The International Journal of Human Resource

Management

ISSN: (Print) (Online) Journal homepage: [Link]/journals/rijh20

Humanized AI in hiring: an empirical study of a virtual

AI job interviewer’s social skills on applicants’ reactions
and experience

Céline Clavel, Sophie d’Armagnac, Suzanne Hebrard, Théophile Hesters &

Delphine Potdevin

To cite this article: Céline Clavel, Sophie d’Armagnac, Suzanne Hebrard, Théophile Hesters
& Delphine Potdevin (2025) Humanized AI in hiring: an empirical study of a virtual AI job
interviewer’s social skills on applicants’ reactions and experience, The International Journal of
Human Resource Management, 36:2, 206-234, DOI: 10.1080/09585192.2024.2440784

To link to this article: [Link]

View supplementary material

Published online: 18 Dec 2024.

Submit your article to this journal

Article views: 397

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

[Link]
The International Journal of Human Resource Management
2025, VOL. 36, NO. 2, 206–234
[Link]

Humanized AI in hiring: an empirical study of a

virtual AI job interviewer’s social skills on
applicants’ reactions and experience
Céline Clavela, Sophie d’Armagnacb , Suzanne Hebrarda,
Théophile Hestersc and Delphine Potdevina,c
a
Université Paris-Saclay, CNRS, Laboratoire Interdisciplinaire des Sciences du Numérique, Orsay,
France; bTBS Business School, 1 Place Alphonse Jourdain - CS 66810 31068 TOULOUSE Cedex 7,
France; cDAVI-Les Humaniseurs, Puteaux, France

ABSTRACT KEYWORDS
In this study, we investigate the effects of the use of AI job Interviewer;
advanced technologies based on AI in HRM, and particularly anthropomorphic
the use of virtual agent embedding anthropomorphic fea- features; social skills;
tures in recruitment interview (hereafter referred to as AI job artificial intelligence;
digital recruitment
interviewer). We question the extent to which the AI job interview; applicant
interviewer endowed with social skills affects the nature of reactions
the interaction, and specifically, applicants’ perceptions,
experience and reactions during the interview. We focus on
intimacy as a core social skill. Data were gathered from 94
applicants who completed an interview training session with
an AI job interviewer that exhibited either multimodal
expression of intimacy behaviors or no intimacy behaviors.
Following the interaction, participants completed a ques-
tionnaire asking about their perceptions of intimacy, user
experience, impression management and fairness. The find-
ings demonstrate the reciprocity of intimacy behaviors and
a positive user experience in the intimacy condition. By illus-
trating how AI job interviewers with social skills can affect
the social relation established during a recruitment interview
and provide satisfactory experiences, the present study
offers promising avenues for improving the quality and
effectiveness of human resource tools using AI devices, while
also highlighting some managerial and ethical concerns.

Introduction
The advent of recent technological developments based on Artificial
Intelligence (AI) has led to significant changes in the HRM processes
(Meijerink et al., 2021; Vrontis et al., 2022). In the context of recruitment,
AI-based tools provide managers with automated procedures that assist them

CONTACT Sophie d’Armagnac [Link]@[Link] TBS Business School, 1 Place Alphonse Jourdain.
Supplemental data for this article can be accessed online at [Link]
© 2024 Informa UK Limited, trading as Taylor & Francis Group
 The International Journal of Human Resource Management 207

in the assessment and selection of candidates (Woods et al., 2020; Langer

et al., 2020; Pan et al., 2022; Myllymäki, 2021). With the recent implementa-
tion of asynchronous virtual interviews (AVI) conducted by a virtual agent,
the questions raised by AI-based tools tend to extend beyond the domain of
decision-making to encompass the social and emotional aspects of an inter-
action with a virtual agent (Kim et al., 2022). The HRM literature still under-
estimates the extent of change introduced by the anthropomorphic features
embedded in the design of the interaction and in particular, the agent design.
This paper investigates whether these AVI conducted by human-like virtual
agents, represent a new turn in digital recruitment.
Hereafter named ‘AI job interviewers’, they are embodied conversational
agents that conduct a recruitment interview with an applicant, via a web
platform. The agent is programmed to interact with humans by using nat-
ural language and conversational skills and by mimicking emotions and
nonverbal behaviors such as gestures, facial expressions, and idle move-
ments (Pelachaud, 2009). If the HRM literature suggests the importance of
the social dimension of digital recruitment through the consideration for
applicants’ reactions to the design of the digital mediation (Potosky, 2008;
Lukacik et al., 2022), it also suggests the multi-dimensionality of its effects
(Langer et al., 2019), without investigating them.
In the human-computer interaction (hereafter HCI) literature, several stud-
ies show that these agents that mimic social behaviours increase users’ percep-
tion of social presence (Oh et al., 2018), the perception that a social interaction
is occurring, and the quality of the interaction (Loveys et al., 2020; Potdevin
et al., 2021b). In marketing, AI-based assistants endowed with human likeness,
attachment, and perceived interactivity improve the relation with customers
and trigger positive perceptions about the whole purchase process (Kim et al.,
2022). Research in the HRM field is not so conclusive. First, the research
tends to analyze AI job interviewers as a modality of interface among others,
and not a specific, different social relation. Studies compare the human-like
virtual recruiter with platform interaction or chatbot to measure outcomes like
applicants’ behavior, perceptions, and performance (Lukacik et al., 2022), but
do not address applicants’ relationships building with the human-like virtual
recruiter. Second, the desirability of such interface in HRM is subject to
debates: an AI job interviewer exhibiting humanlike verbal or nonverbal fea-
tures can elicit a sense of strangeness among interviewees (Langer et al., 2017,
2019), trigger anxiety (Melchers et al., 2020; Schneider et al., 2019) and other
negative reactions due to the context (i.e. the stakes associated to job inter-
views as illustrated in Langer et al., 2019). While the understanding of the
impact of anthropomorphic features in the AVI design can rely on Lukacik
et al. (2022) framework, we suggest introducing specific theoretical insights to
illuminate the different nature of the relationship.
208 C. CLAVEL ET AL.

By relying on the HCI and HRM literatures, we propose to study the

relation between the AI job interviewer’s anthropomorphic features, and
applicants’ reactions and experience. We use the concept of intimacy to
illustrate an AI job interviewer’s anthropomorphic features. Intimacy is
an interpersonal process of emotional communication associated with
positive outcomes in interpersonal relationships (Lomanowska & Guitton,
2016; Prager, 1997). As a complex process involving multimodal commu-
nication (verbal and nonverbal behaviors), intimacy relies on both expres-
sive and perceptive mechanisms (Prager, 1997) that are incorporated into
the job interview application embedding conversational AI components
and real-time three-dimensional animation. According to HCI literature,
an AI agent in the intimacy conditions has a strong social presence and
develops a social relation (Potdevin et al., 2021b).
This paper therefore addresses the following research question: does
the AI job interviewers’ intimacy affect applicants’ perceptions, reactions
and experience?
We developed an empirical design with two versions of an AI job
interviewer that conducts autonomous and responsive digital job inter-
view sessions: one with anthropomorphic social skills and one without.
The former AI agent skills were elaborated based on the intimacy model.
The originality and strength of our study is the design of the AI job
interviewer for the goal of the research.
Our findings validate the construction of a social relation in the inter-
view session that depend on the AI agent intimacy: the intimacy expressed
by the AI job interviewer leads to a more positive experience compared
with an agent with no intimacy. The findings also indicate that this rela-
tion is not a simulacrum that the user would confuse with a real con-
versation but another modality of social relation. The managerial
contribution derived from our results relate to the interview outcomes
and ethical concerns. If the AI job interviewer endowed with intimacy
affects the applicants’ emotions and improves the users’ experience, this
finally questions its design, and the ethics related to the practice of
resorting to emotional communication by a machine. By focusing on the
AI job interviewer as a different social relation rather than just a sophis-
ticated interviewing medium, our findings reveal cross-cutting insights
for HRM, organizational psychology, and human-computer interactions.

