Conference PaperPDF Available

Can Humanizing Voice Assistants Unleash the Potential of Voice Commerce?

Authors:

Abstract and Figures

Voice commerce allows customers to carry out sales dialogues with voice assistants (VAs) through natural spoken language. However, its adoption remains limited. To help determine how to overcome existing barriers to adoption, we conducted a series of three empirical pre-studies and a laboratory experiment (N = 323) investigating the role of VAs' humanness in interactions with customers; research has reached no consensus on this matter. Our results reveal that humanizing VAs increases customers' perceptions of social presence and parasocial interaction, thereby enhancing perceived relationship quality and ultimately leading to increased intentions to shop using the VA. Although, we also find a negative direct effect of humanization on parasocial interaction, it is offset by the larger positive indirect effect via social presence. This may provide one explanation for the inconsistencies in the literature. For practitioners, our findings highlight the importance of careful design in humanizing VAs to increase voice commerce adoption.
Content may be subject to copyright.
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
1
Can Humanizing Voice Assistants Unleash
the Potential of Voice Commerce?
Completed Research Paper
Fabian Reinkemeier
University of Goettingen
Goettingen, Germany
fabian.reinkemeier@
wiwi.uni-goettingen.de
Ulrich Gnewuch
Karlsruhe Institute of Technology (KIT)
Karlsruhe, Germany
ulrich.gnewuch@kit.edu
Waldemar Toporowski
University of Goettingen
Goettingen, Germany
wtoporo@uni-goettingen.de
Abstract
Voice commerce allows customers to carry out sales dialogues with voice assistants (VAs)
through natural spoken language. However, its adoption remains limited. To help
determine how to overcome existing barriers to adoption, we conducted a series of three
empirical pre-studies and a laboratory experiment (N = 323) investigating the role of
VAs’ humanness in interactions with customers; research has reached no consensus on
this matter. Our results reveal that humanizing VAs increases customers’ perceptions of
social presence and parasocial interaction, thereby enhancing perceived relationship
quality and ultimately leading to increased intentions to shop using the VA. Although, we
also find a negative direct effect of humanization on parasocial interaction, it is offset by
the larger positive indirect effect via social presence. This may provide one explanation
for the inconsistencies in the literature. For practitioners, our findings highlight the
importance of careful design in humanizing VAs to increase voice commerce adoption.
Keywords: Conversational Agent, Voice Assistants, Voice Commerce, Humanness, Social
Response Theory, Parasocial Interaction Theory, Experiment
Introduction
Until recently, technologies struggled to understand and correctly interpret human speech; it was
impossible to hold conversations with them using voice alone. This is now feasible, thanks to major
advancements in artificial intelligence (AI), including natural language processing. Such developments
have fostered the rise of voice assistants (VAs) with over a third of U.S. adults having access to them via
smart speakers, such as Amazon Echo or Google Home (Voicebot 2020).
This trend presents companies with a new means of reaching customers: they can develop their own VAs to
be employed in various functions, such as online shopping. This so-called voice commerce offers
convenience by enabling users to verbalize product wishes; ask context-related questions; and order
products in a hands-free, speech-based dialogue with round-the-clock availability (Puntoni et al. 2021;
Rzepka et al. 2020; Son and Oh 2018).
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
2
Despite its advantages, voice commerce use remains limited. Only 14.3% of smart-speaker owners turn to
them regularly for purchases (Voicebot 2020). Initial qualitative research has revealed that users value the
convenience of voice commerce but are still uncomfortable using it (Rzepka et al. 2020; Voicebot and Voysis
2018). A key barrier to a pleasant buying experience and positive intentions to use VAs for voice commerce
is the failure to establish a good relationship with VAs (Cowan et al. 2017; Mari and Algesheimer 2021;
Rzepka et al. 2020).
According to the literature, the key drivers of success in traditional sales are enabling social interactions
and building relationships with customers, which allow companies to provide a comfortable shopping
experience (Bickmore and Picard 2005; Neslin and Greenhalgh 1983). In voice commerce, a company’s VA
can act as a virtual salesperson for its customers, as users recognize VAs as entities in their own right (Pitardi
and Marriott 2021). The VA can provide services such as product recommendations or orders in a speech-
based conversation (Son and Oh 2018), which customers are more likely to rely on as they build a relationship
with the VA (Mari and Algesheimer 2021). However, customers primarily perceive humanness and build
social relationships through interpersonal conversations, which are generally lacking in the online context
(Bickmore and Picard 2005). Therefore, the key questions for research and practice are how can companies
provide more social aspects in voice commerce and whether more humanness is beneficial in this context.
Research on humancomputer interaction (HCI) has shown that companies can augment their VAs with
humanlike design elements that relate to facets of human behaviorsuch as tone of voice or speech rate;
Nass and Gong 2000)also known as social (anthropomorphic) cues (Diederich et al. 2022; Feine et al.
2019; Riquel et al. 2021; Schanke et al. 2021; Seeger et al. 2021). According to the “computers are social
actors” (CASA) paradigm (Nass and Moon 2000; Nass et al. 1994; Reeves and Nass 1996), integrating social
cues into a technology can increase its perceived humanness and make interacting with it feel more social.
In the context of VAs, implementing social cues has been shown to have both positive and negative impacts.
On the one hand, it reduces user concerns about privacy invasion by, for example, addressing the need for
social contact and compensating for a lack of trust (Benlian et al. 2019). Furthermore, a more humanlike
voice leads users to draw certain assumptions about the VA, seeing it as more advanced and competent
(Cowan et al. 2015; Guzman 2019). On the other hand, however, as VAs already convey a certain level of
social characteristics through their more natural and intuitive form of interaction (Chattaraman et al. 2019;
McLean and Osei-Frimpong 2019), humanizing could lead users to compare the VAs to real people. This
could trigger unrealistic expectations and thus alienate users (Puntoni et al. 2021). Furthermore, it can be
challenging to find the right balance of humanness to prevent users from perceiving the interaction as
awkward or manipulative, which can result in less affinity for the technology (“uncanny valley
phenomenon”; Diederich et al. 2020; Mori et al. 2012). In summary, there is no consensus in the literature
as to whether humanizing VAs is helpful for the broader adoption of voice commerce.
To exploit the full potential of voice commerce, the obstacles to its adoption must be mitigated. In the offline
world, having a positive customersalesperson relationship can increase one’s intention to shop at the
store. In the case of voice commerce, however, it is unclear to what extent companies can foster such a
relationship through perceived humanness and social aspects such as social presence (the extent to which
people interacting with a VA feel they are in the presence of another social entity; Gefen and Straub 2003)
and parasocial interaction (users viewing VAs as more social, friend-like interlocutors). Therefore, recent
research has called for examining the role of humanness in interactions with VAs (e.g., Maedche et al. 2019;
Puntoni et al. 2021). We address this research gap and pose the following research question:
RQ: How does humanizing VAs influence social presence, parasocial interaction, the userVA
relationship, and ultimately users’ intentions to use the VA for voice commerce?
