2
3
4
5
6
7
Female
Male
Score
Figure 6: Perception scores for male and female personas. Female personas were perceived as more agreeable and conscientious.
Empathy
PP
Usefulness
Realism
Completeness
Extraversion
Agreeableness
Conscien%ousness
Stability
Openness
0
1
2
3
4
5
6
7
Midd le Eastern
African
European
Score
Figure 7: Perception scores for the three persona ethnicities. Pain points (PP) of personas of Middle Eastern and African origin were
observed more strongly than those of personas of European origin.
4.4 RQ2 Eect of Participant Gender and Ethnic Background
Participants of Asian origin indicated less empathy for the personas ( B = -0.388, p < .01) and perceived the personas as
less realistic (B = -0.498, p < .05) and open (B = -0.420, p < .05) than participants of European origin (see Figure 8).
Participants of Latin origin tended to perceive the pain points of a persona more strongly (B = 0.622, p < 0.05) than
those of European origin. Additionally, male participants tended to view personas as less agreeable when compared to
female participants (B = -0.235, p < 0.01), as well as less open to experience (B = -0.208, p < 0.05) (see Figure 9). Apart
from these eects, participants with dierent genders or ethnic backgrounds provided consistent responses.
Figure 8: Persona scores by participant ethnic background. Participants of Asian origin indicated less empathy for personas and
viewed them as less realistic and open. Participants of Latin origin perceived the personas’ pain points more strongly.
Empathy
PP
Usefulness
Realism
Completeness
Extraversion
Agreeableness
Conscien%ousness
Stability
Openness
0
1
2
3
4
5
6
7
Female
Male
Score
Figure 9: Perception scores for male and female participants. Male participants viewed the personas as less agreeable and open.
4.5 RQ3: Users’ Engagement with the Persona
H7: Picture of happiness increases the users’ engagement with the persona. To address our third research
question, namely, whether participants spend less time viewing unhappy personas than happy ones, we conducted a
set of t-tests to compare the amount of time (in seconds) the participants spent with each persona type in the survey.
is information was recorded by an invisible form eld in altrics that enables observing participants’ time spent
16
on a given survey page. We included this question format in each persona page, and thus, we could retrieve the time
spent with “happy” and “unhappy” personas. is duration was used as a measure of engagement.
We used a t-test variant whose assumptions are robust to unequal variances (two-sample t-test assuming unequal
variances [67]) as the groups had unequal variances. ere was no signicant eect for duration, t(303.113) = -1.3, p =
0.194, despite “happy” personas (M = 68.4 seconds, SD = 166.3) aaining higher scores than “unhappy” personas (M =
53.2, SD = 63.9). is test used the assigned happy/unhappy persona treatments. e participants spent somewhat
more time with the happy personas, but this dierence was not statistically signicant. erefore, H7 was not
supported; the use of happy or unhappy persona pictures does not aect the level of engagement (i.e., duration of
interaction) with the persona.
4.6 RQ4: User Initial Solution Designs to a Persona’s Pain Points
e goal of the qualitative analysis is to investigate how the use of (un)happy persona pictures alters the initial design
outcomes in a user-centric product development task, specically the depth and breadth of the designed solutions that
the participants immediately came up with to remedy the pain points of happy versus unhappy personas. Since there
is no direct measure to assess and compare the eectiveness and depth of the initial design solutions, we relied on
several qualitative methods that analyzed the participants’ responses that describe their solutions. We surmised that
the initial design outcomes could be qualitatively compared by answering questions such as the following: (1) Did the
participants feel empathy toward the personas? (2) Did the participants understand the pain points of the personas? (3) Did
the participants nd the persona information useful in coming up with a solution design? (4) Did the participants nd the
persona realistic?
To address these questions, we performed a qualitative semantic coding of the data through a lexical analysis
[11,21] based on the calculation of word frequencies. We also qualitatively revisited the textual data based on the
context and semantically grouped lexicons, as needed. In our analysis, we qualitatively picked or determined lexicon
groups that could be associated with our research questions and ordered the words in the lexicon based on their
contextual valence [80]. Valence scoring can be done by a group of researchers based on their sense of positive versus
negative context [80] or by a single researcher [53]. Since all our semantic lexicons only contained a small number of
words (i.e., descriptors), a single researcher ordered their valence using their contextual importance to the research
question.
4.6.1 Expressions of Empathy
Our rst premise was aimed at seeing the eects of the persona pictures on participants’ empathy toward the persona.
To investigate this question further, we focused on the persona-address descriptors—the specic words, names, and
adverbs the participants used to address the persona in their wrien responses. We ordered the descriptors by their
potential to show empathy, ranging from “no mention,” which shows the least empathy, to “you,” which shows the
most empathy. e results are given in Table 6. Examples of each category are given below:
No mention of or bypassing the persona: e participant’s text either does not address the persona in
any way or bypasses the persona. Typically, the comments that make no mention of the personas only focus
on the product and its functionalities (e.g., “A system that would separate the work environment on a
computer from the personal environment. […] e system can only be turned on the following morning
during oce hours. e system would automatically save all work in progress, so no work is lost during the
automatic shutdown.” Participant #308 or P308 for short). Occasionally some comments bypass the persona
17
by extrapolating their product to a group of people instead of the specic persona provided (e.g., “My
product is an application for people who are not familiar with the use of technology.” P140).
Persona (or Person, Individual, User, Character, Customer) coupled with It or ey5: “is persona
needs a soware […that] could provide detailed information on how they are spending their workday so
they can see the paerns and improve their performance overall” (P1).
Persona (Person, Guy, Lady) coupled with He or She: “is persona has several issues about smart
working because he barely can separate work from everyday life” (P6); “is guy denitely needs a router
[…]” (P57); “I think routine and structure day will help this lady.” (P77)
First-name coupled with He or She: “[…] James is presented as very unhappy and struggling to balance
his work and home lives.” (P7); “I think the issues Sarah has are mainly service issues and lack of computer
usage experience. So, my solution would be a program installed in her computer […]” (P66)
You: “You don’t need a degree to start this business. However, you should polish your skills rst.” “Try
enrolling in an online course or reading graphic design books.” (P4); “Calm Down Buddy! is app is for you.
If you’re struggling separating work from leisure well, you’re not alone.” (P89)
Table 6: The frequency of persona-address descriptors based on specific personas. Percentages above 25% are highlighted. A
weighted score was also calculated, with “no mention” being worth 0 points (minimum) and “you” being worth 4 points at 1-point
increments for each category.
Persona Happines
s
No
Mention
Persona w/
It or ey
Persona w/
He or She
First-name
w/ He or She
You Weighted
Score (0-4)
James
Unhappy 5.1%
(n=2)
30.8%
(n=12)
48.7%
(n=19)
10.3%
(n=4)
5.1%
(n=2)
1.79
Happy 20.5%
(n=8)
17.9%
(n=7)
41.1%
(n=16)
17.9%
(n=7)
2.6%
(n=1)
1.64
Sarah
Unhappy 7.7%
(n=3)
15.4%
(n=6)
46.2%
(n=18)
28.2%
(n=11)
2.6%
(n=1)
2.03
Happy 7.5%
(n=3)
20.0%
(n=8)
50.0%
(n=20)
15.0%
(n=6)
7.5%
(n=3)
1.95
Jane
Unhappy 15.8%
(n=6)
21.1%
(n=8)
57.9%
(n=22)
2.6%
(n=1)
2.6%
(n=1)
1.55
Happy 20.0%
(n=8)
15.0%
(n=6)
47.5%
(n=19)
10%
(n=4)
7.5%
(n=3)
1.70
Khaled
Unhappy 10.0%
(n=4)
40.0%
(n=16)
35.0%
(n=14)
15.0%
(n=6)
0 1.55
Happy 18.4%
(n=7)
13.2%
(n=5)
50.0%
(n=19)
13.2%
(n=5)
5.3%
(n=2)
1.74
Roger
Unhappy 21.4%
(n=9)
19.1%
(n=8)
52.4%
(n=22)
4.8%
(n=2)
2.4%
(n=1)
1.48
Happy 18.9%
(n=7)
5.4%
(n=2)
51.4%
(n=19)
21.6%
(n=8)
2.7%
(n=1)
1.84
Hind
Unhappy 16.2%
(n=6)
13.5%
(n=5)
54.1%
(n=20)
13.5%
(n=5)
2.7%
(n=1)
1.73
Happy 14.3%
(n=6)
14.3%
(n=6)
61.9%
(n=26)
7.1%
(n=3)
2.4%
(n=1)
1.69
5 Although we recognize that singular “they” can used as a generic third-person pronoun [36] or, purposefully, to
avoid misgendering, in our persona cases wherein gender was specied clearly, we accepted it as a form of distancing
language.
18
e average weighted score for unhappy personas was x = 1.69, and for happy personas, it was x = 1.76, which was
a marginal dierence. e outliers were unhappy James and Khaled scoring lower and unhappy Sarah scoring higher,
thus, falling outside the mean. is led us to believe that there might be a gender-based treatment dierence. Men
scored lower than women both in the unhappy category (x = 1.61 vs. x = 1.77) and overall (x = 1.67 vs. x = 1.77). We
see this as an indication that unhappy female visuals (or female avatars in general) garner more empathy than their
male counterparts [38,79].