Conceptual background and hypothesis

The AI job interviewer as a medium
In technology-mediated recruitment interviews, the medium matters:
the efficiency of the procedure as well as applicants and recruiters’ per-
ceptions are affected by the characteristics of the medium (Potosky,
 The International Journal of Human Resource Management 209

2008). According to Chapman et al. (2003) applicants have less favor-

able reactions toward telephone and videoconferencing interviews than
in face-to-face, in terms of perception of fairness and intention to
accept the job. Furthermore, an a-synchronous pre-recorded interview
is likely to suppress interactivity and thus users’ feelings of control over
the experience (Langer et al., 2019; Levashina et al., 2014; Woods et al.,
2020; Melchers et al., 2020). Applicants who feel they lack such control
may worry about how they present themselves. This feeling is exacer-
bated in platform interviews, since they do not receive feedback from
the recruiter. These interviews may prevent applicants from aptly
demonstrating their abilities, leading to frustration that can further
hinder their ability to effectively engage in impression management
(Langer et al., 2019). belangrijk voor hypothese self representation
Compared to a web-based platform interview that inhibits natural con-
versation and follow-up questions (Langer et al., 2017), an AI job inter-
viewer can integrate multimodal aspects of human communication to
enable a more dynamical interaction, including follow-up questions (Su
et al., 2019). In the HCI field, in an experiment comparing a scenario-based
interview and an android robot acting as an interviewer and generating
follow-up questions based on the interviewees’ answers, participants
report a greater appreciation of the android robot: the results indicated
that the quality of the content and the feeling of presence were higher in
the android conversational robot compared with the scenario-based ver-
sion (Inoue et al., 2020). To sum up, a medium that allows the digital
interview to be as similar as possible to face-to-face interview is consid-
ered as the optimal solution for the digital interview outcomes. Yet, in
the HRM literature, several studies suggest that AI job interviewers might
generate applicants’ anxiety (Langer & König, 2018; Melchers et al., 2020).
A recent review by Lukacik et al. (2022) dealing with the relation between
AVI design and applicants’ reactions shows that it is finally difficult to
decipher if applicants’ negative reactions result from the fact that the
recruitment is digitalized or reflect some specific design features. Among
them, the particularity of anthropomorphic features is not mentioned as
a topic addressed by the virtual interview literature. We suggest that the
imitation of human interaction could induce a different status of the
experience compared with a non-human-like conversational digital inter-
viewer, that might be whether positive of negative. This is therefore
important to study the nature and effects of such design. This change in
the ontological status of the relation when an AI job interviewer is imple-
mented is supported by the recent studies in HCI, and specifically the
affective computing field (Lee et al., 2020; Potdevin et al., 2021b), based
on the early CASA paradigm.
210 C. CLAVEL ET AL.

The AI job interviewer as an agent in a social relation

Since the emergence of the affective computing field in early 2000 (Picard,
1999), research on HCI has sought to understand the importance of social
behaviors and perceptions. With the Computers Are Social Actors (CASA)
paradigm, Moon and Nass (1996) argue that embodied conversational
agents should be considered social partners in interactions. Empowered
with verbal communication skills (i.e. ability to understand a question in
natural language and to provide appropriate answers) and nonverbal behav-
iors (i.e. gesturing, gazing, exhibiting facial expressions), these agents can
engage in quasi-natural communication (Pelachaud, 2009).
In a recent systematic literature review, Loveys et al. (2020) show that
virtual agents design affects the users’ relationship quality, social percep-
tions, and behaviors. In HCI research on virtual interviewing, Gebhard
et al. (2014) show that an AI job interviewer exhibiting demanding (ver-
sus understanding) behaviors can induce stress in interviewees. In their
study, the demanding virtual interviewer displayed dominant gestures
and gaze behaviors, negative facial expressions, and a strict language
style. In contrast, the understanding virtual interviewer exhibited friendly
gestures, positive expressions, and politeness. Interviewees felt less com-
fortable with the demanding virtual interviewer and self-reported lower
performance scores for the interview. Another study demonstrated that
manipulation of social communication (friendly versus hostile) through
verbal and nonverbal behaviors of a virtual recruiter could influence
interviewees’ perceptions (Callejas et al., 2014). In sum, endowing virtual
agents with positive social behaviors seems to create positive perceptions
among users.
To convey social presence, the virtual agent must be perceived as a
partner in conversation by the applicant (Lombard & Ditton, 2006). Social
skill comprises socially acceptable behaviors that allow individuals to
respond to and interact with others (Grover et al., 2020). These behaviors
reflect a person’s ability to understand, adapt to, and access intentions,
goals, and emotions in ways that have positive impacts in social contexts
(Gardner & Hatch, 1989). Agents exhibiting intentionally designed social
skills, as emotional communication skills, would be able to induce reci-
procity effects. They are perceived to be more social than agents that are
not designed with such social skills (Potdevin et al., 2020) and influential
on user intentions and behaviors (Bickmore et al., 2009; Potdevin et al.,
2021b; Lee & Choi, 2017; Lee et al., 2020). Among the psychological con-
structs trying to encapsulate social skills and its impact on interpersonal
communication, the concept of intimacy has been extensively studied in
the field of interpersonal psychology because i­ntimacy is typically built
for analyzing interactions: it combines characteristics of emotional
 The International Journal of Human Resource Management 211

communication by individuals, with outcomes in interpersonal relation-

ships behaviors (Laurenceau et al., 2005; Lomanowska & Guitton, 2016;
Prager, 1997). It therefore refers to behavioral, physical, cognitive, and
emotional sharing experiences (Morton, 1978; Prager, 2000). As a com-
plex process involving multimodal communication, intimacy relies on
both expressive and perceptive mechanisms (Prager, 1997). Verbal behav-
iors that increase intimacy within an interpersonal relationship include
revealing personal or private emotional information (i.e. self-disclosure)
and sharing feelings or supporting behaviors. Nonverbal behaviors include
physical proximity, smiling and other positive facial expressions, making
eye contact, and head nodding or tilting.
Although it is typically associated with close social contacts, intimacy
occurs in all kinds of relationships and settings. For example, increased
intimacy between the customer and retailer is related to higher satisfac-
tion, stronger purchase intention, and positive customer emotions (Yim
et al., 2008). In the HCI domain, studies have investigated the impact of
intimacy exhibited by embodied conversational agents on human percep-
tions and behaviors (e.g. Lee & Choi, 2017). For example, Potdevin et al.
(2021a) find a positive impact of a virtual tourism counselor’s multi-
modal intimacy behaviors on tourists’ perceptions and social behaviors.
Inspired by literature in human psychology, the authors propose a con-
ceptual model of virtual intimacy for human–agent interactions compris-
ing three key dimensions: honesty and genuineness (e.g. self-disclosure,
sharing personal thoughts and opinions, revealing vulnerability), positiv-
ity (e.g. positive feelings, commitment), and mutual comprehension (e.g.
understanding, consideration, support, care).
In the field of HRM also, a human-like agent with anthropomorphic
features like intimacy might generate a social experience including emo-
tional communication, reciprocity, and finally, the building of a social
relation. This change could be impactful for interview outcomes, through
a ‘triggering effect’ on applicant’s behavior during the interview (Lukacik
et al., 2022, p. 6), and also, regarding the attractiveness of the hiring
organization (Chapman et al., 2005; Langer et al., 2019; Woods et al.,
2020): applicants consider an organization as attractive to the extant that
they perceive themselves as corresponding to this organization
(person-organization fit), which might be reinforced by the key dimen-
sions of intimacy outlined above (genuineness, commitment, perception
of understanding and consideration). The reception of a social signal
(the intimacy of the agent), and the willingness to respond to such soci-
ality mean that social presence is enhanced, which can positively affect
interview performance (Lukacik et al., 2022). However, the perception by
applicants of an AI job interviewer intimacy and their will to behave
212 C. CLAVEL ET AL.

accordingly might be hindered by typical digital interview features refer-

ring to acceptance related to such experience (e.g. negative affective reac-
tion to an interview conducted by artificial intelligence and creepiness
associated to the AI job interviewer, see Langer & König, 2018).
In sum, expression of virtual intimacy from an AI job interviewer may
create specific conditions for a social relationship, increase applicants’
self-disclosure and commitment, sense of togetherness and willingness to
join the hiring organization. Accordingly, we formulate the following
hypothesis on intimacy perception:
Hypothesis 1: Compared with participants in the no-intimacy condition, partici-
pants in the intimacy condition will perceive the AI job interviewer to be more
intimate.

As previously noted, intimacy is an interpersonal and dynamic process

of communication (Potdevin et al., 2020; Prager, 1997). Intimacy expressed
by an AI job interviewer should not be just perceived but should also
influence the intimacy behaviors of its interlocutor. Therefore, we propose:
Hypothesis 2a: Participants in the intimacy condition will perceive themselves as
behaving more intimately, compared with those in the no-intimacy condition.

We also hypothesize a positive link between the participants’ percep-

tions of intimacy from the AI job interviewer and their perceptions of
their own intimacy behavior. Thus, we propose:
Hypothesis 2b: The more intimacy participants perceive in the AI job interviewer,
the more they will perceive themselves as behaving intimately.