To answer this question, we conducted a laboratory experiment with 323 participants who carried out a
real-time buying process using a VA with either a low or high level of perceived humanness. The key
contribution of our paper lies in revealing the effect of humanizing VAs on intention to use the VA for voice
commerce. First, we provide empirical evidence that humanizing VAs influences the social aspects that
users perceive in interacting with them. In doing so, our results also reveal novel, yet counterintuitive,
insights: Humanizing VAs has direct and indirect effects that point in opposite directions. Contrary to our
theorizing, there is an unexpected negative (rather than positive) direct effect of VA humanization on
parasocial interaction. However, a larger positive indirect effect via social presence leads to a positive total
effect of humanization on parasocial interaction. Second, we contribute to understanding how to overcome
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
3
existing barriers in voice commerce by demonstrating that companies can boost relationship quality by
fostering perceptions of VAs as socially present and thus allowing for more parasocial interactions. Third,
we offer evidence that relationship quality increases users’ intentions to use the VAs for voice commerce.
In the following, we briefly describe the current use of VAs, highlight the importance of social cues, and
introduce social response theory and parasocial interaction theory as a basis for our research model. We
then hypothesize various effects of humanizing VAs in the context of voice commerce, explain our research
design, and present our results. Finally, we discuss these results, derive their theoretical and practical
implications, critically review some limitations of our study, and highlight opportunities for future research.
Theoretical Foundations and Related Work
In recent years, rapid developments in the field of AI have enabled interactions with so-called virtual
conversational agents (CAs) using natural language via text (chatbots) or voice (VAs). To capitalize on these
developments in AI, companies have been rolling out a flood of third-party VAs (e.g., so-called skills for
Amazon Alexa). VA users primarily employ them to listen to music, ask questions (e.g., weather), or use
their favorite third-party VAs (e.g., to track their daily fitness progress with Fitbit or request a ride from
Uber; Voicebot 2020). Although companies have more recently been given the ability to use the VAs to
exploit the enormous potential of voice commerce (Sun et al. 2019), customers still harbor reservations
about this new sales channel (Rzepka et al. 2020). It is therefore important to examine how social presence,
parasocial interaction, and relationship quality influence the success of userVA interactions, as they do for
traditional sales interactions.
Social Response Theory: Computers are Social Actors
The CASA paradigm establishes that interactions between people and technologies are fundamentally social
(Nass et al. 1994). Based on this paradigm, social response theory states that people apply social rules to
technologies that exhibit humanlike traits and therefore treat technologies as if they were human (Nass and
Moon 2000; Nass et al. 1994). Social cues in the context of interacting with technologies can prompt people
to apply social scripts to the situation, eliciting the same social behaviors that they would apply to another
human, despite knowing that the technology has no feelings or human motivations (Moon 2000; Nass and
Moon 2000; Nass et al. 1994).
To leverage the implications of social response theory, prior research on information systems (IS) and HCI
has classified social cues related to VA design into three main categories: auditory, verbal, and invisible
(Feine et al. 2019). Auditory cues are those that can be heard but are nonverbal, such as pitch range (Lee
and Nass 2003; Schroeder and Epley 2016) and voice tempo (Cowell and Stanney 2005). This dimension is
a defining feature of VAs, differentiating them from traditional e-commerce websites and chatbots
(Moriuchi 2021). VAs express verbal cues with spoken words, such as greetings and farewells (“Hi,”
“Goodbye”; Bickmore and Picard 2005), while invisible cues are those that cannot be seen or heard, such
as response time (Cowell and Stanney 2005; Gnewuch et al. 2022; Schanke et al. 2021).
These social cues can be implemented into the VAs design in various ways. For example, Amazon’s Alexa
and Google’s Assistant have their own personalities, voices and slang (Moriuchi 2021), leading some users
to personify Alexa as a “she” instead of an “it” (HernandezOrtega and Ferreira 2021). Hence, VAs can also
be social actors, suggesting that their appearance and social aspects play an important role in helping users
develop positive attitudes toward (Pfeuffer et al. 2019b; Pitardi and Marriott 2021) and intentions to reuse
them (Moriuchi 2021).
Parasocial Interaction Theory
To explain how humanizing VAs affects consumers’ perceptions of social aspects in interacting with them
and influences relationship building, we also draw on the theory of parasocial interaction (Horton and Wohl
1956). Originating from the field of media psychology and communication, the theory of parasocial
interaction describes how the perceived humanness of a media figure can result in the audience
experiencing a sort of two-way human-to-human social interaction, even with no real interaction
(Hartmann 2008; Whang and Im 2021). These parasocial interactions arise because people mentally
connect with the person, establishing links between mass media communication and interpersonal (face-
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
4
to-face) settings (Hartmann 2008; Lee 2018). Lim and Kim (2011) also found that parasocial interaction
can alleviate loneliness in older consumers. This may be particularly relevant during times like COVID-19
pandemic, when people lack companionship (Xie and Pentina 2022).
In recent years, this theory has been developed further and increasingly applied in the fields of psychology
and marketing to better understand consumer perceptions of such interactions (Hartmann 2008). Notable
cases include the analyses of intentions to continue using a mobile app (Lee 2018) and interactions with
websites (Zhou and Jia 2018). In line with these studies, we define parasocial interactions between
consumers and VAs as social, friend-like exchanges in which the consumers feel at ease. Such interactions
may be particularly relevant for VAs, as their greater interactivity (advanced communicative ability with
back-and-forth dialogues) and intelligence (possibility of personalization through AI) already renders them
more conversational and social than traditional technologies (Maedche et al. 2019; McLean and Osei-
Frimpong 2019; Whang and Im 2021; Zierau et al. 2020).
To address the aforementioned challenges facing voice commerce and the uncertainty regarding how VA
humanization, social presence, and parasocial interaction affect it, we conducted a study manipulating a
VA’s design in terms of humanness.
Research Model and Hypotheses
To address our research question, we propose hypotheses specific to voice commerce and present the research
model depicted in Figure 1. We draw on social response theory and parasocial interaction theory as a guiding
theoretical lens for understanding the social aspects of interactions and how they influence relationships and
behavioral intentions (Lee 2018; Nass and Moon 2000; Nass et al. 1994; Reeves and Nass 1996). In our model,
we define humanness as the perception of humanlike characteristics in VAs. We hypothesize that a VA’s
humanness positively affects users perceptions of its social presence and the level of parasocial interaction.