4.6.2 Expressions of Pain Points
Our second premise was aimed at seeing the eects of the persona pictures on participants’ perceptions of the
persona’s pain points. To investigate this question further qualitatively, we created a qualitative codebook [39] by
focusing on the verbs and modal verbs mobilized by the participants as pain-point descriptors in all the responses. We
based the grouping of the codebook on the valence scores of descriptors in an existing framework [53] by checking
how far away they are from the neutral score, whether in a positive or negative direction. As a result, we ended up
with three categories: weak valence, moderate valence, and strong valence. A researcher read all the comments and
coded each relevant sentence into the proper category. Although the coding predominantly followed the verb/valence
groupings that inspired the categories, the advantage of the qualitative assessment of each sentence was that it helped
evaluate some usage more correctly depending on context. e results are provided in Table 7. An example of each
category is given below:
Weak valence (imperatives, should, must, etc.): “Product for him should be complex with many functions.”
(P13); “e person should keep the baby at childcare center […]” (P47); “[…] this persona should work on a product or
organization where he could spend most of his day without having to rely on internet” (P53).
Moderate valence (need, benet, require, suggest, provide, come in handy, could use, be useful, etc.): “He
might benet from some kind of scheduling assistant/app […]” (P17); “I would suggest a phone app to time work and
leisure activities to make sure that Sarah does not get carried away […]” (P24); “I think she needs a product to beer
split her day between work and personal activities” (P21).
Strong valence (help, support, struggle, suer, having [a problem, issue, or trouble], etc.): “I think that I
could help him by giving him a way to get isolated from everyday life” (P6); “e persona struggles with self-discipline
and organizational issues” (P9); “It's important to see what her problems are” (P122).
Table 7: The frequency of pain point descriptors based on specific personas. The dominant category(s) is highlighted. A weighted
score was also calculated, with “weak valence” being worth 0 points (minimum) and “strong valence” being worth 2 points, with a 1-
point increment for each category.
Weak Valence Moderate Valence Strong Valence Weighted
Score (0-2)
James Unhappy 23.1 % (n=9) 35.9% (n=14) 41.0% (n=16) 1.18
Happy 46.2% (n=18) 23.1% (n=9) 30.8% (n=12) .85
Sarah Unhappy 25.6% (n=10) 38.5% (n=15) 35.9% (n=14) 1.10
Happy 27.5% (n=11) 25.0% (n=10) 47.5% (n=19) 1.20
Jane Unhappy 36.8% (n=14) 28.9% (n=11) 34.3% (n=13) .97
Happy 42.5% (n=17) 35.0% (n=14) 22.5% (n=9) .80
Khaled Unhappy 52.5% (n=21) 22.5% (n=9) 25.0% (n=10) .73
Happy 50.0% (n=19) 21.1% (n=8) 28.9% (n=11) .79
Roger Unhappy 45.2% (n=19) 33.4% (n=14) 21.4% (n=9) .76
19
Weak Valence Moderate Valence Strong Valence Weighted
Score (0-2)
Happy 32.4% (n=12) 45.9% (n=17) 21.7% (n=8) .89
Hind Unhappy 40.5% (n=15) 40.5% (n=15) 19.0% (n=7) .78
Happy 52.4% (n=22) 33.4% (n=14) 14.2% (n=6) .62
e average weighted score for unhappy personas was x = 0.92 and for happy personas x = 0.86, another marginal
dierence that highlights that participants were spoke more strongly about the pain points of the unhappy personas.
ere were some stark dierences between individual personas. e problems of unhappy James and Jane were more
recognized, but the problems of happy James and Jane were mentioned less oen. e pain points of Khaled and Hind
were not voiced, regardless of whether they were happy or unhappy. Sarah’s pain points, on the other hand, were
more strongly voiced both when happy and unhappy. is led us to check the results from a racial perspective (see
Table 8).
e results conrmed that the pain points of personas of Middle Eastern origin were not recognized (or were
handled in a weaker way) regardless of whether they looked happy or unhappy. e pain points of personas of
European origin were more easily recognized when they were unhappy; however, the pain points of personas of
African origin were voiced even if they were not unhappy. In other words, (1) it was easy to imagine personas of
European origin struggling when they look unhappy; (2) it was easy to imagine personas of African origin struggling
even if they look happy; (3) it was not easy to imagine the struggles of personas of Middle Eastern origin at all. is
could be explained by bias brought by the participants. Previous studies suggest that race in prole pictures aects
interactions in virtual environments such as Airbnb [17] and online loan services [63].
Table 8: Results of pain-point descriptor analysis from a racial perspective.
Weighted Score (0-
2)
Personas of European
origin
Personas of Middle Eastern
origin
Personas of African origin
Unhappy 1.08 0.75 0.93
Happy 0.82 0.70 1.05
4.6.3 Expressions of Usefulness
e third premise investigated the eects of the persona pictures on participants’ perceptions of the usefulness of the
persona. Although our survey included a self-reported usefulness item for the personas, we wanted to bolster our
results with an additional response analysis for further insights. We accepted usefulness in relation to the details and
depth of the solution and/or product that the participants could come up with. e more descriptive the participants
could get about their solution and/or product, the more useful we considered the persona prole to be. e two
shortcomings of this method are that (1) there is no previous research that mobilizes response length in relation to the
usefulness of a persona, and (2) in some cases, concise answers might be perceived as more meaningful textual
responses. However, previous studies [3,81] oer response length as an imperfect counterpart for the quality of
responses in various survey contexts. Accordingly, we used two types of data to check this: we looked rst at the
percentages of sentences in each response that were wrien to explain the product (thus, the subject of the sentence
was the product/solution) and second at the word count of each product description section. e results are given in
Table 9.
20
Table 9: The percentages of the responses that relate to a solution and/or product description, as well as word counts.
Persona Happiness Percentage (%) Mean (x) Word Count Mean (x)
James Unhappy 63.14 50.86
Happy 57.78 46.94
Sarah Unhappy 74.78 73.27
Happy 54.45 40.87
Jane Unhappy 43.81 38.54
Happy 54.89 34.07
Khaled Unhappy 68.89 59.60
Happy 60.89 56.60
Roger Unhappy 59.23 41.20
Happy 29.36 22.54
Hind Unhappy 50.19 41.47
Happy 49.45 41.94
Male Personas Unhappy 61.96 49.25
Happy 54.82 45.65
Female Personas Unhappy 56.01 51.25
Happy 53.17 38.80
Personas of European
origin
Unhappy 50.80 42.74
Happy 60.66 45.82
Personas of African
origin
Unhappy 67.00 57.24
Happy 41.90 31.70
Personas of Middle
Eastern origin
Unhappy 59.17 50.77
Happy 55.54 49.04
All Personas Unhappy 58.99 50.25
Happy 54.15 42.85
Looking at all the responses, almost half of the text addressed a solution and/or product for the persona, with
unhappy personas sparking a marginally higher density of text both in percentage and word count (58.99% versus
54.15%; 50.25 words versus 42.85 words). One of the outliers was happy Roger, who had signicantly lower solutions
and/or products oered to him. at result qualitatively marks him as a less useful persona in terms of soliciting
solutions from the participants. One explanation could be that since Roger looked happy, so participants did not feel
the need to create solutions for his problems. However, this does not explain why that was not the case for other
happy personas. e other explanation could involve Roger’s racial identity; however, that does not explain the case of
happy Sarah, who is also of African origin. As a result, we believe that this stands at an intersection of persona picture,
gender, and racial identity. Other outliers were unhappy Sarah and Khaled, who had more breadth in their solution
discussions. e density of discussions around unhappy Sarah is signicant and, again, seems to sit at an intersection
of persona picture, gender, and racial identity. Results from all personas showed that happy personas of African origin
and happy female personas had shorter descriptions of solutions and/or products. However, the imbalance between
Sarah and Roger (both personas of African origin) remains unexplained.
4.6.4 Expressions of Realism, Completeness, and Personality
For our H4–H6 (which are aimed at seeing the eects of the persona pictures on participants’ perceptions of a
persona’s realism, completeness, and personality, respectively), we studied the 471 wrien responses, identied and
21
grouped those that mention the said aspects, and coded each group into subcategories. e results are given in Table
10 and Table 11 along with detailed information on each coding.
4.6.4.1 Realism
We coded the segments about the realism of the personas under ve categories: storifying, disbelief, environment,
family, and baby.
Storifying emerges when the participants create additional details about the personas and their lives in ways that
were not mentioned, hinted at, or highlighted in the prole. For example, “As he is very procient and successful, and I
assume also rich then he surely has money for holiday” (P363) or “It make not help [sic] that the baby wakes oen in
the night and he is also tired which will have a negative eect on his work” (P14). Neither case has any information on
the persona’s wealth or sleeping paerns. Storifying is a phenomenon that was previously identied in other persona
studies [46,49,50]. Typically, if a user nds a persona realistic enough (e.g., the persona resembles somebody in the
user’s life or has other parallels to the user’s own life), then the user will start projecting real-life experiences and
stories onto the persona. We accept this as an indication of realism. In our data, most stories emerged for unhappy
male personas, with James being the dominant source. Happy personas displayed almost no storifying eect (7.1%).