AI job interviewer and user experience

If the perception and reciprocity in intimacy means that a social relation
is built, it does not mean this relation is experienced as positive. As
outlined above, negative emotions might finally arise during this social
relation: feelings to be understood and considered might coexist with
forms of anxiety or reluctance with self-disclosure in front of a robot
(Langer & König, 2018). Despite the growing interest in using virtual
agents for business (see for example, Gartner, 2021), emotions, feelings
or subjectivity in interactions with virtual agents are still neglected in
management. In a systematic review of the literature in marketing about
the use of conversational agents, Ling et al. (2021) identify factors related
to usage (e.g. usefulness, ease of use), the agent (e.g. physical appearance,
movements, likeability, social behaviors), the user (e.g. demographics,
psychological, intrinsic motivation), and the user’s attitude and evaluation
(e.g. emotions and satisfaction) that directly or indirectly affect inten-
tions to use and adopt conversational agents. However, the agent
 The International Journal of Human Resource Management 213

appearance and social skills are rarely considered: most studies focus on
one or a few specific dimensions of the user–agent interaction, such as
satisfaction (Lee & Choi, 2017), engagement (Bickmore et al., 2009), or
usefulness (Yang & Lee, 2019), without necessarily considering emotions
as a prism through which AI agent adoption could be studied (for an
exception, see Gursoy et al., 2019). In recent studies however, warmth of
the message of the chatbot (Kull et al., 2021) or authenticity of the avatar
(Jones et al., 2022) are perceived positively by consumers, affecting pos-
itively their engagement toward the brand, loyalty and satisfaction.
In the nascent HRM literature dealing with the acceptance of robots
in the HRM field, the user experience construct is an important dimen-
sion of employee reactions to exchanges based on AI applications,
whether virtual assistants or AI enabled bots (Malik et al., 2022; Dutta
et al., 2023). Despite this statement, the literature on digital recruitment
interview does not directly address the importance of user experience. It
is stated that for automated virtual interviews, users’ anxiety can nega-
tively affect their performance (Melchers et al., 2020; Schneider et al.,
2019). The medium used for selection interviews is considered preferable
when it allows the conversation to be natural and convey similar feeling
and emotions that a human conversation would do (Chapman et al.,
2003; McColl & Michelotti, 2019; Lukacik et al., 2022). Yet, emotions are
not systematically considered as desirable. Malik et al. (2022) observed
that employees interacting with AI enabled bots experienced
hyper-personalization and individualization of HRM practices that
resulted in a positive experience and finally an increased satisfaction and
commitment to the organization. Interestingly, this experience is described
as deprived with emotions and subjectivity. This is on purpose, because
HRM would need control over emotions and subjectivity, compared with
the interactions occurring between human beings. More investigation is
therefore needed about the intimacy of the agent as a predictor of a
positive user experience via the positive emotions conveyed by this agent.
Such investigation requires refining the understanding of user experi-
ence in relation to emotions. User experience formally describes all
aspects of a person’s experience with a system (Norman et al., 1995),
including perceptions and responses resulting from the use and antici-
pated use of a product or service (ISO (International Organization for
Standardization), 2010). Contrary to the concept of acceptance, which
poorly incorporates emotions within its framework, user experience cap-
tures ‘all the users’ emotions, beliefs, preferences, perceptions, physical
and psychological responses, behaviors, and accomplishments that occur
before, during and after use’ (ISO (International Organization for
Standardization), 2010). The Components of User Experience model, as
defined by Mahlke and Lindgaard (2007) and Mahlke (2008), describes
214 C. CLAVEL ET AL.

three dimensions of user experience: user emotions, instrumental aspects

(product usability and usefulness), and non-instrumental aspects (aes-
thetic and symbolic aspects). With emotions positioned as a core dimen-
sion, the model offers an interesting evaluative framework to study how
system properties, user characteristics, and interactions between them
influence the overall experience. Potdevin et al. (2021b) show that user–
agent interactions have a direct influence on emotions as well as indirect
effects via user perceptions of instrumental and non-instrumental quali-
ties of the system. As previously mentioned, adjusting the behavior of
the AI job interviewer (e.g. enhancing perceptions of intimacy) could
improve this user experience.
Virtual intimacy might be a predictor of user experience, particularly
by influencing users’ emotions (Potdevin et al., 2021b). We therefore
make several related predictions regarding a positive association between
the perception of virtual intimacy and the users’ overall experience:
Hypothesis 3a: Participants in the intimacy condition will perceive a more posi-
tive user experience, compared with those in the no-intimacy condition.

Hypothesis 3b: The more intimacy participants perceive in the AI job interviewer,
the more positively they will perceive their user experience.

Hypothesis 4a: The positive user experience predicted in Hypothesis 3 will be

driven by a positive impact on user emotions.

Hypothesis 4b: The more intimacy participants perceive in the AI job interviewer,
the more positive their emotions will be.

The AI job interviewer, an agent of confusion between real and virtual

interview?
Investigating the building of a social relationship with an AI job inter-
viewer also questions the positioning of such relationship: is the AI job
interviewer considered by applicants a real person? Is the digital inter-
view lived as an experience of real conversation? Is there any confusion
between the real and the virtual? If this confusion exits, applicants’ reac-
tions as impression management and fairness perception might be
affected, with differences depending on the sociality of the agent.
Concerning impression management, applicants’ reactions depend on
their feeling of making the right impression (Stevens & Kristof, 1995).
Levashina et al. (2014, p. 256) define impression management as ‘a pro-
cess by which people attempt to influence the images others form of
them during social interaction’. Applicants engage most in impression
management when speaking in person-to-person settings than in video-
conferences (Basch et al., 2021). Accordingly, talking with an AI job
 The International Journal of Human Resource Management 215

interviewer could exacerbate the feeling of poor relevance of impression

management techniques. However, in such conditions, the intimacy of
the AI job interviewer might mitigate this, through friendly or benevo-
lent reactions, and make the applicant get confidence in having the
potential of influencing the course of the interview. This is suggested by
studies in the marketing field: the human-likeness of the virtual assistant
improves the perceived interactivity of this assistant and finally induce
consumers to make a decision, suggesting that the human-like virtual
assistant enhances the feeling of control of consumers over the conversa-
tion (Kim et al., 2022). An AI job interviewer equipped with virtual inti-
macy skills might trigger impression management techniques by
mimicking natural human interactions.
The perception of fairness toward the interview led by an AI job
interviewer could also be affected by this agent features. Users’ fairness
perceptions vary according to the type of interview: they are stronger in
face-to-face interviews than in telephone or videoconferencing interviews
(Chapman et al., 2003; Basch et al., 2021). However, Langer et al. (2017)
find no significant difference in fairness perceptions between videocon-
ferencing and automated asynchronous virtual interviews, perhaps because
the preparation time given in the automated procedure positively affected
perceptions of fairness. Following Lukacik et al. (2022), the opportunity
to perform an AVI might be secured by the fact that the procedure is
automated and similar for all applicants. Taken more globally, the per-
ception of fairness of the whole hiring process depends on the relation-
ship between the importance of the hiring process for the applicant and
the perception of the use of an AI-based device: the perception of fair-
ness could be attenuated by the use of an AI-based job interviewer
instead of a real person, for a high-stake situation (Langer et al., 2019).
However, the perception of fairness could be improved if the sophistica-
tion and technologies embedded in the agent are convincing enough to
show the willingness of the employer to consider candidates. In that
sense, the AI job interviewer’s intimacy might suggest such consideration
and influence fairness perceptions. Therefore, we explore in the present
study how intimacy exhibited (or not exhibited) by the AI job inter-
viewer affects impression management, and their perception of fairness
regarding this type of interview.
Based on results from Langer et al. (2019, 2020) and Basch et al. (2021),
impression management and fairness perception might depend on the inti-
macy condition. Considering the earlier arguments regarding the effects of
interview conditions on fairness, we hypothesize the following:
Hypothesis 5: Participants in the intimacy condition will perceive a higher sense
of impression management in the recruitment process than those in the no-intimacy
condition.
216 C. CLAVEL ET AL.

Hypothesis 6: Participants in the intimacy condition will perceive a higher sense

of fairness of the recruitment process than those in the no-intimacy condition.

Hypotheses
In sum, we argue that AI job interviewers’ behavior change the condi-
tions of social presence, and thus the nature of the interaction in the
digital interviews, which might affect applicants’ perceptions and reac-
tions. Specifically, we formulate and test two conditions in which appli-
cants complete an interview training session with an AI job interviewer
that uses intimacy behaviors (intimacy condition) or does not exhibit any
such behaviors (no-intimacy condition). Below is the conceptual model
that visually represent the hypotheses (See Figure 1).

Experimental study on a job interview app

We developed an autonomous job interview application leveraging
cutting-edge conversational AI components, HR expertise, and real-time
three-dimensional animation rendering to simulate a natural and dynamic
interview process (refer to Figure 2). For a detailed description of the job
interview interface, please see the Supplementary File.

Figure 1. Conceptual model.

 The International Journal of Human Resource Management 217

Figure 2. AI job interviewer interface in the intimacy condition. Applicant’s image in the
upper left corner is blurred for anonymity.