This should in turn improve perceptions of relationship quality and ultimately increase users’ intentions to
use the VA for voice commerce. Our research model is detailed in the subsections below.
Figure 1. Research Model
Perceiving Social Presence
Social presence has emerged as a key success factor for technologies and e-commerce (Gefen and Straub
2003; McLean and Osei-Frimpong 2019). By creating interactions with a sense of social presence, it is not
only people but also technologies that can be perceived as salespersons (Lee and Nass 2003).
Humanizing technologies can have psychological and behavioral effects that lead people to perceive
associated social characteristics, thus giving rise to greater social presence. This is the core tenet of social
response theory, which posits that social cues trigger people to apply social rules when gauging social
presence (Lee and Nass 2003; Nass and Moon 2000). Previous research has supported the positive effect
of humanizing technologies on social presence, evaluating social cues such as customer praise via dialog
box on a computer screen (Fogg 2003) or a combination of cues in chatbots (Diederich et al. 2020).
Furthermore, Qiu and Benbasat (2009) showed that users perceive greater social presence when interacting
with a website-based recommendation agent where the output is a human voice containing many social
cues as opposed to an unnatural, computer-synthesized one. Because comparable social cues can also be
employed in VAs, we propose that their incorporation would have a similar effect in voice commerce,
making the VAs seem more humanlike and engendering a sense of social presence in the user’s mind:
H1: Increasing a VA’s humanness has a positive effect on the user’s perceptions of the VA’s social presence.
Intention to use the
VA for shopping
Humanness
Social presence
Parasocial
interaction
Relationship quality
H1
H3 H6
H4
H5
H2
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
5
Perceiving Parasocial Interaction
An important aspect closely related to humanness and social presence is the phenomenon of parasocial
interaction, which can be extended to interactive settings with technologies (Giles 2002; Hartmann 2008).
While social presence refers to the feelings of warmth, sociability, and human contact created by the VA
(Gefen and Straub 2003), parasocial interaction refers to more interpersonal, friendship-like interactions,
feelings of intimacy, and that users tend to develop mental connections to the VA (Giles 2002; Hartmann
2008; Horton and Wohl 1956; Whang and Im 2021). The phenomenon of parasocial interaction requires
the interaction to have a certain degree of perceived authenticity and social aspects (Hartmann 2008; Zhou
and Jia 2018). When people interact with a more humanlike and socially present technology, they perceive
it to be more real and authentic than artificial, leading to more interpersonal interactions (Hartmann 2008;
Nass and Moon 2000; Reeves and Nass 1996; Whang and Im 2021).
Parasocial interaction theory states that the media audience gather information about a persona as they
would about a friendfor example, observing and interpreting their appearance, voice, and conversations
(Horton and Wohl 1956). Similarly, research on chatbots has shown that people perceive a more humanlike
bot to be more familiar (Diederich et al. 2021). A humanized VA could promote even more similarity and
attraction by, for example, adopting the same language patterns of users in conversations (Cowan et al.
2015). This in turn could increase the perceived authenticity of and intimacy with the VA, leading to greater
parasocial interaction (Hartmann 2008). For example, HernandezOrtega and Ferreira (2021) showed that
consumers’ positive experiences with VAs engender a kind of passion and a stronger overall intimacy with
thema feeling that is closely related to parasocial interaction (Hartmann 2008). Furthermore, Blut et al.
(2021) concluded that people perceive more humanlike service robots to be more intelligent and likeable
because they are more similar to themselves. Moreover, they found that increasing a service robot’s
humanness makes users feel more secure about the risk and privacy invasion when interacting with it. Along
similar lines, Benlian et al. (2019) demonstrated that smart home assistants with more humanlike design
features have lesser or non-intrusive privacy effects, as their humanness compensates for a lack of trust and
increases perceived control. Based on these considerations, we argue that a more humanlike VA increases
users’ sense of safety in interactions, which is generally a characteristic of friendship-like conversations and
thus leads to stronger perceptions of parasocial interactions. Hence, we hypothesize the following:
H2: Increasing a VA’s humanness has a positive effect on the user’s perceptions of parasocial interaction.
Based on social response theory, a technology’s humanlike characteristics make interactions more social
and enjoyable because they address fundamental social needs (e.g., socializing and building trust with
interlocutors and experiencing companionship) and trigger more socially appropriate human behavior
(Benlian et al. 2019; Nass and Gong 2000). When people perceive social presence in VAs, they apply
personality-based social rules such as similarity attraction to them (Nass and Moon 2000); this familiar
behavior should lead users to view interaction with VAs as more parasocial (Han and Yang 2018;
HernandezOrtega and Ferreira 2021). Furthermore, research in the context of websites has shown that
people sometimes personify a brand as a good friend and feel greater intimacy with it (meaning they
experience a more parasocial interaction) when its website provides a virtual experience comparable to
having a real face-to-face dialogue (Zhou and Jia 2018). We therefore propose the following hypothesis:
H3: An increase in a VA’s social presence has a positive effect on the user’s perceptions of parasocial
interaction.
Perceiving Relationship Quality
Relationship quality reflects the overall strength of a customersalesperson bond (Rajaobelina and
Bergeron 2009). Ideally, this connection should be particularly strong in situations where customers face
uncertainties, such as sales transactions in e-commerce (Bickmore and Picard 2005; Crosby et al. 1990).
Prior research has shown that relationship quality is a multidimensional construct that is mostly
conceptualized through trust and satisfaction, which are critical determinants of how customers perceive
the relationship (Bickmore and Picard 2005; Crosby et al. 1990; Rajaobelina and Bergeron 2009).
Traditionally, when customers have in-store conversations with human salespersons, they evaluate the
salesperson’s appearance, social skills, and similarities to themselves. A relationship can be built based on
these social aspects (Bickmore and Picard 2005; Crosby et al. 1990). Little empirical research has been
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
6
conducted on how these social aspects affect the relationship quality in voice commerce. In related research,
it has been found that humanizing VAs can mitigate the technology’s negative effects and increase trust in
the technology by making it seem more socially present (Benlian et al. 2019). This positive influence of
social presence on the perception of trust has been demonstrated in studies focusing on websites (Gefen
and Straub 2003), recommendation agents (Qiu and Benbasat 2009), and VAs (Pitardi and Marriott 2021).
In addition, research has shown that perceiving social presence in chatbots increases user satisfaction with
the technology and service (Verhagen et al. 2014). Such social presence, in that chatbots behave more
socially, can be discerned through a chatbot’s use of social cuesfor example, a name and self-references
(Diederich et al. 2021) or more humanlike response times (Gnewuch et al. 2022). Social presence thus
appears to improve relationship quality by enhancing its key dimensions, trust and satisfaction. Hence:
H4: An increase in a VA’s social presence has a positive effect on the user’s perceptions of relationship quality.