Family and baby categories might be seen as an extension of storifying; however, since they have other
connotations, they were mentioned separately. e proles state that the persona has a 6-month-old baby. is
information is mentioned in the responses in relation to real-life experiences of how challenging it is to live with an
infant. Of the baby mentions, 63% were made for female personas—a cognitive gender bias that was identied
previously between motherhood and labor prospects of women [9]. An interesting reversal was that baby mentions
were more prominent for unhappy female versus happy male personas. Also, there were some instances where the
personas were assumed to have other family members in the form of partners. is was more common for male
personas assumed to have wives (67%) who could/should take care of the baby while the father works.
Table 10: The results of coding in text segments relating to the realism of the personas.
Persona Segments
(N=130)
Storifying
10.8%
(n=14)
Disbelief
11.5%
(n=15)
Environment
18.5%
(n=24)
Family
13.8%
(n=18)
Baby
45.4%
(n=59)
James Unhappy 6 3 2
Happy 2 3 3
Sarah Unhappy 2 1 5 5
Happy 4
Jane Unhappy 1 2 2 10
Happy 3 2 6
Khaled Unhappy 2 2 1 2
Happy 1 1 1 3 5
Roger Unhappy 2 2 3 5
Happy 72 2 5
Hind Unhappy 5 3 6
Happy 1 1 6
James Unhappy
Happy
Male
Personas
Unhappy 10 7 4 9
Happy 1 10 3 8 13
Female Unhappy 3 1 12 5 21
22
Persona Segments
(N=130)
Storifying
10.8%
(n=14)
Disbelief
11.5%
(n=15)
Environment
18.5%
(n=24)
Family
13.8%
(n=18)
Baby
45.4%
(n=59)
Personas Happy 4 2 1 16
e environment coding mentioned the physical working conditions of the personas. Although there was no
textual information about the physical working conditions of the personas, some participants mobilized the content of
the persona photo to make suggestions. e suggestions ranged from using a beer chair or table to being in a room
with a door where isolation would be easier. ese codes were more prominent for unhappy personas and especially
for unhappy women. An interesting side note is three participants claiming that unhappy Sarah and Jane both work at
a kitchen table—neither photo shows the women in the kitchen. is is possibly another cognitive slip driven by
gender bias.
Finally, we found some instances where participants voiced disbelief about the persona, for example, “I nd it hard
to imagine that a successful manager can have not so good [sic] computer skills” (P259), “I do not understand how it is
possible for someone in a management position to work remotely with good result [sic] without having a fast internet
connection” (P333), or “e fact that the persona looks happy makes me think they are okay with working remote
[sic]” (P424). ese instances were mostly focused on happy Roger (70% of the segments from male personas and 67%
of the segments from all personas). Previously, happy Roger was identied as the persona with the fewest solutions
oered.
4.6.4.2 Completeness
We found 26 segments about the completeness of the personas and coded them under two categories: company
information (30.8%, n = 8) and uncertainty (69.2%, n = 18).
e company information segments were all made for unhappy personas (Sarah 25%, Jane 12.5%, Khaled 37.5%, and
Hind 25%) and were complaints or assumptions that the personas’ companies should solve their problems either by
providing them with upgraded equipment or faster connections. Since our persona proles did not mention specic
information about their work and company relationships, this marked a perspective that might be perceived as
incomplete in the persona proles we used. e uncertainty segments marked instances when the participant was not
sure how to help the persona or mentioned that they had lile faith in their solution, such as “He could probably do
with some sort of training … but these don't really seem necessary” (P422), “It’s dicult to say because I don’t have
enough information about the person” (P421), or “I don’t know what kind of product it can be” (P392). ese segments
were mainly concentrated on unhappy Khaled (22.3%, n = 4) and happy Sarah (16.7%, n = 3), with happy Khaled and
Roger each having two, and happy/unhappy James and Jane and unhappy Sarah each having one.
4.6.4.3 Personality
To understand how the participants perceived the personas’ personalities, we looked for words that were used to
describe the personalities, and we found 59 segments that we coded to be either negative (55.9%, n = 33) or positive
(44.1%, n = 26).
Table 11: The results of coding in text segments relating to the personalities of the personas.
Persona Happiness Negative Positive
Male Personas Unhappy 15 4
Happy 4 13
23
Persona Happiness Negative Positive
Female Personas Unhappy 14 6
Happy 3
Personas of European origin Unhappy 14 4
Happy 1 4
Personas of African origin Unhappy 8 5
Happy 6
Personas of Middle Eastern origin Unhappy 7 1
Happy 3 6
All Personas Unhappy 29 10
Happy 4 16
As expected, negative personality descriptors were mostly focused on unhappy personas. Also, unhappy personas
of European origin had almost twice the number of negative descriptors than personas of African or Middle Eastern
origin, which was a surprising result. Although it is hard to pinpoint an exact reason for this racial divide, it may be
that men of European origin are seen as a more advantageous social category, which results in men of European origin
aracting more criticism when seen as unhappy. e most common negative descriptors were stressed (36.4%, n = 12)
and unhappy (9.1%, n = 3), with instances of afraid, anxious, distracted, and pessimism, among others.
Interestingly, both unhappy and happy personas received positive personality descriptors, with happy personas
having more (61.5%). is is mainly because, although happy men signicantly aracted more positive comments
(76.5%) among men, unhappy women aracted more positive comments (66.6%) among women. Much like the
explanation of the previous result, this may be due to women being seen as a comparatively disadvantageous category.
As a result, they received positive comments for support and encouragement when they looked unhappy. In terms of
origins, the unhappy Middle Eastern personas received a signicantly lower number of positive comments than the
happy ones. is was not the case for the happy and unhappy personas of African and European origins, which
received an equal distribution of positive comments. (us, there was less encouragement for unhappy personas of
Middle Eastern origin.) e commonly used positive descriptors were comfortable (7.8%, n = 2), reliable (7.8%, n = 2),
happy (7.8%, n = 2), and focused (7.8%, n = 2), with instances of motivated, open for learning, and successful, among
others.
ese results indicate that for an initial design task to benet from positive perceptions of personalities, a happy
man or an unhappy woman would be the most useful along the lines of gender, and a Middle Eastern persona would
be the least useful in terms of ethnicity. However, the second part could be translated as the need for using a persona
prole that is ethnically closer to the participants.
5 DISCUSSION
5.1 Theoretical Contributions
Determining “why and how personas work” and showing their real value for design outcomes [69] are quintessential
aims in design studies [83]. e quantitative results indicate that happier persona pictures signicantly decrease
realism and pain point perception. e less happy a persona looks, the more realistic it seems and the more intense its
pain points are perceived in a design task. For this part, the results support the use of unhappy pictures as a way to
enhance a designer’s immersive experience with personas (Design Implication 1). Furthermore, the results indicate
no downside of decreasing persona engagement when using unhappy pictures.
24
e fact that a happy picture decreases the users’ pain point perception is reasonable and expected. (If you are
happy, are you really having troubles?) e fact that picture happiness does not aect a persona’s usefulness implies
that the persona’s happiness is not directly related to usefulness. Similarly, our ndings suggest that happiness is not
related to the persona’s completeness. e fact that picture happiness decreases the realism of the persona is in line
with previous ndings concerning the eect of stock photos [72]. Unexpectedly, there was no statistically signicant
dierence between unhappy and happy personas. One would expect someone unhappy to evoke empathy, but this was
not the case.
Our results showed that the rst-impression “mood” of the persona aects how its personality is judged, extending
the work on personas and personality traits [4,5]. ese Big Five personality ratings maer since they may inuence
design outcomes, and dierent design solutions might be created for introverted personas compared to extroverted
ones.
As expected, users tend to extrapolate the persona’s personality from a single picture: therefore, the momentary
“mood” of the persona aects how its personality is judged. To lessen this eect—that is, to increase the range of
perception of personality—multiple pictures could be added to portray the persona in dierent moods (e.g., happy,
unhappy, neutral) (Design Implication 2). is could possibly alleviate the rst-impression eect of users associating
the persona with a specic (positive or negative) type of personality based on a single picture alone.
5.2 Cultural and Gender Eects
Personas provide a vehicle toward inclusive design by portraying users from dierent cultures [27] [13]. Our ndings
show that varying the persona’s origin can have real consequences for how the persona is applied to design tasks. e
qualitative analysis outlined the dierences that existed along racial lines—the easiest to recognize were the pain
points of personas of European origin, and the hardest to recognize were the pain points of personas of Middle Eastern
origin.
Unhappy Roger was found both less useful and less realistic. Although there was no clear explanation for this, we
believe that the phenomenon stands at the intersection of ethnic identity and gender. e participants were more
prone to (1) creating stories around unhappy male personas; (2) bringing up a baby and family for unhappy female
personas; (3) assuming the existence of partners for male personas; and (4) using negative personality descriptors for
unhappy men but, in contrast, positive personality descriptors for unhappy women.
ese results indicate that the participants tried to come up with explanations around the unhappiness of men and
used negative descriptors for them, but the participants were also quick to accept the unhappiness of women due to
having a baby and family duties and used positive descriptors for them. e participants were the least empathetic
toward the personas with unhappy male pictures and found it easier to empathize with unhappy female personas.