Participants and experimental design

The study included 94 BSc students (54 women, 40 men) in their first
year of business and administration management, all French speakers
and aged at least 18 years (M = 18.5, SD = 0.7, min = 18, max = 21).
These participants were enrolled in short, vocational bachelor’s degree
programs aimed at preparing students for rapid integration into the job
market. To examine the impact of the AI job interviewer’s communica-
tion behaviors on participants (specifically, the expression of intimacy),
we manipulated one independent variable: intimacy behaviors in a
between-subjects condition (intimacy vs. no intimacy). This setup aimed
to explore the effects of intimate vs. non-intimate interaction styles,
detailed further in Potdevin et al. (2021b) and the Supplementary File.

Procedure and conditions overview

All participants were students attending a mandatory ‘professional day’

hosted by their university. At the event, students had the opportunity to
complete an interview training with an AI job interviewer to help them
prepare for a subsequent mandatory face-to-face interview with an HR
professional at the same event. These conditions added real stakes to the
conditions of our experiment. After the introduction to the training ses-
sion, students were assigned to a computer with a headset, a microphone,
and a camera. Participants were randomly assigned to a computer
218 C. CLAVEL ET AL.

preprogrammed with either the intimacy (N = 47) or no-intimacy condi-

tion (N = 47). All participants were asked to read and sign a consent form.
The training session included an introduction, seven interview ques-
tions, and a conclusion (Appendix A). Depending on the applicant’s
responses, the AI job interviewer could pose follow-up questions or con-
tinue with the predefined scenario until the interview’s end. After train-
ing, participants were asked to complete an online questionnaire about
their perceptions of the interview training, taking approximately 25 to
30 min in total, followed by the face-to-face interview with the HR
professional.

Intimacy vs. No-intimacy conditions

Appendix A details the scenarios for the intimacy and no-intimacy con-
ditions of the job interview training session. The no-intimacy version
adopts a formal and straightforward tone, presenting the AI job inter-
viewer as a mere intermediary in the interview process. Conversely, the
intimacy version personalizes the interaction, with the AI job inter-
viewer named Sasha adopting a warmer, more personal tone, seeking to
establish a connection with the candidate.

Measures

The online questionnaire was designed to evaluate four main areas: vir-
tual intimacy conveyed by the AI job interviewer, self-perception of inti-
macy (hereafter referred to as ‘virtual self-intimacy’), the interview
process perception, and overall user experience.

Perception of virtual intimacy

This dimension was assessed using a revised version of the virtual inti-
macy scale originally developed by Potdevin et al. (2018, 2020). The revi-
sion was undertaken to accommodate the assessment of participants’
self-perceptions in addition to their perceptions of the virtual job inter-
viewer’s intimacy. This scale consists of 17 items spread across three sub-
dimensions: honesty and genuineness (e.g. ‘The virtual interviewer acted
in a spontaneous and authentic manner’ vs. ‘I acted in a spontaneous
and authentic manner’), positivity (e.g. ‘The virtual interviewer adopted
a positive and pleasant attitude’ vs. ‘I adopted a positive and pleasant
attitude’), and mutual comprehension (e.g. ‘The virtual interviewer sought
to truly understand me’ vs. ‘I sought to truly understand the virtual
interviewer’). This scale uses a sliding scale from 0% to 100%, and its
internal consistency was found to be excellent (Cronbach’s alpha = 0.94).
 The International Journal of Human Resource Management 219

Perception of user experience

Assessed with the meCue questionnaire (Mahlke, 2008), validated in
French by Lallemand and Koenig (2017), this measure uses a 7-point
Likert scale with 30 items divided into four modules: product percep-
tions, emotions, consequences, and global evaluation. Examples of items
include ‘The device’s features are perfectly aligned with my goals’ for
product perceptions (Cronbach’s alpha = 0.89) and ‘Using this device
makes me feel joyful’ for emotions (Cronbach’s alpha = 0.84). The unique
approach of analyzing mean scores from the product perceptions and
emotions modules separately was inspired by the Components of User
Experience model, yielding a comprehensive Cronbach’s alpha of 0.92 for
the overall user experience measure.

Perception of impression management

This four-item scale developed by Basch et al. (2021) assesses applicants’
ability to present themselves attractively during the interview process. An
example item is ‘During the interview, I was able to describe my skills
and abilities attractively’. This measure uses a 7-point Likert scale, with
a Cronbach’s alpha of 0.74 indicating good reliability.

Perception of fairness and justice

The fairness perception scale by Basch et al. (2021) was adapted for this
study to focus on job-relatedness and chance to perform, using a subset
of 5 items. Example items include ‘Succeeding in this interview means I
can succeed in the job I am applying for’. This scale also uses a 7-point
Likert scale, demonstrating satisfactory internal consistency (Cronbach’s
alpha = 0.79). Scales Reliability and correlation matrix are available in
Appendix B.

Description of analytical methods

In this study, we employed a combination of descriptive, comparative, and
regression analyses to explore the data collected from our participants.
For comparative analysis, we utilized independent samples t-tests to
investigate differences between groups, specifically looking at the effects
of perceived intimacy on applicant perceptions. This choice was moti-
vated by our focus on comparing mean scores across different conditions
and groups, aligning with our hypotheses regarding the impact of AI job
interviewer behaviors.
In exploring the predictive relationships between perceived intimacy
from the AI job interviewer and outcomes such as applicants’
self-perception of intimacy and user experience, we conducted linear
regression analyses. These models were chosen to quantify the strength
220 C. CLAVEL ET AL.

and direction of these relationships, allowing us to test our hypotheses

regarding the influence of perceived intimacy on various aspects of the
applicant experience.

Results
Descriptive analysis
A distribution analysis of our data using Shapiro-Wilk tests supports the
assumption of normality (p > 0.05), except for user experience perception
and its related subdimensions. Nevertheless, the sample size is large
enough (N = 94) to be consistent with the use of parametric tests (Dudley,
2014). On average, the job interview training lasted 9.45 min, and partic-
ipants spoke for 7.15 min. Participants answered 3.07 additional follow-up
questions. We also tested gender bias and found a difference only for the
perception of honesty and genuineness of the AI job interviewer (t[92]
= −2.14, p = < 0.035, d = 0.45, M = 49.25, SD = 20.02 vs. M = 40.28,
SD = 20.16).
Before presenting our analysis outcomes, we reiterate our study’s
hypotheses to align the forthcoming results with our initial conjectures.
This study posits that (1) intimacy behaviors by the AI job interviewer
enhance applicants’ perceptions of intimacy towards the interviewer and
themselves, and (2) such perceived intimacy positively influences the
user experience and emotional response, and perceptions of impression
management and fairness during the AI-conducted interview process.

Intimacy behaviors affect applicants’ intimacy perceptions of the AI job

interviewer and of themselves
Using t-tests, we compared perceptions of the AI job interviewer’s inti-
macy behavior and the three related factors, as well as the self-perception
of intimacy, in both experiment conditions. As shown in Figure 3, we
observed significantly higher scores for the AI job interviewer in the
intimacy condition (t[92] = 8.478, p = < 0.001, d = 1.748, M = 67.19,
SD = 15.23 vs. M = 40.68, SD = 15.10). This effect was repeated for all
three virtual intimacy scores. A second analysis revealed significantly
higher mean scores for self-perception of intimacy in the intimacy con-
dition (t[92] = 2.77, p = 0.003, d = 0.57, M = 55.23, SD = 20.99 vs. M = 44.30,
SD = 17.11). Again, the effect was repeated across the three related scores
of virtual intimacy.
A linear regression analysis also showed that perceived virtual inti-
macy of the AI job interviewer significantly predicted applicants’
self-perception of intimacy (F[1, 92] = 71.85, p < 0.001, α = 14.59, SE =
0.008, β = 0.652, R2 = 0.432).
 The International Journal of Human Resource Management 221

Figure 3. Applicants’ perception of AI job interviewer’s intimacy (left) and their self-perception
of intimacy (right), by intimacy condition.

These outcomes lend support to Hypothesis 1 by providing evidence

against the null hypothesis, suggesting participants perceived the AI job
interviewer with intimacy behaviors as more intimate than without.
Additionally, the results endorse Hypothesis 2, illustrating that applicants
felt more intimate with the AI job interviewer exhibiting intimacy behav-
iors. We thus corroborate a positive linkage between perceiving virtual
intimacy from an AI job interview and self-perception of intimacy
(Hypothesis 2b). Group descriptives for both the perception of virtual
intimacy toward the AI job interviewer and the applicants, as well as a
correlation matrix, are available in Appendix B.