In addition to social presence, research has indicated that parasocial interaction can influence relationship
quality. Zhou and Jia (2018) found that parasocial interaction can act as a mediator between website quality
and relationship quality. Interestingly, Whang and Im (2021) noted that consumers may view VAs to be
more of independent entities than they do commercial websites that more or less directly convey a brands
message. Therefore, parasocial interaction with VAs should have even greater effects on relationship quality
compared to websites. In the context of social chatbots, research has found that the humantechnology
relationship is stronger if the technology has more social and friend-like attributes, such as being
understanding (Skjuve et al. 2021; Xie and Pentina 2022). Furthermore, a more personalized interaction
(which is presumably more parasocial) enhances trust in the VA (Reinkemeier and Gnewuch 2022) and
satisfaction with the experience (Verhagen et al. 2014). Recently, HernandezOrtega and Ferreira (2021)
showed that natural interactions with a VA can lead users to establish intimate relationships with the
technology, as if it were human. Overall, research has shown that people react socially to technology (Fogg
2003) and that having parasocial interactions makes the exchange more comfortable (Zhou and Jia 2018).
This relaxed interaction is important for how users perceive their relationships with technologies (Skjuve
et al. 2021) and should lead to a more positive assessment. We therefore propose the following:
H5: An increase in parasocial interaction with a VA has a positive effect on the user’s perceptions of
relationship quality.
Impact on Intention to Use the VA for Shopping
Consumers’ shopping intentions reflect their intentions to use a specific sales channelin our case a VA
to purchase products and to recommend using the channel to others (Arnett et al. 2003). It is important for
companies in e-commerce to maximize these intentions, allowing them to reach new customers without the
typical high costs of acquirement and ideally boost revenue (Rajaobelina and Bergeron 2009). To achieve
this, companies must cultivate high-quality relationships, as they are closely connected to users’ future
actions (Bickmore and Picard 2005; Crosby et al. 1990).
The relationship marketing literature and studies in contexts such as recommendation agents (Qiu and
Benbasat 2009) have shown that relationship quality (or certain aspects of it) has a positive effect on use
and recommendation intentions. For example, a closer relationship can reduce uncertainties in the
purchasing process and lead to a more convenient shopping experience, which is likely to encourage both
use and endorsement of the channel (Qiu and Benbasat 2009; Wong et al. 2007). Consistent with prior
research, we therefore propose a final hypothesis:
H6: An increase in relationship quality with a VA has a positive effect on the user’s intentions to use the
VA for shopping.
Research Design
In our main study, we tested our hypotheses by conducting a laboratory experiment employing a between-
subjects design with two conditions (low vs. high level of VA humanness). In a simulated purchasing
process, participants were tasked to order a specific book using natural language via VA. Before conducting
the main experiment, we carried out three pre-studies to (1) determine suitable products for the purchasing
process, (2) develop a reliable experimental design, and (3) verify our stimuli. Table 1 provides an overview
of our four data-collection phases, which are explained in greater detail below.
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
7
Study
Purpose of Study
Sample
Pre-study 1
Choosing a suitable product category
N = 82
Pre-study 2
Building and training a fluent prototype
N = 8
Pre-study 3
Validating the manipulation (selecting social cues)
N = 157
Main study
Analyzing the impact of a VA’s humanness
N = 323
Table 1. Overview of Studies Conducted
Pre-study 1: Choosing a Suitable Product Category
Similar to previous studies with VAs (e.g., Whang and Im 2021), we conducted an online survey (N = 82) to
determine a suitable product category for the subsequent experiments. We examined four categories (personal
care, books, groceries, beverages) that have been used in previous research and represent realistic voice-
commerce scenarios (i.e., are relatively easy to order/reorder by voice and do not typically require intensive
sales conversations) (Sun et al. 2019). As participants indicated that they were significantly more likely to
purchase books via VA than items from the other categories (p < .001), we chose books as our focal category.
Pre-study 2: Building and Training a Fluent Prototype
Next, we built and trained a VA prototype to carry out the main experiment. We built the VAs using
Amazon’s Alexa Skills Kit, which enabled us to customize various functions (using coding schemes on
speech synthesis markup language with JavaScript to implement specific humanlike design elements) and
create our own interaction model. We employed Amazon’s Alexa engine for natural language processing
and chose to run it on Amazon Echo Dot (3rd Generation) smart speakers, which allow only voice for in-
and output communication and have no human embodiment. To minimize associations with the Amazon
brand, we anonymized the appearance of the smart speakers and did not use the words “Amazon” or “Alexa”
at any point in our experiments. We trained our interaction model with various conversational expressions
to allow it to recognize different formulations of the same conversational intention from participants and
thereby enable engagement (Moriuchi 2021). During this process, participants (N = 8) were asked to test
the VA prototype several times and provide direct qualitative feedback. Thus, we collected various sample
questions to ensure more fluent communication. We used the same trained underlying interaction model
in all subsequent experiments to minimize external influences on our findings.
Pre-study 3: Manipulating the Voice Assistant’s Degree of Humanness
We aimed to develop a VA exhibiting a higher degree of humanness in the treatment condition than in the
control. Previous research has found that a combination of several social cues in CAs can conflict with one
another and that increasing the number of cues does not necessarily lead to a more humanlike design
(Seeger et al. 2018; Seeger et al. 2021). Therefore, we conducted a between-subjects laboratory experiment
(N = 157) with four different humanness conditions to establish which design users perceive as most
humanlike. We selected social cuessuch as natural speech pauses, greetings and farewells, a VA name,
and acknowledgmentsbased on the following criteria: First, the cues had to be able to convey a more
humanlike design, as confirmed by numerous previous studies using comparable cues (e.g., Cowell and
Stanney 2005; Fogg 2003; Gnewuch et al. 2022; Schanke et al. 2021). Second, the design elements had to
be practically feasible, meaning that designers could implement them when modifying VAs. While the
control version included no intentionally implemented social cues, the three treatment versions integrated
progressively more of them. Participants were randomly assigned to one of the four conditions and were
instructed to order an assigned book via VA. Afterward, they rated the VA’s humanness on a 9-point
semantic differential scale (Holtgraves and Han 2007), which has been validated in the context of chatbots
(Diederich et al. 2019). As the participants perceived the VA with the highest number of social cues to be
most humanlike, we used this design as the treatment condition in our main study. Unlike Seeger et al.
(2018) discovered for text-based CAs, we did not find that social cues applied in conjunction conflict with
one another in terms of fostering perceived humanness; the data from our post-experimental survey reveal
that the VA with the highest number of intentionally implemented social cues was perceived as the most
humanlike. Hence, there does not appear to be an overload of cues in our case.