As such, the ndings tie back to the discussion on stereotypical thinking associated with personas and if
stereotypes are indeed inevitable when designing with personas [84]. Our ndings suggest that stereotyping takes
place; however, it is not certain if this aects the design outcomes negatively. It appears that the design outcomes can
be detailed despite culture and gender aecting their content. erefore, the question of stereotyping requires further
study to delineate when stereotyping becomes harmful and when it is acceptable (i.e., non-harmful) for design
outcomes.
Overall, we speculate that the gender stereotypes in society can explain some of the dierences we observed.
Characterizing this stereotype, we can conclude that society does not like unhappy (should we dare to say, weak) men
but wants to protect and support unhappy women. As a result, a design lesson could be as follows (Design
25
Implication 3): If the task requires the participants to empathize with the persona, use a happy man instead of an
unhappy one, but also use an unhappy woman instead of a happy one. On the other hand, unhappy white men are
described very negatively. It almost appears as if the participants were thinking, “You are already a member of an
advantaged social group; why are you complaining?”
Also, ethnicity is a factor. e pain point perception was the highest for the personas of Middle Eastern origin.
However, this does not necessarily mean that the participants sympathize with the personas. e fact that Middle
Eastern personas are seen to struggle with remote work could also be perceived from the point of view of racial
proling so that Middle Eastern people are seen as less capable with technology (a negative stereotype). In fact, the
qualitative ndings suggest that Middle Eastern personas are at a disadvantage of being understood and empathized
with. ese ndings could be explained by the participants’ ethnicities (predominance of people of European origin); if
more participants were from the Middle East, the results could have been dierent.
While more research is needed to ascertain to which degree ethnic (dis)similarity aect perceiving and using
personas for design, given our limited evidence, we would encourage persona creators to “play it safe” by ensuring
that the cultural match between the personas and participants is not too wide to negatively aect empathy (Design
Implication 4). is recommendation may be controversial since the purpose of personas is to increase empathy (e.g.,
bridge cultural gaps). Nonetheless, our results do not show this purpose being realized; hence, it is uncertain how well
personas mitigate stereotypes and how much they instead reinforce them.
5.3 Limitations and Future Research Directions
As with all research, there are limitations and areas for further analysis. A limitation of this work involves the limited
ethnic diversity of the participants: We would have preferred having more varied ethnic backgrounds, but the sample
mostly consisted of people of European origin. Future studies can recruit participants from other ethnic backgrounds.
Moreover, future research could experiment with more personas of dierent ethnic backgrounds (e.g., Asian and
Latinx). Due to practical limits imposed by our study design, we were constrained to three personas. We decided that
the inclusion of personas of Middle Eastern origin was important, as previous studies have applied personas of Asian
origin [63,64], but as far as we know, perceptions toward personas of Middle Eastern origin have not been tested in
previous research.
Future research could also investigate if and how “unhappy” personas stand out if mixed with “happy” personas. In
particular, unhappy personas may beer stand out from happy ones if the majority of the personas are happy (and
vice versa). e happy persona thus becomes more visible and will receive more aention and be beer remembered
than its visual surroundings [26]. is might have (un)intended consequences for the design process, depending on
whether the goal is to highlight a certain persona or to try to ensure the even aention of users to all personas. In
conjunction with this, additional persona perception aributes, such as likeness, could be studied. Also, instead of
images, the research could be expanded to investigate the eect of the textual information only expressing the pain
points with neutral images. Although we counterbalanced the text and images in this study, it would be interesting to
tease apart the eect of the text only.
Furthermore, the degree or nuances of unhappiness in the pictures could be varied. Most of the pictures we used
could be characterized as “moderately unhappy”; however, there is also a picture material where the person appears to
be strongly dissatised, even depressed (i.e., there is a wide range of emotion). e unhappiness can be associated with
sadness or frustration. e strength of the unhappiness could further aect how the persona is perceived; higher levels
26
of agony might result in heightened pain point perception. erefore, purposefully varying the levels and types of
unhappiness provides a logical continuum to our work.
While the ndings support the association between a persona’s happiness and personality aribution, how this
association aects design outcomes should be studied further. With our experiment seing [47] focusing primarily on
the eect of happy and unhappy pictures on perceptions, we were limited in only investigating the initial design
outputs [69]. Future research employing eld study methodologies could investigate the eect perceptions of personas
along a spectrum of tasks in the design process, especially communication within the design team and with other
stakeholders, in a more realistic seing [47]. Finally, future research should focus on varying the design task. It may be
that for some tasks, happy pictures are beer; for other tasks, unhappy ones may be beer. Task types may require
dierent levels of user empathy and pain point immersion, so the implications of the use of pictures might dier, even
when the personas contain pain points. In our research design, participants performed the tasks alone. Future research
that relies on executing the tasks in a group environment is also needed.
6 CONCLUSION
Should designers of personas use happy or unhappy pictures? It seems that it depends, as there are trade-os. Users
perceive the pain points of the persona more strongly when using unhappy pictures than when using happy pictures.
Users also nd personas with unhappy pictures more realistic than those with happy pictures. However, happy
pictures seem to present a more positive impression of the personas in terms of personality traits. In general, the
evidence supports the use of unhappy pictures to strengthen the users’ perception of personas as real people with real
problems.
REFERENCES
[1] Basant Agarwal. 2014. Personality detection from text: A Review. International Journal of Computer System 1, 1 (2014).
[2] American Psychological Association. 2020. Racial and ethnic identity. Racial and Ethnic Identity. Retrieved December 20, 2020 from
hps://apastyle.apa.org/style-grammar-guidelines/bias-free-language/racial-ethnic-minorities
[3] Mark Andrews. 2005. Who is being heard? Response bias in open-ended responses in a large government employee survey. In 60th Annual
Conference of the American Association for Public Opinion Research, Miami Beach, FL.
[4] F. Anvari, D. Richards, M. Hitchens, and M. A. Babar. 2015. Eectiveness of Persona with Personality Traits on Conceptual Design. In 2015
IEEE/ACM 37th IEEE International Conference on Soware Engineering, 263–272. DOI:hps://doi.org/10.1109/ICSE.2015.155
[5] Farshid Anvari, Deborah Richards, Michael Hitchens, Muhammad Ali Babar, Hien Minh i Tran, and Peter Busch. 2017. An empirical
investigation of the inuence of persona with personality traits on conceptual design. Journal of Systems and Soware 134, (December 2017),
324–339. DOI:hps://doi.org/10.1016/j.jss.2017.09.020
[6] Farshid Anvari and Hien Minh Tri Tran. 2013. Persona ontology for user centred design professionals. In e ICIME 4th International Conference
on Information Management and Evaluation, Ho Chi Minh City, Vietnam, 35–44.
[7] M. Aoyama. 2007. Persona-Scenario-Goal Methodology for User-Centered Requirements Engineering. In Proceedings of the 15th IEEE
International Requirements Engineering Conference (RE 2007), Delhi, India, 185–194. DOI:hps://doi.org/10.1109/RE.2007.50
[8] Alixandra Barasch, Emma E. Levine, and Maurice E. Schweitzer. 2016. Bliss is ignorance: How the magnitude of expressed happiness inuences
perceived naiveté and interpersonal exploitation. Organizational Behavior and Human Decision Processes 137, (2016), 184–206.
[9] Stephen Benard, In Paik, and Shelley J. Correll. 2007. Cognitive bias and the motherhood penalty. Hastings LJ 59, (2007), 1359.
[10] Ellen Berscheid and Elaine Walster. 1974. Physical Aractiveness11Preparation of this paper was facilitated by National Institute of Health
Grant MH 16729, and National Science Foundation Grants GS-35157X and GS-30822X. We should also like to thank Dr. John Arrowood and Dr.
Marshall Dermer for their helpful comments. In Advances in Experimental Social Psychology, Leonard Berkowitz (ed.). Academic Press, 157–215.
DOI:hps://doi.org/10.1016/S0065-2601(08)60037-4
27
[11] Richard Bolden and Jean Moscarola. 2000. Bridging the quantitative-qualitative divide: the lexical approach to textual data analysis. Social
science computer review 18, 4 (2000), 450–460.
[12] Susanne Buecker, Marlies Maes, Jaap JA Denissen, Maike Luhmann, and Odilia M. Laceulle. 2020. Loneliness and the Big Five Personality Traits:
A Meta–Analysis. European Journal of Personality 34, 1 (2020), 8–28.
[13] Daniel G. Cabrero, Heike Winschiers-eophilus, and José Abdelnour-Nocera. 2016. A Critique of Personas As Representations of “the Other” in
Cross-Cultural Technology Design. In Proceedings of the First African Conference on Human Computer Interaction (AfriCHI’16), ACM, Nairobi,
Kenya, 149–154. DOI:hps://doi.org/10.1145/2998581.2998595
[14] Souti Chaopadhyay, Ishita Prasad, Austin Z. Henley, Anita Sarma, and Titus Barik. 2020. What’s Wrong with Computational Notebooks? Pain
Points, Needs, and Design Opportunities. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1–12.
[15] Alan Cooper. 1999. e Inmates Are Running the Asylum: Why High Tech Products Drive Us Crazy and How to Restore the Sanity (1 edition ed.).