Intimacy behaviors positively influence user emotions and overall user

experience

In this section, we examine the influence of intimacy behaviors on user

emotions and overall user experience through a series of t-tests, with a
particular emphasis on distinguishing between the objectives and out-
comes of each test.
The first t-test aimed to explore the overarching effect of intimacy
behaviors on the overall user experience score. This analysis sought to
determine whether the presence of intimacy behaviors by the AI job
interviewer could significantly enhance the general perception of the
interview process by participants. Our findings indicate a significant
effect of intimacy on the overall user experience score (t(92)= 2.11,
p = 0.019, d = 0.44, M = 4.45, SD = 0.95 vs. M = 4.04, SD = 0.91, as well as
on the scores related to Module 1 - Production Perception, t(92) = 2.24,
p = 0.014, d = 0.46, M = 4.40, SD = 0.97 vs. M = 3.94, SD = 1.05 and Module
2 - Emotions t(92) = 1.82, p = 0.036, d = 0.38, M = 4.43, SD = 1.00 vs.
M = 4.04, SD = 1.04.), suggesting that participants in the intimacy
222 C. CLAVEL ET AL.

condition rated their user experience more positively than those in the
no-intimacy condition.
Subsequently, a second series of t-tests was conducted to specifically
assess the impact of intimacy behaviors on distinct emotional responses
- namely, positive and negative emotions. These tests were designed to
isolate and compare the effects of intimacy behaviors on different facets
of participants’ emotional experiences during the Job Interview. Results
from this analysis revealed that only positive emotions were significantly
impacted by the presence of intimacy behaviors (t(92) = 1.74, p = 0.043,
d = 0.36, M = 3.21, SD = 1.53 vs. M = 2.66, SD = 1.55), while no significant
effect was observed for negative emotions. This distinction underscores
the targeted influence of intimacy behaviors on enhancing positive emo-
tional responses, without a corresponding increase in negative emotions.
By clarifying the separate objectives and findings of these two sets of
t-tests, we aim to provide a more nuanced understanding of how inti-
macy behaviors exert differential impacts on various aspects of the user
experience and emotional responses during job interviews.
Figure 4 summarizes the results.
Using a linear regression analysis, we tested the relationship between
participants’ perceptions of the AI job interviewer’s intimacy and their
overall user experience. We found that the virtual intimacy score signifi-
cantly predicted the mean user experience score (F[1, 92] = 30.52, p <
0.001, α = 2.98, SE = 0.004, β =0.023, R2 = 0.24) and the emotions score
(F[1, 92] = 22.31, p < 0.001, α = 3.01, SE = 0.005, β =0.023, R2 = 0.19). A
correlation matrix is proposed in Appendix B.
These findings confirm that intimacy-related behaviors of the AI job
interviewer affected the user experience (Hypothesis 3) and that posi-
tive experiences were driven by a positive impact on user emotions
(Hypothesis 4). Additional analyses indicated that the effect on user
emotions was not solely responsible for the increased score for user
experience. The product perception module was also sensitive to the
intimacy condition. Overall, the results support our Hypotheses 3b and
4b identifying perceived virtual intimacy as a significant predictor of
the overall user experience and user emotions. Group descriptives for
the user experience and its related subdimensions are available in
Appendix B.

Perception of impression management and fairness

Finally, we studied the impact of the AI job interviewer’s intimacy-related

behaviors on the applicants’ perceptions. Because of technical issues, per-
ceptions of impression management and fairness could not be collected
from all participants; thus, the analysis included a subset of 67
 The International Journal of Human Resource Management 223

Figure 4. Mean user experience and related modules. The figure presents perceptions of
instrumental and non-instrumental system qualities (production perception module) and
positive and negative emotions (emotions module) separately.

participants (20 in the intimacy condition and 47 in the no-intimacy

condition). Descriptive analyses revealed that, on average, participants
attributed a score of 4.00 (SD = 1.31, min = 1.00, max = 7.00) to their per-
ception of impression management and 3.57 (SD = 1.27, min = 1.00,
max = 6.20) to their perception of fairness in the interview. Additionally,
no difference was found between the intimacy and no-intimacy condi-
tions in perceptions of impression management (t[65] = 0.44, p = 0.33,
d = 0.118) or fairness (t[65] = 0.35, p = 0.73, d = 0.092). These findings sug-
gest a non-rejection of the null hypothesis for Hypotheses 5 and 6 due
to the lack of significant effects from the AI job interviewer’s
intimacy-related behaviors on these outcomes. Table 1 below recapitulates
hypotheses and their outcomes.
In sum, according to our results, when the virtual interviewer is
endowed with anthropomorphic features as intimacy, applicants perceive
the AI agent social skills (here, honesty and genuineness, positivity,
understanding and support) and they develop reciprocally a social behav-
ior. This unveils the social dimension of the relationship. These anthro-
pomorphic features are not basically the expression of a sophistication of
the design of the digital interview: they introduce a more complex rela-
tionship, from an interaction more or less efficient and acceptable, to a
social relation including mixed emotions and reciprocal behaviors. No
confusion could be demonstrated, however, through the measure of
impression management and fairness: since the hypothesis are not vali-
dated (see Table 1), it suggests that applicants in the intimacy condition
did not find relevant to implement more impression management tactics
compared to the neural agent, because they kept aware of the virtuality
of the agent. The absence of impact on fairness associated to the agent
in the intimacy condition seem to indicate a similar awareness. A rela-
tion so, but without confusion with a human-human relation. This can
be attributed to the device limits or to the user persistent awareness of
the virtuality of the experience.
224 C. CLAVEL ET AL.

Table 1. Summary of hypotheses and outcomes.

Hypotheses Outcome
Hypothesis 1 Compared with participants in the no-intimacy condition, participants in the Supported
intimacy condition will perceive the AI job interviewer to be more intimate.
Hypothesis 2a Participants in the intimacy condition will perceive themselves as behaving Supported
more intimately, compared with those in the no-intimacy condition.
Hypothesis 2b The more intimacy participants perceive in the AI job interviewer, the more Supported
they will perceive themselves as behaving intimately.
Hypothesis 3a Participants in the intimacy condition will perceive a more positive user Supported
experience, compared with those in the no-intimacy condition
Hypothesis 3b The more intimacy participants perceive in the AI job interviewer, the more Supported
positively they will perceive their user experience.
Hypothesis 4a The positive user experience predicted in Hypothesis 3 will be driven by a Supported
positive impact on user emotions
Hypothesis 4b The more intimacy participants perceive in the AI job interviewer, the more Supported
positive their emotions will be.
Hypothesis 5 Participants in the intimacy condition will perceive a higher sense of Not Supported
impression management in the recruitment process than those in the
no-intimacy condition
Hypothesis 6 Participants in the intimacy condition will perceive a higher sense of fairness of Not Supported
the recruitment process than those in the no-intimacy condition

Discussion
HRM strategies and activities are changed by recent technological devel-
opments based on IA (Vrontis et al., 2022; Pan et al., 2022). If the HRM
literature has been investigating algorithms and AI applications devel-
oped for HRM, the anthropomorphic features embedded in the digi-
talized experience are still a nascent topic. In a socio-material approach
of HRM, when situated activities constituted by human agents and tech-
nologies are implemented, they have both a social and material existence
and agency (Myllymäki, 2021). In that sense, it is surprising that AVIs
with a human-like conversational agent have not received much attention
in HRM: sociality and materiality are tightly associated in situations that
matter for applicants and hiring organizations. The use of the AI job
interviewer affects the meaning of the interview by being at the same
time mimetic to a human conversation, and different. For applicants, the
attribution of social skills like intimacy to the AI job interviewer, as
opposed to just keeping the device the most neutral as possible, ques-
tions their ability and willingness to deal with the mimetic dimension of
the experience. For the hiring organization, the attribution of social skills
like intimacy to the AI job interviewer means a change in the approach
of digital recruitment that induces organizational and ethical concerns.

The anthropomorphic features of the AI job interviewer and applicants’

experience

According to our results, designing intimate behaviors for an AI job

interviewer is well received by users. This is also consistent with studies
 The International Journal of Human Resource Management 225

in marketing outlining the importance of incorporating warmth or soci-

ality in the design of AI assistants (Kim et al., 2022; Jones et al., 2022).
In the context of recruitment, interactions are highly structured, and
applicants’ perceived control and social presence are low, while anxiety is
reported (Langer et al., 2017; Levashina et al., 2014; Melchers et al.,
2020; Woods et al., 2020). Nevertheless, we observe a self-perception of
intimacy in this setting, even with a virtual interlocutor. Applicants do
not reject the human-likeness of the AI job interviewer behavior: a pos-
itive experience is reported. Applicants who perceive the AVI format to
be more personal are more likely to feel comfortable in the process. This
observation suggests that two important principles from the HCI field
are relevant in the HRM context. First, the AI job interviewer is an
interaction partner, supporting the Computers Are Social Actors para-
digm and the related social response theory (Reeves & Nass, 1996).
Second, the intimate interaction with a virtual agent involves an interper-
sonal and dynamic communication process (Potdevin et al., 2020; Prager,
1997). Our results suggest that the social component of the
technology-mediated interview can be enhanced by using intelligent
agents capable of adapting to and building a relationship with the user
through verbal and nonverbal expressions of intimacy that foster recip-
rocal communication, which could finally benefit the affective dimension
in job interviews (Melchers et al., 2015).
Although our results show that intimacy promotes a better user expe-
rience, we challenge our hypotheses concerning impression management
and fairness. Applicants in the intimacy condition did not indicate that
they had more opportunities than those in the no-intimacy condition to
deploy impression management techniques, nor did they consider the sit-
uation to be fairer. Several reasons explain this absence of difference.
First, applicants might not consider that the AI job interviewer is a
human partner in a human-human conversation. Second, in applicants’
mind, the AI job interview in itself might not be strongly associated with
decisions-making in the selection procedure. For these reasons, the AI
job interviewer might not be perceived as a level of decision to influence,
through impression management, nor a step in the process that could
issue an unfair decision. The fact that the AI job interviewer exhibit
intimacy does not modify such perception.
The context of the study may also explain this result. The experiment
occurred during a job fair as a preliminary training step before a real
person-to-person encounter. This is finally a limited stake for our study
participants. Nevertheless, participants gave above-average ratings for
impression management (over 4) and fairness (over 3.5). Acknowledging
the limitations imposed by the small sample size and its potential impact
on the robustness of the results, further experiments are needed in
226 C. CLAVEL ET AL.

real-context selection interview, to see if intimacy positively affects

impression management and fairness.