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
8
Main Study
Treatment, Data Collection, Experimental Procedure, and Sample
After analyzing the third pre-study’s results which included qualitative feedback from the survey’s free-form
field, and conducting a sparring session with two consulting experts in the branch, we made minor technical
improvements and adjustments for a more fluent interaction. Finally, we created two updated versions of
the VAs, which were identical except for the use of different social cues. The control VA exhibited a lower
level of humanness. We speak of a lower level of humanness because, for our study, interaction should still
be natural and not intentionally more machinelike; a VA inherently possesses some degree of humanness
(Chattaraman et al. 2019). Following a similar procedure used in prior research on chatbots (e.g., Diederich
et al. 2021; Riquel et al. 2021; Schanke et al. 2021), the treatment condition VA included a combined set of
social cues to render the interaction more humanlike and social. We used voices of the same gender (female)
for both VAs to avoid gender-related effects and because most current VAs are designed to be female by
default (Reinkemeier and Gnewuch 2022). Figure 2 illustrates a sample of the two manipulated dialogues
for the main study.
Figure 2. Example of the Two Dialogues and Applied Social Cues
We recruited our participants directly at the university via flyers, recruitment booth, and personal networks.
To ensure high enrollment, we rewarded participation with immediate incentives and a raffle entry.
Participation in the experiment involved several steps, including completing a purchase in real time via VA:
(1) To avoid participants behaving in an impaired manner due to the presence of another human, each was
left alone in a separate experimentation room. Participants were directed to state the anonymous ID they
were given and read the experiment’s scenario instructions on tablet PCs. This included their assigned task:
imagining being in their living rooms and buying a specific novel or academic textbook. (2) Next,
participants had to activate the smart speaker using the wake word “computer” and a predefined neutral
invocation name. (3) Upon waking the speaker, participants confirmed their comprehension of the task,
triggering the VA to begin the buying process. (4) Participants were randomly assigned to one of two
scenarios: low or high VA humanness. To ensure comparability, we followed similar research (e.g., Pfeuffer
et al. 2019a; Riquel et al. 2021) in selecting a rather guided dialogue flow that was the same for both
experiment groups, although participants received tailored answers based on their choices. The buying
process included eight interaction points where participants had to communicate naturally with the VA and
make decisions (see Figure 3). (5) Afterward, they filled out a survey on tablet PCs.
A total of 365 participants completed our laboratory experiment without any technical issues or suspicious
click-through behavior. We omitted results from 4 respondents who had already taken part in the second
or third pre-study and 38 who failed to pass the two attention checks or the comprehension question. The
final sample consisted of 323 participants (50.2% female, average age = 24.68 years), almost equally
distributed between the two groups.
1
Dialogue with a
lower level of humanness
How can I help?
Hi, Emma. I would like to
purchase […]
Hi. I'm Emma, your personal
shopping assistant.
How can I help you?
Hi, I would like to purchase
[…]
There are two alternatives.
The first book [product 1] or the
second book [product 2].
Which book should be chosen?
Okay! Ahh, I have two choices
for you. On the one hand
[product 1], on the other hand
[product 2]. Which of the two
books shall I choose for you?
Dialogue with a
higher level of humanness
Social cues
implemented
Verbal:
Greeting
Formality
Express name
Verbal:
Acknow-
ledgment
Greater lexical
diversity
Invisible:
Delayed response
Auditory:
Greater pitch
range
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
9
Figure 3. Dialogue Flow of the Configured VAs
Measurement of Constructs, Checks, and Controls
To ensure validity and reliability, we measured the latent constructs with validated self-report scales (7-
point Likert scales) from prior research and adapted them to our research environment (see Appendix for
an item overview). The questionnaire captured the perceptions of social presence adapted from Gefen and
Straub (2003) and parasocial interaction adapted from Lee (2018) and Zhou and Jia (2018). As in similar
research, we measured relationship quality as a reflectivereflective higher-order construct (Rajaobelina
and Bergeron 2009; Wong et al. 2007) that includes the dimensions of trust (McKnight et al. 2002; Qiu
and Benbasat 2009) and satisfaction (Collier and Sherrell 2010; Han and Yang 2018). This approach
allowed us to reduce the number of structural model relationships while summarizing more essential
aspects in this single multidimensional construct, as we did not seek to explain variance in relationship
qualities (Hair et al. 2018). Furthermore, we measured users’ intentions to use the VA for shopping with
established dimensions of intention to use the VA to purchase products and recommend it to others (Arnett
et al. 2003; Venkatesh and Davis 2000). Finally, as control variables to minimize confounding effects in
our research model (Nwankpa and Datta 2022), we gathered measures on age, gender, experience with
smart speakers, resistance to using new technologies for purchases (Kim and Kankanhalli 2009), and need
for interaction (Dabholkar and Bagozzi 2002).
Data Analysis and Results
Manipulation and Randomization Check
To ensure that our manipulation worked as intended, we asked participants to assess the VA’s humanness
(Cronbach’s α = .81) on a 9-point semantic differential scale (Holtgraves and Han 2007). Perceived
humanness was significantly higher in the high than in the low humanness condition (Mhigh = 7.003, Mlow
= 6.348; p < .001). This showed that our manipulation was successful. Moreover, the experiment groups
presented no significant differences with respect to the control variables (for all controls: p .816),
suggesting that the randomization was also successful.
Evaluation of the Measurement Model
We tested the research model and hypotheses with the PLS-SEM approach in SmartPLS 3 (Ringle et al. 2015).
This approach is also suitable for including our independent variable as a binary experimental variable (0 =
low, 1 = high humanness). Because we are more interested in the higher-level estimates of our higher-order
construct, we used the two-stage approach, as recommended in the literature (Sarstedt et al. 2019).
We assessed our measurement with satisfying results (see Table 2), lying above the required thresholds:
After dropping two items, all factor loadings are higher than .6 (see Appendix) (Gefen and Straub 2005).
Cronbach’s alpha (α) and the composite reliability (CR) are above .7 (Nunnally and Bernstein 1994). The
average variance extracted (AVE) of each construct is above .5 (Bagozzi and Yi 1988), and all square roots
Start (1) Ask for
customer ‘s demand
(2) Book
options
Alternativ e 2
Alternativ e 1 Yes
(4) Book
format
Paperback
Hardcover
(5) Paymen t
options
Credit car d
Invoice
(6) Buy
now?
Add to car t
Buy now
End (8) Ask for
feedback (7 ) Confirmation
of purchase
(3) Continue
purchasing?