Sams - Pearson Education, Indianapolis, IN.
[16] Sarah D’Angelo and Darren Gergle. 2018. An eye for design: gaze visualizations for remote collaborative work. In Proceedings of the 2018 CHI
Conference on Human Factors in Computing Systems, 1–12.
[17] Benjamin Edelman, Michael Luca, and Dan Svirsky. 2017. Racial discrimination in the sharing economy: Evidence from a eld experiment.
American Economic Journal: Applied Economics 9, 2 (2017), 1–22.
[18] Paul Ekman and Wallace V. Friesen. 1971. Constants across cultures in the face and emotion. Journal of personality and social psychology 17, 2
(1971), 124.
[19] P. Ekmann. 1973. Universal facial expressions in emotion. Studia Psychologica 15, 2 (1973), 140.
[20] Denae Ford, Margaret-Anne Storey, omas Zimmermann, Christian Bird, Sonia Jae, Chandra Maddila, Jenna L. Butler, Brian Houck, and
Nachiappan Nagappan. 2020. A Tale of Two Cities: Soware Developers Working from Home During the COVID-19 Pandemic.
arXiv:2008.11147 [cs] (August 2020). Retrieved December 21, 2020 from hp://arxiv.org/abs/2008.11147
[21] Henrique Freitas, Jean Moscarola, and Milton Jenkins. 1998. Content and lexical analysis: a qualitative practical application. ISRC, Merrick School
of Business, University of Baltimore (MD, EUA), WP ISRC 070498 (1998), 35.
[22] Teri Kwal Gamble and Michael W. Gamble. 2013. Interpersonal communication: Building connections together. Sage Publications.
[23] Samuel D. Gosling, Peter J. Rentfrow, and William B. Swann. 2003. A very brief measure of the Big-Five personality domains. Journal of
Research in personality 37, 6 (2003), 504–528.
[24] Katharine H. Greenaway and Elise K. Kalokerinos. 2017. Suppress for success? Exploring the contexts in which expressing positive emotion can
have social costs. European Review of Social Psychology 28, 1 (2017), 134–174.
[25] Joseph F Hair, William C Black, Barry J Babin, and Rolph E Anderson. 2014. Multivariate data analysis. Pearson Education Limited, Essex.
[26] Jillian G. Hamilton. 2009. Identifying with an avatar: a multidisciplinary perspective. In Proceedings of the Cumulus Conference: 38o South:
Hemispheric Shis Across Learning, Teaching and Research, Swinburne University of Technology and and RMIT University.
[27] Richard Herrio and Birgie Geert Jensen. 2013. Students’ responses to inclusive design. Design studies 34, 4 (2013), 438–453.
[28] Ursula Hess and Shlomo Hareli. 2015. e role of social context for the interpretation of emotional facial expressions. In Understanding facial
expressions in communication. Springer, 119–141.
[29] Charles G. Hill, Maren Haag, Alannah Oleson, Chris Mendez, Nicola Marsden, Anita Sarma, and Margaret Burne. 2017. Gender-Inclusiveness
Personas vs. Stereotyping: Can We Have it Both Ways? In Proceedings of the 2017 CHI Conference, ACM Press, Denver, Colorado, USA, 6658–
6671. DOI:hps://doi.org/10.1145/3025453.3025609
[30] Ying Hu, Asal Baragchizadeh, and Alice J. O’Toole. 2020. Integrating faces and bodies: Psychological and neural perspectives on whole person
perception. Neuroscience & Biobehavioral Reviews 112, (2020), 472–486.
[31] Chiungjung Huang. 2019. Social network site use and Big Five personality traits: A meta-analysis. Computers in Human Behavior 97, (2019),
280–290.
[32] Sampsa Hyysalo, Tatu Marila, So Perikangas, and Karoliina Auvinen. 2019. Codesign for transitions governance: Amid-range pathway
creation toolset foraccelerating sociotechnical change. Design Studies 63, (July 2019), 181–203. DOI:hps://doi.org/10.1016/j.destud.2019.05.002
[33] Bernard Jansen, Joni Salminen, Soon-gyo Jung, and Kathleen Guan. 2021. Data-Driven Personas (1st ed.). Morgan & Claypool Publishers.
Retrieved February 10, 2021 from hps://www.morganclaypool.com/doi/abs/10.2200/S01072ED1V01Y202101HCI048
28
[34] Edward E. Jones and Keith E. Davis. 1965. From acts to dispositions the aribution process in person perception. In Advances in experimental
social psychology. Elsevier, 219–266.
[35] Shirli Kopelman, Ashleigh Shelby Rosee, and Leigh ompson. 2006. e three faces of Eve: Strategic displays of positive, negative, and
neutral emotions in negotiations. Organizational Behavior and Human Decision Processes 99, 1 (2006), 81–101.
[36] Chelsea Lee. 2019. Welcome, singular “they.” hps://apastyle.apa.org. Retrieved January 21, 2021 from hps://apastyle.apa.org/blog/singular-
they
[37] Frank Long. 2009. Real or imaginary: e eectiveness of using personas in product design. In Proceedings of the Irish Ergonomics Society Annual
Conference, Irish Ergonomics Society Dublin.
[38] Esther MacCallum-Stewart. 2008. Real boys carry girly epics: Normalising gender bending in online games. Eludamos. Journal for Computer
Game Culture 2, 1 (2008), 27–40.
[39] Kathleen M. Maceen, Eleanor McLellan, Kelly Kay, and Bobby Milstein. 1998. Codebook development for team-based qualitative analysis.
Cam Journal 10, 2 (1998), 31–36.
[40] Florence H. Manning. 2014. On Leveraging the First Impression: Learning, Achievement Motivation, and the Design of Digital Tasks. PhD
esis. University of Illinois at Chicago.
[41] Terry Marks and Constance L. Hammen. 1982. Interpersonal mood induction: Situational and individual determinants. Motivation and Emotion
6, 4 (1982), 387–399.
[42] Nicola Marsden and Maren Haag. 2016. Stereotypes and politics: reections on personas. In Proceedings of the 2016 CHI Conference on Human
Factors in Computing Systems, 4017–4031.
[43] Nicola Marsden and Maren Haag. 2016. Evaluation of GenderMag personas based on persona aributes and persona gender. In International
Conference on Human-Computer Interaction, Springer, 122–127.
[44] Tara Mahews, Tejinder Judge, and Steve Whiaker. 2012. How do designers and user experience professionals actually perceive and use
personas? In Proceedings of the 2012 ACM annual conference on Human Factors in Computing Systems - CHI ’12, ACM Press, Austin, Texas, USA,
1219. DOI:hps://doi.org/10.1145/2207676.2208573
[45] Francis T. McAndrew and Chelsea Rae De Jonge. 2011. Electronic person perception: What do we infer about people from the style of their e-
mail messages? Social Psychological and Personality Science 2, 4 (2011), 403–407.
[46] Leslie Zebrowitz McArthur and David L. Post. 1977. Figural emphasis and person perception. Journal of Experimental Social Psychology 13, 6
(1977), 520–535.
[47] Joseph E. McGrath. 1995. Methodology Maers: Doing Research in the Behavioral and Social Sciences. In Readings in Human–Computer
Interaction, Ronald M. Baecker, Jonathan Grudin, William A. S Buxton and Saul Greenberg (eds.). Morgan Kaufmann, 152–169.
DOI:hps://doi.org/10.1016/B978-0-08-051574-8.50019-4
[48] Keith A McNeil, Isadore Newman, and Francis J Kelly. 1996. Testing research hypotheses with the general linear model. SIU Press.
[49] Roberta M. Melvin and Andrea Bunt. 2012. Designed for work, but not from here: rural and remote perspectives on networked technology. In
Proceedings of the Designing Interactive Systems Conference, 176–185.
[50] Tomasz Miaskiewicz and Kenneth A. Kozar. 2011. Personas and user-centered design: How can personas benet product design processes?
Design Studies 32, 5 (2011), 417–430.
[51] Tomasz Miaskiewicz, Tamara Sumner, and Kenneth A. Kozar. 2008. A latent semantic analysis methodology for the identication and creation
of personas. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, ACM, 1501–1510. Retrieved from
hp://dl.acm.org/citation.cfm?id=1357290
[52] Douglas C Montgomery, Elizabeth A Peck, and G Georey Vining. 2021. Introduction to linear regression analysis. John Wiley & Sons.
[53] Finn \AArup Nielsen. 2011. A new ANEW: Evaluation of a word list for sentiment analysis in microblogs. arXiv preprint arXiv:1103.2903 (2011).
[54] Lene Nielsen. 2002. From User to Character: An Investigation into User-descriptions in Scenarios. In Proceedings of the 4th Conference on
Designing Interactive Systems: Processes, Practices, Methods, and Techniques (DIS ’02), ACM, New York, NY, USA, 99–104.
DOI:hps://doi.org/10.1145/778712.778729
[55] Lene Nielsen. 2019. Personas - User Focused Design (2nd ed. 2019 edition ed.). Springer, New York, NY, USA.
29
[56] Lene Nielsen, Kira Storgaard Hansen, Jan Stage, and Jane Billestrup. 2015. A Template for Design Personas: Analysis of 47 Persona Descriptions
from Danish Industries and Organizations. International Journal of Sociotechnology and Knowledge Development 7, 1 (2015), 45–61.