Managerial insights for the hiring organization

Digital recruitment interfaces with anthropomorphic features embedded

in the design of the interaction raise managerial and ethical questions.
First, as a managerial concern, the consequences of these digital recruit-
ment interfaces on professional HR activities must be addressed. These
professionals might experience situations of substitution, a change in work
organizations, the transformation of their role, a requirement to collabo-
rate with AI-devices, and finally their professional knowledge might be
challenged by the system explicability (Klein & Watson-Manheim, 2021).
Second, an ethical question is raised by these new technologies, concern-
ing both fairness of the recruitment outcome and even, the procedure.
About the outcome, it is observed that technology tends to standardize
recruitment conditions in favor of one interpretation of the desired pro-
file, rendering the selection process no longer accessible to more atypical
profiles (Blacksmith et al., 2016). Self-learning algorithms in HRM adapt
selection criteria to continuously improve the prediction of candidate’s
performance in their future job (Meijerink et al., 2021). The sophistica-
tion of algorithms questions their transparency for the recruiter: biases
could be introduced and reproduced (Vrontis et al., 2022). Similarly, the
AI job interviewer features could reproduce interactional schemas that
could be in favour of some candidates and penalize others, for reasons
that would not be related to the job needs. If the selection process intro-
duces interactional schemas that do not ensure accurate and equitable
assessments and decisions, then the use of this technology for recruitment
raises potential ethical concerns (Weiskopf & Hansen, 2023).
About the procedure, affective anthropomorphism raises questions
about the status conferred on the conversational virtual agent, especially
as users tend to reveal themselves more during an exchange with a vir-
tual agent than with a human; in short, they are less wary of the agent
(Mell et al., 2017). In the context of our study, we relied on and sought
to reinforce this affective dimension by assuming that participants would
perceive an AI job interviewer that had intimacy behaviors as less judg-
mental than a neutral one and thus would be more comfortable revealing
personal or critical information in our setting. The ethical dimension in
digitalized social interaction is thus complex to question. On the one
hand, it appears to be crucial establishing interactions in which appli-
cants feel safe to open up and reveal themselves without artifice. On the
other hand, in the context of a recruitment interview, users remain aware
that digital devices reflect the recruitment purpose, and the
 The International Journal of Human Resource Management 227

overconfidence triggered by the intimate dimension might either remain

relatively limited or constitute a problem. Research dealing with the
anthropomorphic features of digital interaction do not only require the
association of HCI and management literature, it also requires opening a
conversation with management ethics and law literatures.

Adopting a different paradigm for analyzing digital interviews design

Finally, our findings suggest that the study of digital interview, when
anthropomorphic features are embedded, would benefit a change in para-
digm. Searle and Al-Sharif (2018) consider that three different paradigms
for recruitment and selection determine the way in which research addresses
recruitment issues: psychometric or predictivist paradigm (the organization
achieves the best applicant selection for the job), the social process para-
digm (the job role is negotiated) and the person-organization fit paradigm
(an individuals interacts with an organization through an environment
which influences the interaction) (see Searl and Al-Sharif, 2008, pp.216-219).
Digital interviews are mainly considered in the HRM literature from a pre-
dictivist paradigm: since they are used in pre-selection, the most interac-
tional and negotiation aspects of recruitment come later, when the pool of
best relevant applicants are selected. Subjects of importance for virtual
interview design are whether the design will allow accessibility, interview
performance, job attractiveness, job offer acceptance, and also organiza-
tional attraction (Lukacik et al., 2022). This question of attraction, linked
to the applicant’s reactions, is very important in the case of an AI job
interviewer. It suggests that person-organization fit is a relevant approach
to analyze such digital interviews. Our results indicate that person-organization
fit is a relevant perspective to analyze digital interview embedding anthro-
pomorphic features. The agent establishes a relation incorporating an emo-
tional dimension, and this relation is lived as a genuine experience with the
hiring organization. The studies about reactions to digital interviews could
be offered a more complete and comprehensive scope of analysis if the AI
job interviewer is really conceived as an organizational agent.
Finally, in practice, our findings should aid in the design of mediated
interaction systems, particularly by highlighting the usefulness of endow-
ing intelligent virtual agents with social skills. Such attributes can enhance
the user experience or may at least avoid degrading applicants’ percep-
tions of hiring companies that use such tools in their recruitment process.

Limits of the study and avenues for research

Based on that main result, the design of the study deserves special atten-
tion to delineates precisely the domain of relevance of results. Notably,
228 C. CLAVEL ET AL.

our study assessed a job interview as completed by students before real

face-to-face interviews in a job fair. One limitation of this study is the
choice to use a sample composed solely of first-year BSc students from
vocational programs. These students were specifically chosen due to their
orientation towards rapid professional integration, making them relevant
for a study focused on job market preparedness. However, this sample
may limit the generalizability of the results to a broader student popula-
tion or to individuals in other educational stages, such as those in more
advanced or non-vocational programs.
Another potential limitation concerns the order in which participants
responded to questions about perceived intimacy in the AI job inter-
viewer versus self-reported intimacy. Research on method biases suggests
that participants may seek consistency or avoid rating themselves as ‘less
intimate’ than a non-human agent, potentially leading to social desirabil-
ity bias (Podsakoff et al., 2003). While our study focused on real-time
interaction, where the perception of virtual intimacy is not inherently
assured, future studies could investigate the effects of question order
more thoroughly, as this may influence participant responses.
Further study should be envisioned in real-world interview settings,
but such a research design would be challenging to implement due to the
high stakes of selection interviews. Here, the situation involves a stake
because the candidates anticipate the next step, and not because a deci-
sion is taken based on this digital interview. To refine this result, an
additional study might assess applicants’ reactions in both experimental
conditions to investigate how expression of intimacy affects interview
quality and applicants’ employability.
Another refinement could concern the self-perception of intimacy, as
reflecting the level of intimacy communicated by the agent, by directly
recording and analyzing verbal and nonverbal behavior (direct measure).
In the present experimental design, our result is based on a post exper-
iment self-reported questionnaire (indirect measure), which could inter-
estingly be complemented by direct measure. By measuring these
behavioral data, more insights into the emotional side of the interaction
could be reported, for example by measuring the contagion effect.
Our research team is also particularly interested in the feelings and
opinions of young users regarding digital recruitment tools, as this audi-
ence will confront various aspects of the digitalization in HRM. To that
end, we aim to explore applicants’ overall experience with the AI job
interviewer through a qualitative design based on content analysis meth-
ods, to study their opinions of these devices. We expect that this com-
plementary approach will provide insights on the specific aspects of
AI-based HR devices that make the interview experience pleasant or
unpleasant for applicants.
 The International Journal of Human Resource Management 229