No
Confirma tion
of task
Ask for
code
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
10
of the AVEs for each construct are greater than the variance shared with other constructs (Fornell and
Larcker 1981). As recent research has questioned the Fornell–Larcker test’s ability to reliably detect
discriminant validity problems (Henseler et al. 2015), we also used the heterotraitmonotrait ratio of
correlations (HTMT) criterion to support our findings; as recommended, the value is below the threshold
of .9 and thus discriminant validity can be established (Teo et al. 2009). Furthermore, the reliability and
validity assessment of the higher-order construct relationship quality (CR = .902; AVE = .821) draws on its
relationship with its lower-order components of trust and satisfaction (Sarstedt et al. 2019). Finally, to
ensure the correctness of results and the validity of our coming conclusions, we performed multicollinearity
diagnostics; all variance inflation factors are lower than the desired threshold of 5 (Hair et al. 2018). With
these satisfying results, we can turn to the evaluation of our hypothesis tests.
Construct
α
CR
AVE
1
2
3
4
5
6
1 Humanness
NA
NA
NA
NA
2 Social presence
.898
.925
.712
.505
.844
3 Parasocial interaction
.731
.848
.650
.197
.578
.806
4 Satisfaction
.881
.926
.807
.171
.529
.688
.899
5 Trust
.837
.877
.505
.010
.371
.563
.649
.711
6 Intention to use the
VA for shopping
.940
.961
.892
.171
.454
.609
.665
.428
.945
Note. Bold numbers on the diagonal = square root of the AVE of the given construct.
Table 2. Cronbach’s α, CR, AVE, and Inter-construct Correlations
Structural Model and Hypothesis Testing
We used bootstrap resampling methods with 5,000 samples to analyze the postulated relationships
simultaneously in one structural equation model (SEM) and conducted additional analyses. Figure 4
presents the model along with the adjusted R2 values, path coefficients, and their significance levels. Our
findings show all relationships to be significant but offer support for only five of our six hypotheses (all but
H2): In support of H1, humanness has a strong positive relationship with social presence (β = .502, p < .001).
However, contrary to our expectations in H2, humanness has not a positive but rather a significant negative
effect on parasocial interaction (β =.124, p = .014). Social presence has a strong positive effect on parasocial
interaction (β = .621, p < .001) and a significant positive effect on relationship quality (β = .159, p = .001),
supporting H3 and H4, respectively. Similarly, we found evidence for H5 with parasocial interaction having a
strong positive effect on relationship quality (β = .586, p < .001). Finally, as hypothesized in H6, we found that
the relationship quality has a strong positive effect on intention to use the VA for shopping (β = .571, p < .001).
Figure 4. Structural Model Results
To further investigate H2 in term of the direct and indirect effects of humanness on parasocial interaction,
we conducted a post-hoc mediation analysis (Zhao et al. 2010) similar to recent research (e.g., Liu et al.
2022; Nwankpa and Datta 2022; Wolf 2019). We used bootstrap tests with PLS-SEM for a more accurate
assessment of direct and indirect mediation effects (Hair et al. 2019). We found a mediation effect in that
the interaction of the indirect effects (humanness on social presence as well as social presence on parasocial
interaction) is significant (p < .001). We then classified the mediation type following Zhao et al. (2010). As
Intention to use the
VA for shopping
(R2 = .443)
Humanness
Social presence
(R2 = .249)
Parasocial
interaction
(R2 = .387)
.502***
.621*** .571***
.159**
.586***
Controls: Age, Gender,
Resistance, Experience,
Need for Interaction
Note.
***p< .001, **p< .01, *p< .05
.124*
Relationship
quality
(R2 = .520)
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
11
the indirect effect via social presence is positive (β = .312, p < .001) while the direct effect is negative (β =
.124, p = .014), we find a competitive mediation. Seeing that the total effect is significant and positive
= .187, p < .001), our post-hoc analysis reveals in terms of H2 an overall positive effect of humanness on
parasocial interaction. However, our analysis points to the existence of another mediator besides social
presence that might explain part of the relationships seen here.
We also investigated the effects of the control variables on all latent variables directly in our SEM, in line
with previous research (e.g., Liu et al. 2022; Nwankpa and Datta 2022; Wolf 2019). Significant paths
include the following: Male users seem to perceive poorer relationship quality with the VA than female users
do = −.129, p = .001). In addition, users who tend to reject new purchasing technologies perceive
interaction with the VA to be significantly less parasocial (β = −.184, p < .001) and have lower intentions to
use the VA for shopping (β = −.236, p < .001). Furthermore, users who have a greater need for interaction
perceive significantly less parasocial interaction with the VA (β = −.106, p = .015).
Discussion
In this study, we examined the impact of humanizing VAs on users’ intentions to use the VAs for voice
commerce. In line with our expectations, we found that humanizing VAs increases social presence, which
in turn increases parasocial interaction and enhances relationship quality. Moreover, parasocial interaction
has a positive effect on relationship quality, while relationship quality boosts users’ intentions to use the
VAs for shopping. However, we found a competitive mediation effect whereby humanizing VAs has a negative
(rather than the expected positive) direct effect on parasocial interaction, but a positive indirect effect via
social presence, resulting in a positive total effect. We discuss important implications in the following.
Implications for Theory
With our findings, we contribute to research on the effects of humanness and social aspects in VAs. This
stream of research offers inconsistent results and recommendations in various contexts and is lacking in-
depth research for VAs in voice commerce, leading scholars to call for research in this domain (e.g., Feine
et al. 2019; Puntoni et al. 2021). Our research addresses this gap and suggests that the inconsistencies in
the literature could be caused by the complex interplay among humanness, social presence, and parasocial
interaction that together shape users’ perceptions of VAs and intentions to use them for shopping. On the
one hand, our findings demonstrate that humanizing VAs can lead to overall positive outcomes in voice
commerce, expanding our knowledge of how people perceive increased humanness in VAs (e.g., Benlian et
al. 2019; Cowan et al. 2017). On the other hand, contrary to our theorizing, our results show a negative
direct effect of VA humanization on parasocial interaction. However, the total effect is positive, owing to a
larger positive indirect effect via social presence. The negative direct effect could be explained in various
ways: First, increasing the humanness of VAs could lead users to adjust their expectations of VAs more
toward human capabilities, as they now compare them closer to real humans (Blut et al. 2021) and therefore
pay greater attention to the nonhuman imperfections of VAs (Diederich et al. 2020; MacDorman et al.
2009). Second, the fact that VAs are still technologies might cause those interacting with them to sense a
lack of authenticity (Giles 2002; Hartmann 2008; Wuenderlich and Paluch 2017) and experience
uneasiness in accordance with the uncanny valley phenomenon (Diederich et al. 2020; Mori et al. 2012).
Finally, a humanlike design may also lead to a reduced perception of familiarity (Diederich et al. 2021) or
even cause irritation if a VA aims to closely mimic human conversation but instead appears strange and
eerie (Gnewuch et al. 2022). Our mediation analysis reveals that humanness can have both negative and
positive effects on users’ perceptions of VAs, although the negative effect can be mediated if the VA elicits
feelings of social presence. Hence, humanizing a VA will only have positive effects on users’ perception if it
evokes feelings of warmth, human contact, and sociability in interactions.