DOI:hps://doi.org/10.4018/ijskd.2015010104
[57] Lene Nielsen, Soon-Gyo Jung, Jisun An, Joni Salminen, Haewoon Kwak, and Bernard J. Jansen. 2017. Who Are Your Users?: Comparing Media
Professionals’ Preconception of Users to Data-driven Personas. In Proceedings of the 29th Australian Conference on Computer-Human Interaction
(OZCHI ’17), ACM, New York, NY, USA, 602–606. DOI:hps://doi.org/10.1145/3152771.3156178
[58] Lene Nielsen and Kira Storgaard Hansen. 2014. Personas is applicable: a study on the use of personas in Denmark. In Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems, ACM, 1665–1674.
[59] James E. Nieters, Subbarao Ivaturi, and Iikhar Ahmed. 2007. Making personas memorable. In CHI ’07 extended abstracts on Human factors in
computing systems - CHI ’07, ACM Press, San Jose, CA, USA, 1817. DOI:hps://doi.org/10.1145/1240866.1240905
[60] Aki Nikolaidis, Diana Paksarian, Lindsay Alexander, Jacob DeRosa, Julia Dunn, Dylan M. Nielson, Irene Droney, Minji Kang, Ioanna Douka, and
Evelyn Bromet. 2020. e Coronavirus Health and Impact Survey (CRISIS) reveals reproducible correlates of pandemic-related mood states
across the Atlantic. medRxiv (2020).
[61] Stefan Palan and Christian Schier. 2018. Prolic. ac—a subject pool for online experiments. Journal of Behavioral and Experimental Finance 17,
(2018), 22–27.
[62] Eyal Peer, Laura Brandimarte, Sonam Samat, and Alessandro Acquisti. 2017. Beyond the Turk: Alternative platforms for crowdsourcing
behavioral research. Journal of Experimental Social Psychology 70, (2017), 153–163.
[63] Devin G. Pope and Justin R. Sydnor. 2011. What’s in a Picture? Evidence of Discrimination from Prosper. com. Journal of Human resources 46, 1
(2011), 53–92.
[64] John Prui and Jonathan Grudin. 2003. Personas: Practice and eory. In Proceedings of the 2003 Conference on Designing for User Experiences
(DUX ’03), ACM, San Francisco, California, USA, 1–15. DOI:hps://doi.org/10.1145/997078.997089
[65] Kari Rönkkö. 2005. An Empirical Study Demonstrating How Dierent Design Constraints, Project Organization and Contexts Limited the
Utility of Personas. In Proceedings of the Proceedings of the 38th Annual Hawaii International Conference on System Sciences - Volume 08 (HICSS
’05), IEEE Computer Society, Washington, DC, USA. DOI:hps://doi.org/10.1109/HICSS.2005.85
[66] Kari Rönkkö, Mats Hellman, Bria Kilander, and Yvonne Dirich. 2004. Personas is Not Applicable: Local Remedies Interpreted in a Wider
Context. In Proceedings of the Eighth Conference on Participatory Design: Artful Integration: Interweaving Media, Materials and Practices - Volume
1 (PDC 04), ACM, Toronto, Ontario, Canada, 112–120. DOI:hps://doi.org/10.1145/1011870.1011884
[67] Graeme D. Ruxton. 2006. e unequal variance t-test is an underused alternative to Student’s t-test and the Mann–Whitney U test. Behavioral
Ecology 17, 4 (2006), 688–690.
[68] Joni Salminen, Kathleen Guan, Lene Nielsen, Soon-gyo Jung, Shammur Absar Chowdhury, and Bernard J. Jansen. 2020. A Template for Data-
Driven Personas: Analyzing 31 antitatively Oriented Persona Proles. In Human Interface and the Management of Information. Designing
Information. HCII 2020., S. Yamamoto and H. Mori (eds.). Springer, Copenhagen, Denmark, 125–144.
[69] Joni Salminen, Soon-gyo Jung, Shammur Absar Chowdhury, Sercan Sengün, and Bernard J Jansen. 2020. Personas and Analytics: A
Comparative User Study of Eciency and Eectiveness for a User Identication Task. In Proceedings of the ACM Conference of Human Factors in
Computing Systems (CHI’20), ACM, Honolulu, Hawaii, USA. DOI:hps://doi.org/10.1145/3313831.3376770
[70] Joni Salminen, Soon-gyo Jung, Ahmed Mohamed Sayed Kamel, João M. Santos, and Bernard J. Jansen. 2020. Using articially generated pictures
in customer-facing systems: an evaluation study with data-driven personas. Behaviour & Information Technology 0, 0 (November 2020), 1–17.
DOI:hps://doi.org/10.1080/0144929X.2020.1838610
[71] Joni Salminen, Soon-gyo Jung, João M. Santos, and Bernard J. Jansen. 2019. e Eect of Smiling Pictures on Perceptions of Personas. In
UMAP’19 Adjunct: Adjunct Publication of the 27th Conference on User Modeling, Adaptation and Personalization, ACM, Larnaca, Cyprus.
DOI:hps://doi.org/10.1145/3314183.3324973
[72] Joni Salminen, Soon-gyo Jung, João M. Santos, and Bernard J. Jansen. 2019. Does a Smile Maer if the Person Is Not Real?: e Eect of a Smile
and Stock Photos on Persona Perceptions. International Journal of Human–Computer Interaction 0, 0 (September 2019), 1–23.
DOI:hps://doi.org/10.1080/10447318.2019.1664068
30
[73] Joni Salminen, Soon-gyo Jung, João M. Santos, Ahmed Mohamed Kamel, and Bernard J. Jansen. 2021. Picturing It!: e Eect of Image Styles on
User Perceptions of Personas. In Proceedings of ACM Human Factors in Computing Systems (CHI’21), ACM, Virtual conference, 1–6.
DOI:hps://doi.org/10.1145/3411764.3445360
[74] Joni Salminen, Haewoon Kwak, João M. Santos, Soon-gyo Jung, Jisun An, and Bernard J. Jansen. 2018. Persona Perception Scale: Developing and
Validating an Instrument for Human-Like Representations of Data. In Extended Abstracts of the 2018 CHI Conference on Human Factors in
Computing Systems - CHI ’18, ACM Press, Montreal QC, Canada, 1–6. DOI:hps://doi.org/10.1145/3170427.3188461
[75] Joni Salminen, Lene Nielsen, Soon-gyo Jung, Jisun An, Haewoon Kwak, and Bernard J Jansen. 2018. “Is More Beer?”: Impact of Multiple Photos
on Perception of Persona Proles. In Proceedings of ACM CHI Conference on Human Factors in Computing Systems (CHI2018), ACM, Montréal,
Canada. DOI:hps://doi.org/10.1145/3173574.3173891
[76] Joni Salminen, Rohan Gurunandan Rao, Soon-gyo Jung, Shammur A. Chowdhury, and Bernard J. Jansen. 2020. Enriching Social Media Personas
with Personality Traits: A Deep Learning Approach Using the Big Five Classes. In International Conference on Human-Computer Interaction,
Springer, 101–120.
[77] Joni Salminen, Joao M. Santos, Soon-gyo Jung, Motahhare Eslami, and Bernard J. Jansen. 2019. Persona Transparency: Analyzing the Impact of
Explanations on Perceptions of Data-Driven Personas. International Journal of Human–Computer Interaction 0, 0 (November 2019), 1–13.
DOI:hps://doi.org/10.1080/10447318.2019.1688946
[78] Joni Salminen, Joao M. Santos, Haewoon Kwak, Jisun An, Soon-gyo Jung, and Bernard J. Jansen. 2020. Persona Perception Scale: Development
and Exploratory Validation of an Instrument for Evaluating Individuals’ Perceptions of Personas. International Journal of Human-Computer
Studies 141, (April 2020), 102437. DOI:hps://doi.org/10.1016/j.ijhcs.2020.102437
[79] Sercan Sengün. 2015. Why do I fall for the elf, when i am no orc mysel? e ımplıcatıons of vırtual avatars ın dıgıtal communıcatıon.
Comunicação e Sociedade 27, (2015), 181–193.
[80] Mostafa Al Masum Shaikh, Helmut Prendinger, and Ishizuka Mitsuru. 2007. Assessing sentiment of text by semantic dependency and contextual
valence analysis. In International conference on aective computing and intelligent interaction, Springer, 191–202.
[81] Jolene D. Smyth, Don A. Dillman, Leah Melani Christian, and Mallory McBride. 2009. Open-ended questions in web surveys: Can increasing the
size of answer boxes and providing extra verbal instructions improve response quality? Public Opinion arterly 73, 2 (2009), 325–337.
[82] Norman Makoto Su and Gloria Mark. 2008. Designing for nomadic work. In Proceedings of the 7th ACM conference on Designing interactive
systems, 305–314.
[83] Cameron Tonkinwise. 2011. A taste for practices: Unrepressing style in design thinking. Design Studies 32, 6 (2011), 533–545.
[84] Phil Turner and Susan Turner. 2011. Is stereotyping inevitable when designing with personas? Design studies 32, 1 (2011), 30–44.