Conclusions and perspectives

By studying the impact of anthropomorphic features attributed to a virtual
recruiter on various individual-level HRM-relevant outcome variables, we
make an important contribution to the HRM literature concerning the social
dimension of the interaction in AI-based interview procedures. The present
study assessed an interactive job interview experience in which an AI virtual
job interviewer exhibited verbal and nonverbal intimacy expressions or no
intimacy expressions. Applicants perceived themselves and the AI job inter-
viewer as more intimate in the intimacy condition, compared with applicants
in the no-intimacy condition. These results underline that highly reciprocal
intimacy processes can emerge even in socially limited and heavily structured
environments, such as a formal job interview. Additionally, our findings
demonstrated that expression of intimacy by the AI job interviewer had a
positive impact on overall user experience by increasing applicants’ positive
emotions. The strength of the results lies in the real context and dynamic
nature of the job interview session, which was specifically designed for the
experiment. By investigating perceptions of applicants who would be com-
pleting a real job interview (with a human recruiter) right after the digital
training session, we could approximate real-world hiring conditions to ensure
that participants took the digital interview session seriously.
By pointing out the importance of AI job interviewers endowed with social
skills to improve applicants’ social perceptions and create a satisfactory can-
didate experience, the present study offers promising avenues for improving
the quality and effectiveness of HR tools using AI devices. It also raises fun-
damental questions on organizational and ethical aspects related to the use of
artificial systems for HR tasks. How much can the affective dimension be
developed, to mitigate anxiety or negative perception effects, without posing
ethical concerns? How could these tools be integrated into the organization
to assist HR teams without depriving recruiters of a key aspect of their work
and affecting HR managers’ professional identities? Accordingly, following the
distinction in recruitment paradigms by Searl and Al-Sharif (2088), we also
suggest a change in paradigm concerning AI job interview, from a predictiv-
ist approach in early phases of recruitment, to a person-organization fit
approach, where the AI job interviewer is an organizational agent. Such might
draw a focus on the social and emotional aspects of digital interviews and its
role in organizational attraction.
Accordingly, for further research, the management side of the use of vir-
tual agents for recruitment would be particularly important to explore further,
in order to understand how HRM reconciles the multiple concerns induced
by the implementation of AI job interviewers, including organizational attrac-
tivity, ethics of the recruitment procedure, efficiency of recruitment inter-
views, HR professionals’ role, and applicants’ experience, among others.
230 C. CLAVEL ET AL.

Acknowledgments
We also thank all students who participated in the present study.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by a research collaboration. We convey our gratitude to the
company DAVI, which helped develop and design the research prototype for the experiment.

ORCID
Sophie d’Armagnac [Link]

Data availability statement

The data that support the findings of this study are available from the corresponding
author, upon reasonable request.

References
Basch, J. M., Melchers, K. G., Kurz, A., Krieger, M., & Miller, L. (2021). It takes more
than a good camera: Which factors contribute to differences between face-to-face in-
terviews and videoconference interviews regarding performance ratings and interview-
ee perceptions? Journal of Business and Psychology, 36(5), 921–940. [Link]
org/10.1007/s10869-020-09714-3
Bickmore, T., Schulman, D., & Yin, L. (2009). Engagement vs. deceit: Virtual humans
with human autobiographies. In Z. Ruttkay, M. Kipp, A. Nijholt, & H. H. Vilhjálmsson
(Eds.), International Conference on Intelligent Virtual Agents (pp. 6–19). Springer.
[Link]
Blacksmith, N., Willford, J. C., & Behrend, T. S. (2016). Technology in the employment
interview: A meta-analysis and future research agenda. Personnel Assessment and
Decisions, 2(1), 12–20. [Link]
Callejas, Z., Ravenet, B., Ochs, M., & Pelachaud, C. (2014). A computational model of
social attitudes for a virtual recruiter [Paper presentation]. International Conference
On Autonomous Agents and Multi-Agent Systems (AAMAS) (pp. 93–100). https://
[Link]/doi/10.5555/2615731.2615749
Chapman, D. S., Uggerslev, K. L., & Webster, J. (2003). Applicant reactions to face-to-
face and technology-mediated interviews: A field investigation. The Journal of Applied
Psychology, 88(5), 944–953. [Link]
Chapman, D. S., Uggerslev, K. L., Carroll, S. A., Piasentin, K. A., & Jones, D. A. (2005).
Applicant attraction to organizations and job choice: A meta-analytic review of the
correlates of recruiting outcomes. The Journal of Applied Psychology, 90(5), 928–944.
[Link]
 The International Journal of Human Resource Management 231

Dudley, R. M. (2014). Uniform central limit theorems (Vol. 142). Cambridge University
Press. [Link]
Dutta, D., Mishra, S. K., & Tyagi, D. (2023). Augmented employee voice and employee
engagement using artificial intelligence-enabled chatbots: A field study. The International
Journal of Human Resource Management, 34(12), 2451–2480. [Link]
9585192.2022.2085525
Gardner, H., & Hatch, H. (1989). Multiple intelligences to school: Educational implica-
tions of the theory of multiple intelligence. Educational Researcher, 18(8), 4–9. https://
[Link]/10.2307/1176460
Gartner. (2021). Emerging technologies and trends impact radar: 2022. [Link]
[Link]/en/documents/4008322
Gebhard, P., Baur, T., Damian, I., Mehlmann, G., Wagner, J., & André, E. (2014). Exploring
interaction strategies for virtual characters to induce stress in simulated job interviews [Paper
presentation]. Proceedings of the international conference on autonomous agents and
multi-agent systems (pp. 661–668). [Link]
Grover, R. L., Nangle, D. W., Buffie, M. L., & Andrews, L. A. (2020). Defining social
skills. In D.W. Nangle, C. A. Erdley, & R. Schwartz-Mette (Eds.), Social skills across
the lifespan: Theory, assessment, and intervention (pp. 3–24). Elsevier. [Link]
org/10.1016/B978-0-12-817752-5.00001-9
Gursoy, D., Chi, O. H., Lu, L., & Nunkoo, R. (2019). Consumers acceptance of artificial-
ly intelligent (AI) device use in service delivery. International Journal of Information
Management, 49, 157–169. [Link]
Inoue, K., Hara, K., Lala, D., Yamamoto, K., Nakamura, S., Takanashi, K., & Kawahara,
T. (2020). Job interviewer Android with elaborate follow-up question generation [Paper
presentation]. Proceedings of the 2020 International Conference on Multimodal
Interaction, October 25–29, 2020. Virtual event, Netherlands. ACM, New York, NY,
USA, 9 pages. [Link]
ISO (International Organization for Standardization). (2010). ISO Standard 9241-210. Ergonomics
of human-system interaction—Part 210: Human-centred design for interactive systems.
Jones, C. L. E., Hancock, T., Kazandjian, B., & Voorhees, C. M. (2022). Engaging the
Avatar: The effects of authenticity signals during chat-based service recoveries. Journal
of Business Research, 144, 703–716. [Link]
Kim, J., Kang, S., & Bae, J. (2022). Human likeness and attachment effect on the per-
ceived interactivity of AI speakers. Journal of Business Research, 144, 797–804. https://
[Link]/10.1016/[Link].2022.02.047
Klein, S., & Watson-Manheim, M. B. (2021). The (re-) configuration of digital work in
the wake of profound technological innovation: Constellations and hidden work.
Information and Organization, 31(4), 100377. [Link]
100377
Kull, A. J., Romero, M., & Monahan, L. (2021). How may I help you? Driving brand
engagement through the warmth of an initial chatbot message. Journal of Business
Research, 135, 840–850. [Link]
Lallemand, C., & Koenig, V. (2017). How could an intranet be like a friend to me?” Why
standardized UX scales don’t always fit [Paper presentation]. Proceedings of the
European Conference on Cognitive Ergonomics 2017 (pp. 9–16). ACM. [Link]
org/10.1145/3121283.3121288
Langer, M., König, C. J., & Hemsing, V. (2020). Is anybody listening? The impact of automat-
ically evaluated job interviews on impression management and applicant reactions. Journal
of Managerial Psychology, 35(4), 271–284. [Link]
232 C. CLAVEL ET AL.