Our findings also contribute in various ways to the field of IS, especially the streams of research on the
effects of social cues and social aspects in HCI (e.g., Diederich et al. 2022; Pfeuffer et al. 2019b; Schanke et
al. 2021; Seeger et al. 2021). First, our results reveal that using a combined set of social cues in a VA’s design
can make it appear more humanlike and increase user perceptions of social presence; this strengthens social
response theory and extends it to the field of voice commerce. Second, our findings show that when VAs are
more humanlike and exhibit social presence, they can foster users’ perceptions of parasocial interaction.
This demonstrates that social response theory is directly connected to parasocial interaction theory in the
context of voice commerce. Third, our results indicate that the combined effects of these two theories
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
12
(greater social presence and parasocial interaction) enhance userVA relationships; the principles of both
theories have proven to be key drivers of relationship quality and can be extended to voice commerce.
Finally, our study enriches the under-researched field of voice commerce (e.g., Mari and Algesheimer 2021;
Rzepka et al. 2020; Son and Oh 2018; Whang and Im 2021) by highlighting that increasing VA humanness
can help in transferring paradigms from offline sales into the world of e-commerce. Furthermore, this work
expands upon the relationship marketing literature (e.g., Rajaobelina and Bergeron 2009) by showing that
social aspects in interactions between buyer and seller are important not only in traditional sales (Crosby
et al. 1990; Neslin and Greenhalgh 1983) but also for transactions in voice commerce. Along these lines,
our results supplement previous research on the impact of relationship quality on desired business
outcomes (e.g., HernandezOrtega and Ferreira 2021; Rajaobelina and Bergeron 2009; Wong et al. 2007),
particularly in terms of revealing its effect on intentions to use the relevant VAs for voice commerce.
Implications for Voice Commerce Practice
This research also has practical implications for managers and designers in voice commerce. First, our
findings suggest that to be successful in this sales channel, companies must take a closer look at the
interactions between VAs and customers. Companies designing VAs should capitalize on the plentiful
possibilities offered by existing humanlike design elements, such as auditory, verbal, and invisible cues.
Such cues can make the voice-shopping experience more social and generate desired behavioral outcomes.
For example, VA designers should mimic human speech more closely in VA conversations. This could be
achieved through features such as a voice output set to a natural speed (Cowell and Stanney 2005) and
varied pitch (Lee and Nass 2003). Furthermore, VAs should use a social communication style with their
customersas salespeople in the offline world ideally doby including elements of courtesy such as saying
hello and goodbye (Bickmore and Picard 2005; Chattaraman et al. 2019). They should also praise customers
purposefully; commending a good choice, for example, can give customers the sense that they have
performed well and boost their moods (Fogg 2003). Second, our findings highlight that humanizing VAs
can have both negative and positive effects on customers’ perceptions of VAs. Companies should therefore
carefully decide whether, which, and how many social cues to implement in humanizing their VAs. To
counteract any unwanted effects between VA humanization and parasocial interaction, companies should
ensure that the overarching goal is to enrich the VA’s social presence. They should thoroughly test VA
designs before going live and avoid adopting a one-size-fits-all strategy. Finally, as we experienced in our
second pre-study, companies must train the VA’s natural language processing engine with enough
conversational expressions to avoid errors such as not providing (adequate) answers to user requests.
Limitations and Opportunities for Future Research
As with any research, our study has some limitations that offer opportunities for further research. First,
because social cues tend to not act in isolation (Feine et al. 2019), we employed a combination of cues to create
a more humanlike perception of the VA. However, we are aware that our study does not address whether one
cue has a more significant effect than another. Future research could therefore investigate particular social
cues (e.g., tone of voice) in isolation and in different combinations. In addition, we analyzed the VAs
humanness as a binary variable, but in the future this could also be measured at multiple levels of (perceived)
humanness. Second, although ordering books via VA reflects a realistic scenario, future research could include
other products and shopping goals to increase the generalizability of the results. Third, our mediation analysis
revealed that competitive mediation is at play in our research model, with a direct effect of humanness on
parasocial interaction and a mediated effect via social presence both existing and pointing in opposite
directions (Zhao et al. 2010). This raises the question of whether there are further mediators on the direct
path between humanness and parasocial interaction that future research could explore, such as VA’s perceived
authenticity (Giles 2002; Hartmann 2008; Wuenderlich and Paluch 2017) and perceived uncanniness
(MacDorman et al. 2009; Mori et al. 2012). Fourth, in practice, it is still important to improve VA technologies
in terms of accurately understanding customers (Rzepka et al. 2020), as research in the related case of social
chatbots has shown that a lack of understanding is problematic for relationship development (Skjuve et al.
2021). Although sufficient training and constant improvement by providers such as Amazon itself will
eventually render VAs quite proficient in this regard, until then, future research could explore whether
humanizing VAs can have negative effects in the case of failure to provide an adequate response. Fifth,
although female VAs are the current default in practice and research has found positive effects of female versus
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
13
male VAs (Reinkemeier and Gnewuch 2022), future research should also consider gender-neutral and male
voices to avoid reinforcing potentially harmful stereotypes in the design of CAs (e.g., Feine et al. 2020). Finally,
our study addresses users’ intentions to use the VAs for shopping activities in one-off scenarios. This means
that the experiment reflects the starting point of using VAs and initial relationship quality, which is legitimate
but different from insights into real usage behavior and long-term orientations (Skjuve et al. 2021; Whang
and Im 2021). As VAs are often embedded in users’ daily lives, future research should extend our research
into the field, investigating elements such as long-term perceptions of humanized VAs and how users might
change their preferences for VA design over the course of a continuous relationship and establishing loyalty.
Conclusion
Voice commerce offers companies the unique opportunity to conduct speech-based sales conversations
while also granting customers numerous advantages. With our research, we aim to help unleash its potential
by offering guidance on how to overcome existing challenges surrounding the humanness of VAs in voice
commerce. In summary, our study offers three main contributions to this timely topic: First, its findings
suggest that customers who interact with more humanlike and social VAs establish higher-quality
relationships, leading to increased intentions to use them for voice commerce. However, our study also
reveals that although the total effect is positive, deploying more humanlike VAs can have certain negative
effects as well. Second, our findings add to the controversial discussion in the literature about humanizing
technology and contribute further insights to the growing knowledge base of voice commerce. Third, as a
practical contribution, our study reveals that instead of presenting machinelike VAs, companies should
bestow them with humanlike design elements, offering customers a more intuitive way to interact. Making
the userVA interaction more humanlike while ensuring that the VA is socially present could be the key to
greater VA acceptance and boosting sales in voice commerce.