[85] Juliet A. Usher-Smith, Laragh LW Harvey-Kelly, Sabrina H. Rossi, Hannah Harrison, Simon J. Grin, and Grant D. Stewart. 2020. Acceptability
and potential impact on uptake of using dierent risk stratication approaches to determine eligibility for screening: A population-based
survey. Health Expectations (2020).
[86] Jan Van Dijk and Kenneth Hacker. 2003. e digital divide as a complex and dynamic phenomenon. e information society 19, 4 (2003), 315–
326.
[87] Nele M. Wessels, Johannes Zimmermann, Jeremy C. Biesanz, and Daniel Leising. 2020. Dierential associations of knowing and liking with
accuracy and positivity bias in person perception. Journal of Personality and Social Psychology 118, 1 (2020), 149.
[88] Christopher R. Wilkinson and Antonella De Angeli. 2014. Applying user centred and participatory design approaches to commercial product
development. Design Studies 35, 6 (2014), 614–631.
31
Appendix A – Persona treatments
Pain Point Profile 1
32
33
Pain Point Profile 2
34
35
Appendix B – Residual plots
36
37
38
39
... Expectedly responding to human behavior makes DAs closer to human beings, enhancing system UX among the user base [31,33,34]. Emotional diverse DAs need not only account for the positive experiences among the user base but also represent the pain points, challenges, and negative emotions experienced by the user groups they represent [60]. It is common that user representations, such as personas [50], include pain points, concerns, and issues of the user groups the personas represent; not only positive aspirations. ...
Conference Paper
Full-text available
When using deepfake technology to represent users, there is a need to convey a reasonable range of emotions to be able to portray different circumstances ranging from positive to negative experiences (e.g., personal struggles). Because it is not known how well deepfake avatars embody emotional diversity, we investigated this aspect among 202 deepfake avatars. Our findings suggest an overall positivity bias in deepfake avatars’ emotions. We also found significant gender differences in several emotional expressions, with, male deepfakes scoring higher in “smile” and “calm” emotions, and female deepfake avatars scoring higher in “surprised”, “fear”, and “happy” emotions. In terms of ethnicity, European and Hispanic deepfake avatars demonstrate the broadest range of “smile”, “happy”, and “calm” compared to other ethnic groups. Age had no notable bias. No emotion score was normally distributed, suggesting that the range of emotional representativeness among the tested deepfake avatars is skewed. We outline the implications for academics and professionals regarding future development and responsible deployment of deepfake avatars.
... While in this first iteration of Persona-L we haven't made deliberate decisions about Persona Image and we used generic cartoonish images to represent personas, our findings indicate that images do matter. [35] has noted that the facial expression used in persona profiles significantly affects user perceptions. Additionally, [41] indicates how persona's appearance can impact perceptions of trustworthiness and competence. ...
Preprint
Full-text available
We present Persona-L, a novel approach for creating personas using Large Language Models (LLMs) and an ability-based framework, specifically designed to improve the representation of users with complex needs. Traditional methods of persona creation often fall short of accurately depicting the dynamic and diverse nature of complex needs, resulting in oversimplified or stereotypical profiles. Persona-L enables users to create and interact with personas through a chat interface. Persona-L was evaluated through interviews with UX designers (N=6), where we examined its effectiveness in reflecting the complexities of lived experiences of people with complex needs. We report our findings that indicate the potential of Persona-L to increase empathy and understanding of complex needs while also revealing the need for transparency of data used in persona creation, the role of the language and tone, and the need to provide a more balanced presentation of abilities with constraints.
... Healthcare and Therapy [8,29,39,74,96,110,114,116,120,129,166,173] Children and Education [1,3,11,41,55,68,72,92,102,105,106,131,147,150,153,160,162,163,171,177] Design and Development [7,51,75,86,91,95,101,111,118,136,137,139,158,159], Social Interaction and Communication [4,11,26,27,42,48,61,66,73,79,87,115,128,141,152,157,170] Accessibility and Inclusion [12, 19, 53, 80-82, 84, 90, 94, 97, 113, 143, 165, 174] Virtual and Conversational Agents [5,25,52,63,64,104,112,117,126,134,138,149,162,164] Human-Robot Interaction [6,16,31,140,144,151,172] Digital and Virtual Environments [15,70] Transportation [13,98,130,161,176] Empathic Technology Development [32,35,54,57,67,85,135,167] Miscellaneous [18,24,30,76,78,122,145,154,156,175] ...
Article
Full-text available
Empathy is considered a crucial construct within HCI and CSCW, yet our understanding of this complex concept remains fragmented and lacks consensus in existing research. In this scoping review of 121 articles from the ACM Digital Library, we synthesize the diverse perspectives on empathy and scrutinize its current conceptualization and operationalization. In particular, we examine the various interpretations and definitions of empathy, its applications, and the methodologies, findings, and trends in the field. Our analysis reveals a lack of consensus on the definitions and theoretical underpinnings of empathy, with interpretations ranging from understanding the experiences of others to an affective response to the other's situation. We observed that despite the variety of methods used to gauge empathy, the predominant approach remains self-assessed instruments, highlighting the lack of novel and rigorously established and validated measures and methods to capture the multifaceted manifestations of empathy. Furthermore, our analysis shows that previous studies have used a variety of approaches to elicit empathy, such as experiential methods and situational awareness. These approaches have demonstrated that shared stressful experiences promote community support and relief, while situational awareness promotes empathy through increased helping behavior. Finally, we discuss a) the potential and drawbacks of leveraging empathy to shape interactions and guide design practices, b) the need to find a balance between the collective focus of empathy and the (existing and dominant) focus on the individual, and c) the careful testing of empathic designs and technologies with real-world applications.
... 642-663;Grudin & Pruitt, 2002). This emotional connection can lead to greater empathy for the user group the persona represents (Salminen, Ş engün, Santos, Jung, & Jansen, 2022). For example, a study found that, when using pictures where the personas looked unhappy, users perceived such persona profiles as more realistic and containing more severe pain points . ...
Article
HCI research is facing a vital question of the effectiveness of AI-generated personas. Addressing this question, this research explores user perceptions of AI-generated personas for textual content (GPT-4) and two image generation models (DALL-E and Midjourney). We evaluate whether the inclusion of images in AI-generated personas impacts user perception or if AI text descriptions alone suffice to create good personas. Recruiting 216 participants, we compare three GPT-generated personas without images and those with either DALL-E or Midjourney-created images. Contrary to initial expectations from persona literature, the presence of images in AI-generated personas did not significantly impact user perceptions. Rather, the participants generally perceived AI-generated personas to be of good quality regardless of the inclusion of images. These findings suggest that textual content, i.e., the persona narrative, is the primary driver of user perceptions in AI-generated personas. Our findings contribute to the ongoing AI-HCI discourse and provide recommendations for designing AI-generated personas.
... A study by Salminen and colleagues showed that "[…] a smile enhances the perceived similarity with the persona, similar personas are more liked, and that likability increases the willingness to use a persona (Lau, 1982;Salminen et al., 2020). In follow-up research Salminen added to this that personas with happy pictures are perceived as more extroverted, agreeable, open, conscientious, and emotionally stable (Salminen et al., 2022). ...
Article
Full-text available
Objectives Emotional eating is recognized as a potential contributor to weight gain. Emotional eaters often hide their problems because of feelings of shame about their behavior, making it challenging to provide them with the necessary support. The introduction of a virtual coach might offer a potential solution in assisting them. To find out whether emotional eaters are receptive to online personalized coaching, we presented emotional eaters with two essential proto-typical problem situations for emotional eaters: “experiencing cravings” and “after giving in to cravings,” and asked them whether they preferred one of the three coaching strategies presented: Validating, Focus-on-Change and Dialectical. Methods An experimental vignette study (2 × 3 design) was carried out. The vignettes featured two distinct personas, each representing one of the two common problem scenarios experienced by emotional eaters, along with three distinct coaching strategies for each scenario. To identify potential predictors for recognition of problem situations, questionnaires on emotional eating (DEBQ), personality traits (Big-5), well-being (PANAS), and BMI were administrated. Results A total of 62% of the respondents identified themselves with “after giving in to cravings” and 47% with “experiencing cravings.” BMI, emotional eating and emotional stability appeared to be predictors in recognizing both the problem situations. In “experiencing cravings,” the participating women preferred Dialectical and the Validation coaching strategies. In the “after giving in to cravings” condition, they revealed a preference for the Dialectical and the Focus-on-Change coaching strategies. Conclusion Using vignettes allowed a less threatening way of bringing up sensitive topics for emotional eaters. The personas representing the problem situations were reasonably well recognized. To further enhance this recognition, it is important for the design and content of the personas to be even more closely related to the typical problem scenarios of emotional eaters, rather than focusing on physical characteristics or social backgrounds. This way, users may be less distracted by these factors. With the knowledge gained about the predictors that may influence recognition of the problem situations, design for coaching can be more customized. The participants represented individuals with high emotional eating levels, enhancing external validity.
... Audit studies have been commonly used to study biases in how economic agents make decisions (Mullainathan et al., 2012;Kline et al., 2021;Salminen et al., 2022). Our method applies GANs to address the confounding problem by generating images that differ in only the selected dimension; ...