Langer, M., König, C. J., & Krause, K. (2017). Examining digital interviews for personnel
selection: Applicant reactions and interviewer ratings. International Journal of Selection
and Assessment, 25(4), 371–382. [Link]
Langer, M., & König, C. J. (2018). Introducing and testing the creepiness of situation
scale (CRoSS). Frontiers in Psychology, 9, 2220. [Link]
Laurenceau, J.‑P., Barrett, L. F., & Rovine, M. J. (2005). The interpersonal process mod-
el of intimacy in marriage: A daily‑diary and multilevel modeling approach. Journal
of Family Psychology, 19, 314–323. [Link]
Langer, M., König, C. J., & Papathanasiou, M. (2019). Highly automated job interviews:
Acceptance under the influence of stakes. International Journal of Selection and
Assessment, 27(3), 217–234. [Link]
Lee, S., & Choi, J. (2017). Enhancing user experience with conversational agent for mov-
ie recommendation: Effects of self-disclosure and reciprocity. International Journal of
Human-Computer Studies, 103, 95–105. [Link]
Lee, Y. C., Yamashita, N., & Huang, Y. (2020). Let’s work together: Integrating human
support with conversational agents. [Link]
prod/2020/07/[Link]
Levashina, J., Hartwell, C. J., Morgeson, F. P., & Campion, M. A. (2014). The structured
employment interview: Narrative and quantitative review of the research literature.
Personnel Psychology, 67(1), 241–293. [Link]
Ling, E. C., Tussyadiah, I., Tuomi, A., Stienmetz, J., & Ioannou, A. (2021). Factors in-
fluencing users’ adoption and use of conversational agents: A systematic review.
Psychology & Marketing, 38(7), 1031–1051. [Link]
Lomanowska, A. M., & Guitton, M. J. (2016). Online intimacy and well-being in the
digital age. Internet Interventions, 4, 138–144. [Link]
Lombard, M., & Ditton, T. (2006). At the heart of it all: The concept of presence. Journal of
Computer-Mediated Communication, 3(2), 0–0. [Link]
tb00072.x
Loveys, K., Sebaratnam, G., Sagar, M., & Broadbent, E. (2020). The effect of design fea-
tures on relationship quality with embodied conversational agents: A systematic re-
view. International Journal of Social Robotics, 12(6), 1293–1312. [Link]
s12369-020-00680-7
Lukacik, E. R., Bourdage, J. S., & Roulin, N. (2022). Into the void: A conceptual model
and research agenda for the design and use of asynchronous video interviews. Human
Resource Management Review, 32(1), 100789. [Link]
Mahlke, S. (2008). User experience of interaction with technical systems [Unpublished doctoral
thesis, Technischen Universität Berlin]. [Link]
Mahlke, S., & Lindgaard, G. (2007). Emotional Experiences and Quality Perceptions of
Interactive Products. In: Jacko, J.A. (Eds.), Human-computer interaction. Interaction
design and usability. HCI 2007. Lecture Notes in Computer Science, vol. 4550. Springer.
[Link]
Malik, A., Budhwar, P., Patel, C., & Srikanth, N. R. (2022). May the bots be with you!
Delivering HR cost-effectiveness and individualised employee experiences in an MNE.
The International Journal of Human Resource Management, 33(6), 1148–1178. https://
[Link]/10.1080/09585192.2020.1859582
McColl, R., & Michelotti, M. (2019). Sorry, could you repeat the question? Exploring
video-interview recruitment practice in HRM. Human Resource Management Journal,
29(4), 637–656. [Link]
 The International Journal of Human Resource Management 233

Melchers, K. G., Ingold, P. V., Wilhelmy, A., & Kleinmann, M. (2015). Beyond validity:
Shedding light on the social situation in employment interviews. In I. Nikolaou & J.
K. Oostrom (Eds.), Employee recruitment, selection, and assessment: Contemporary is-
sues for theory and practice (pp. 154–171). Routledge/Taylor & Francis Group. https://
[Link]/10.5167/uzh-110989
Melchers, K. G., Petrig, A., Basch, J. M., & Sauer, J. (2020). A comparison of conven-
tional and technology-mediated selection interviews with regard to interviewees’ per-
formance, perceptions, strain, and anxiety. Frontiers in Psychology, 11, 603632. https://
[Link]/10.3389/fpsyg.2020.603632
Meijerink, J., Boons, M., Keegan, A., & Marler, J. (2021). Algorithmic human resource
management: Synthesizing developments and cross-disciplinary insights on digital
HRM. The International Journal of Human Resource Management, 32(12), 2545–2562.
[Link]
Mell, J., Lucas, G., & Gratch, J. (2017). Prestige questions, online agents, and gender-driven
differences in disclosure. In J. Beskow, C. Peters, G. Castellano, C. O’Sullivan, I. Leite,
& S. Kopp (Eds.), Intelligent virtual agents lecture notes in computer science. Springer.
[Link]
Moon, Y., & Nass, C. (1996). How “real” are computer personalities? Psychological re-
sponses to personality types in human-computer interaction. Communication Research,
23(6), 651–674. [Link]
Morton, T. L. (1978). Intimacy and reciprocity of exchange: A comparison of spouses
and strangers. Journal of Personality and Social Psychology, 36(1), 72–81. [Link]
org/10.1037/0022-3514.36.1.72
Myllymäki, D. (2021). Beyond the ‘e-’in e-HRM: Integrating a sociomaterial perspective.
The International Journal of Human Resource Management, 32(12), 2563–2591. https://
[Link]/10.1080/09585192.2021.1913624
Norman, D., Miller, J., & Henderson, A. (1995). What you see, some of what’s in the
future, and how we go about doing it: HI at Apple computer in Computing Systems
[Paper presentation]. In CHI ‘95: Conference Companion on Human Factors (p. 155).
[Link]
Oh, C. S., Bailenson, J. N., & Welch, G. F. (2018). A systematic review of social pres-
ence: Definition, antecedents, and implications. Frontiers in Robotic and AI, 5, 1–34.
[Link]
Pan, Y., Froese, F., Liu, N., Hu, Y., & Ye, M. (2022). The adoption of artificial intelligence in
employee recruitment: The influence of contextual factors. The International Journal of Human
Resource Management, 33(6), 1125–1147. [Link]
Pelachaud, C. (2009). Modelling multimodal expression of emotion in a virtual agent.
Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences,
364(1535), 3539–3548. [Link]
Picard, R. W. (1999). Affective computing for HCI [Paper presentation]. Proceedings of
HCI International (the 8th International Conference on Human-Computer Interaction)
on Human-Computer Interaction: Ergonomics and User Interfaces (vol. I, pp. 829–
833). L. Erlbaum Associates.
Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common meth-
od biases in behavioral research: a critical review of the literature and recommended
remedies. Journal of applied psychology, 88(5), 879.
Potdevin, D., Clavel, C., & Sabouret, N. (2018). Virtual intimacy, this little something between
us: A study about human perception of intimate behaviors in embodied conversational
agents [Paper presentation]. Proceedings of the 18th International Conference on Intelligent
Virtual Agents (pp. 165–172). [Link]
234 C. CLAVEL ET AL.

Potdevin, D., Sabouret, N., & Clavel, C. (2020). Intimacy perception: Does the artificial
or human nature of the interlocutor matter? International Journal of Human-Computer
Studies, 142, 102464. [Link]
Potdevin, D., Clavel, C., & Sabouret, N. (2021a). Virtual intimacy in human-embodied con-
versational agent interactions: The influence of multimodality on its perception. Journal on
Multimodal User Interfaces, 15(1), 25–43. [Link]
Potdevin, D., Clavel, C., & Sabouret, N. (2021b). A virtual tourist counselor expressing
intimacy behaviors: A new perspective to create emotion in visitors and offer them a
better user experience? International Journal of Human-Computer Studies, 150, 102612.
[Link]
Potosky, D. (2008). A conceptual framework for the role of the administration medium
in the personnel assessment process. Academy of Management Review, 33(3), 629–648.
[Link]
Prager, K. J. (1997). The psychology of intimacy. Guilford Press.
Prager, K. J. (2000). Intimacy in personal relationships. In C. Hendrick & S. S. Hendrick
(Eds.), Close relationships: A sourcebook (pp. 228–243). Sage Publications.
Reeves, B., & Nass, C. I. (1996). The media equation: How people treat computers, televi-
sion, and new media like real people and places. Cambridge university press.
Schneider, L., Powell, D. M., & Bonaccio, S. (2019). Does interview anxiety predict job
performance and does it influence the predictive validity of interviews? International
Journal of Selection and Assessment, 27(4), 328–336. [Link]
Searle, R. H., & Al‑Sharif, R. (2018). Recruitment and selection. In Human Resource
Management: A Critical Approach (2nd ed.). Collings, D. G., Wood, G. T., & Szamosi,
L.T. (Eds.). (pp. 215–237). Routledge.
Stevens, C. K., & Kristof, A. L. (1995). Making the right impression: A field study of
applicant impression management during job interviews. Journal of Applied Psychology,
80(5), 587–606. [Link]
Su, M. H., Wu, C. H., & Chang, Y. (2019). Follow-up question generation using neural
tensor network-based domain ontology population in an interview coaching system
[Paper presentation] Interspeech 2019, Graz, Austria (pp. 4185–4189). [Link]
org/10.21437/Interspeech.2019-1300
Vrontis, D., Christofi, M., Pereira, V., Tarba, S., Makrides, A., & Trichina, E. (2022).
Artificial intelligence, robotics, advanced technologies and human resource manage-
ment: A systematic review. The International Journal of Human Resource Management,
33(6), 1237–1266. [Link]
Weiskopf, R., & Hansen, H. K. (2023). Algorithmic governmentality and the space of
ethics: Examples from ‘People Analytics’. Human Relations, 76(3), 483–506. [Link]
org/10.1177/00187267221075346
Woods, S. A., Ahmed, S., Nikolaou, I., Costa, A. C., & Anderson, N. R. (2020). Personnel
selection in the digital age: A review of validity and applicant reactions, and future
research challenges. European Journal of Work and Organizational Psychology, 29(1),
64–77. [Link]
Yang, H., & Lee, H. (2019). Understanding user behavior of virtual personal assistant
devices. Information Systems and e-Business Management, 17(1), 65–87. [Link]
org/10.1007/s10257-018-0375-1
Yim, C. K., Tse, D. K., & Chan, K. W. (2008). Strengthening customer loyalty through inti-
macy and passion: Roles of customer–firm affection and customer–staff relationships in
services. Journal of Marketing Research, 45(6), 741–756. [Link]