... The pipeline that we propose is particularly suitable for audit studies. Audit studies have been commonly used to study biases in how economic agents make decisions (Mullainathan et al., 2012;Kline et al., 2021;Salminen et al., 2022). Our method applies GANs to address the confounding problem by generating images that differ in only the selected dimension; additionally, GAN-generated images are realistic, which allows us to study the effect of the feature in a life-like setting. ...
Preprint
Online platforms often face challenges being both fair (i.e., non-discriminatory) and efficient (i.e., maximizing revenue). Using computer vision algorithms and observational data from a micro-lending marketplace, we find that choices made by borrowers creating online profiles impact both of these objectives. We further support this conclusion with a web-based randomized survey experiment. In the experiment, we create profile images using Generative Adversarial Networks that differ in a specific feature and estimate it's impact on lender demand. We then counterfactually evaluate alternative platform policies and identify particular approaches to influencing the changeable profile photo features that can ameliorate the fairness-efficiency tension.
Conference Paper
Full-text available
Though photographs of real people are typically used to portray personas, there is little research into the potential advantages or disadvantages of using such images, relative to other image styles. We conducted an experiment with 149 participants, testing the effects of six different image styles on user perceptions and personality traits that are attributed to personas by the participants. Results show that perceptions of clarity, completeness, consistency, credibility, and empathy for a persona increase with picture realism. Personas with more realistic pictures are also perceived as more agreeable, open, and emotionally stable, with higher confidence in these assessments. We also find evidence of the uncanny valley effect, with realistic cartoon personas experiencing a decrease in the user perception scores.
Article
Full-text available
Data-driven personas are a significant advancement in the fields of human-centered informatics and human-computer interaction. Data-driven personas enhance user understanding by combining the empathy inherent with personas with the rationality inherent in analytics using computational methods. Via the employment of these computational methods, the data-driven persona method permits the use of large-scale user data, which is a novel advancement in persona creation. A common approach for increasing stakeholder engagement about audiences, customers, or users, persona creation remained relatively unchanged for several decades. However, the availability of digital user data, data science algorithms, and easy access to analytics platforms provide avenues and opportunities to enhance personas from often sketchy representations of user segments to precise, actionable, interactive decision-making tools—data-driven personas! Using the data-driven approach, the persona profile can serve as an interface to a fully functional analytics system that can present user representation at various levels of information granularity for more task-aligned user insights. We trace the techniques that have enabled the development of data-driven personas and then conceptually frame how one can leverage data-driven personas as tools for both empathizing with and understanding of users. Presenting a conceptual framework consisting of (a) persona benefits, (b) analytics benefits, and (c) decision-making outcomes, we illustrate applying this framework via practical use cases in areas of system design, digital marketing, and content creation to demonstrate the application of data-driven personas in practical applied situations. We then present an overview of a fully functional data-driven persona system as an example of multi-level information aggregation needed for decision making about users. We demonstrate that data-driven personas systems can provide critical, empathetic, and user understanding functionalities for anyone needing such insights.
Article
Full-text available
Background Using risk stratification approaches to determine eligibility has the potential to improve efficiency of screening. Objectives To compare the public acceptability and potential impact on uptake of using different approaches to determine eligibility for screening. Design An online population‐based survey of 668 adults in the UK aged 45‐79 including a series of scenarios in the context of a potential kidney cancer screening programme in which eligibility was determined by age, sex, age and sex combined, a simple risk score (age, sex, body mass index, smoking status), a complex risk score additionally incorporating family history and lifestyle, or a genetic risk score. Outcome measures We used multi‐level ordinal logistic regression to compare acceptability and potential uptake within individuals and multivariable ordinal logistic regression differences between individuals. Results Using sex, age and sex, or the simple risk score were less acceptable than age (P < .0001). All approaches were less acceptable to women than men. Over 70% were comfortable waiting until they were older if the complex risk score or genetics indicated a low risk. If told they were high risk, 85% would be more likely to take up screening. Being told they were low risk had no overall influence on uptake. Conclusions Varying the starting age of screening based on estimated risk from models incorporating phenotypic or genetic risk factors would be acceptable to most individuals and may increase uptake. Patient or Public Contribution Two members of the public contributed to the development of the survey and have commented on this paper.
Article
Full-text available
We conduct two studies to evaluate the suitability of artificially generated facial pictures for use in a customer-facing system using data-driven personas. STUDY 1 investigates the quality of a sample of 1,000 artificially generated facial pictures. Obtaining 6,812 crowd judgments, we find that 90% of the images are rated medium quality or better. STUDY 2 examines the application of artificially generated facial pictures in data-driven personas using an experimental setting where the high-quality pictures are implemented in persona profiles. Based on 496 participants using 4 persona treatments (2 × 2 research design), findings of Bayesian analysis show that using the artificial pictures in persona profiles did not decrease the scores for Authenticity, Clarity, Empathy, and Willingness to Use of the data-driven personas. ARTICLE HISTORY
Preprint
Full-text available
The COVID-19 pandemic and its social and economic consequences have had adverse impacts on physical and mental health worldwide and exposed all segments of the population to protracted uncertainty and daily disruptions. The CoRonavIruS health and Impact Survey (CRISIS) was developed for use as an easy to implement and robust questionnaire covering key domains relevant to mental distress and resilience during the pandemic. In the current work, we demonstrate the feasibility, psychometric structure and construct validity of this survey. We then show that pre-existing mood states, perceived COVID risk, and lifestyle changes are strongly associated with negative mood states during the pandemic in population samples of adults and in parents reporting on their children in the US and UK. Ongoing studies using CRISIS include international studies of COVID-related ill health conducted during different phases of the pandemic and follow-up studies of cohorts characterized before the COVID pandemic.
Chapter
Full-text available
Following the proliferation of personified big data and data science algorithms, data-driven user personas (DDPs) are becoming more common in persona design. However, the DDP templates are seemingly diverse and fragmented, prompting a need for a synthesis of the information included in these personas. Analyzing 31 templates for DDPs, we find that DDPs vary greatly by their information richness, as the most informative layout has more than 300% more information categories than the least informative layout. We also find that graphical complexity and information richness do not necessarily correlate. Furthermore, the chosen persona development method may carry over to the information presentation, with quantitative data typically presented as scores, metrics, or tables and qualitative data as text-rich narratives. We did not find one “general template” for DDPs and defining this is difficult due to the variety of the outputs of different methods as well as different information needs of the persona users.
Chapter
Full-text available
To predict personality traits of data-driven personas, we apply an automatic persona generation methodology to generate 15 personas from the social media data of an online news organization. After generating the personas, we aggregate each personas’ YouTube comments and predict the “Big Five” personality traits of each persona from the comments pertaining to that persona. For this, we develop a deep learning classifier using three publicly available datasets. Results indicate an average performance increase of 4.84% in F1 scores relative to the baseline. We then analyze how the personas differ by their detected personality traits and discuss how personality traits could be implemented in data-driven persona profiles, as either scores or narratives.
Conference Paper
Full-text available
Personas are a well-known technique in human computer interaction. However, there is a lack of rigorous empirical research evaluating personas relative to other methods. In this 34-participant experiment, we compare a persona system and an analytics system, both using identical user data, for efficiency and effectiveness for a user identification task. Results show that personas afford faster task completion than the analytics system, as well as outperforming analytics with significantly higher user identification accuracy. Qualitative analysis of think-aloud transcripts shows that personas have other benefits regarding learnability and consistency. However, the analytics system affords insights and capabilities that personas cannot due to inherent design differences. Findings support the use of personas to learn about users, empirically confirming some of the stated benefits in the literature, while also highlighting the limitations of personas that may necessitate the use of accompanying methods.
Article
Players in online games frequently choose the opposite gender when they select an avatar. Previously, this has been attributed to a player's unconscious sexual anxieties and the need to experiment through the anonymous location of the avatar. However, this paper argues that the development of choice in games, where players have frequently selected the female form for ludic reasons, means that this choice has become normalised through a historical process. The avatar is frequently considered as a tool, with gender regarded as a freely admitted aesthetic pleasure. The player does not see this as a site of tension, or seeks to absolve this tension publicly as an act of appropriation typical to Jenkins¹ textual poachers. Overall, the act of gender switching is not considered deviant within gaming; more, it is embraced as a common practise with historical precedents to support it.
Article
This paper addresses the importance of improving users’ satisfaction and efficiency for obtaining health information through health information systems. In order to understand how task characteristics, such as type, complexity, and difficulty, affect user interaction with online health information systems in China, this paper investigates if and how various types of task and information systems lead to different user experience, and, how tasks affect users’ evaluation of the interaction design of specific systems. A user experiment was carried out with 24 participants completing two types of tasks with two levels of complexity on three popular health information systems in China. The two types of tasks comprise a high objective complexity task and a medium complexity task. The three selected health systems include the 39 Health Website, Quick Doctors, and Good Doctor Online. Both quantitative and qualitative data were analyzed. The results indicate that the difference of both tasks and systems affect users’ judgment concerning task difficulty and complexity, as well as users’ evaluation of the interaction design of the systems. In order to better assist users with obtaining health information, this paper provides suggestions for improving the interaction design of health information systems from four aspects: website structure and information organization, webpage display, usability, and system performance. The limitations and future work are also discussed.