References
Arnett, D. B., Laverie, D. A., and Meiers, A. 2003. “Developing parsimonious retailer equity indexes using
partial least squares analysis: a method and applications,” Journal of Retailing (79:3), pp. 161170.
Bagozzi, R. P., and Yi, Y. 1988. “On the evaluation of structural equation models,” Journal of the Academy
of Marketing Science (16:1), pp. 7494.
Benlian, A., Klumpe, J., and Hinz, O. 2019. “Mitigating the Intrusive Effects of Smart Home Assistants by
using Anthropomorphic Design Features: A Multi-Method Investigation,” Information Systems
Journal, pp. 143.
Bickmore, T. W., and Picard, R. W. 2005. “Establishing and maintaining long-term human-computer
relationships,” ACM Transactions on Computer-Human Interaction (12:2), pp. 293327.
Blut, M., Wang, C., Wünderlich, N. V., and Brock, C. 2021. “Understanding anthropomorphism in service
provision: a meta-analysis of physical robots, chatbots, and other AI,” Journal of the Academy of
Marketing Science.
Chattaraman, V., Kwon, W.-S., Gilbert, J. E., and Ross, K. 2019. “Should AI-Based, conversational digital
assistants employ social- or task-oriented interaction style? A task-competency and reciprocity
perspective for older adults,” Computers in Human Behavior (90), pp. 315330.
Collier, J. E., and Sherrell, D. L. 2010. “Examining the influence of control and convenience in a self-service
setting,” Journal of the Academy of Marketing Science (38:4), pp. 490509.
Cowan, B. R., Branigan, H. P., Obregón, M., Bugis, E., and Beale, R. 2015. “Voice anthropomorphism,
interlocutor modelling and alignment effects on syntactic choices in human−computer dialogue,”
International Journal of Human-Computer Studies (83), pp. 2742.
Cowan, B. R., Pantidi, N., Coyle, D., Morrissey, K., Clarke, P., Al-Shehri, S., Earley, D., and Bandeira, N.
2017. “”What can I help you with?: Infrequent Users Experience of Intelligent Personal Assistants,” in
Proceedings of the 19th International Conference on Human-Computer Interaction with Mobile
Devices and Services, Vienna, Austria, pp. 112.
Cowell, A. J., and Stanney, K. M. 2005. “Manipulation of non-verbal interaction style and demographic
embodiment to increase anthropomorphic computer character credibility,” International Journal of
Human-Computer Studies (62:2), pp. 281306.
Crosby, L. A., Evans, K. R., and Cowles, D. 1990. “Relationship Quality in Services Selling: An Interpersonal
Influence Perspective,” Journal of Marketing (54:3), pp. 6881.
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
14
Dabholkar, P. A., and Bagozzi, R. P. 2002. “An Attitudinal Model of Technology-Based Self-Service:
Moderating Effects of Consumer Traits and Situational Factors,” Journal of the Academy of Marketing
Science (30:3), pp. 184201.
Diederich, S., Brendel, A. B., and Kolbe, L. M. 2020. “Designing Anthropomorphic Enterprise
Conversational Agents,” Business & Information Systems Engineering (62:3), pp. 193209.
Diederich, S., Brendel, A. B., Morana, S., and Kolbe, L. M. 2022. “On the Design of and Interaction with
Conversational Agents: An Organizing and Assessing Review of Human-Computer Interaction
Research,” Journal of the Association for Information Systems (23:1), pp. 96138.
Diederich, S., Janssen-Müller, M., Brendel, A. B., and Morana, S. 2019. “Emulating Empathetic Behavior
in Online Service Encounters with Sentiment-Adaptive Responses: Insights from an Experiment with
a Conversational Agent,” in Proceedings of the 40th International Conference of Information Systems
(ICIS), Munich, Germany.
Diederich, S., Lembcke, T.-B., Brendel, A. B., and Kolbe, L. M. 2021. “Understanding the Impact that
Response Failure has on How Users Perceive Anthropomorphic Conversational Service Agents: Insights
from an Online Experiment,” AIS Transactions on Human-Computer Interaction (13:1), pp. 82103.
Feine, J., Gnewuch, U., Morana, S., and Maedche, A. 2019. “A Taxonomy of Social Cues for Conversational
Agents,” International Journal of Human-Computer Studies (132), pp. 138161.
Feine, J., Gnewuch, U., Morana, S., and Maedche, A. 2020. “Gender Bias in Chatbot Design,” in Chatbot
Research and Design, A. Følstad, T. Araujo, S. Papadopoulos and E. L.-C. Law (eds.), Cham: Springer
International Publishing, pp. 7993.
Fogg, B. J. 2003. “Computers as persuasive social actors,” in Persuasive Technology: Using Computers to
Change What we Think and Do, B. J. Fogg (ed.), San Francisco, CA, USA: Morgan Kaufmann
Publishers, pp. 89120.
Fornell, C., and Larcker, D. F. 1981. “Evaluating Structural Equation Models with Unobservable Variables
and Measurement Error,” Journal of Marketing Research (18:1), pp. 3950.
Gefen, D., and Straub, D. 2003. “Managing User Trust in B2C e-Services,” e-Service Journal (2:2), pp. 724.
Gefen, D., and Straub, D. 2005. “A Practical Guide To Factorial Validity Using PLS-Graph: Tutorial And
Annotated Example,” Communications of the Association for Information Systems (16), pp. 91109.
Giles, D. C. 2002. “Parasocial Interaction: A Review of the Literature and a Model for Future Research,”
Media Psychology (4:3), pp. 279305.
Gnewuch, U., Morana, S., Adam, M. T. P., and Maedche, A. 2022. “Opposing Effects of Response Time in
Human–Chatbot Interaction,” Business & Information Systems Engineering, pp. 139.
Guzman, A. L. 2019. “Voices in and of the machine: Source orientation toward mobile virtual assistants,”
Computers in Human Behavior (90), pp. 343350.
Hair, J. F., Black, W. C., Babin, B. J., and Anderson, R. E. 2019. Multivariate data analysis, Hampshire,
UK: Cengage Learning EMEA.
Hair, J. F., Sarstedt, M., Ringle, C. M., and Gudergan, S. 2018. Advanced issues in partial least squares
structural equation modeling, Thousand Oaks, CA: SAGE.
Han, S., and Yang, H. 2018. Understanding adoption of intelligent personal assistants,” Industrial
Management & Data Systems (118:3), pp. 618636.
Hartmann, T. 2008. “Parasocial interactions and paracommunication with new media characters,” in
Mediated Interpersonal Communication, E. Konijn, S. Utz, M. Tanis and S. Barnes (eds.), New York:
Routledge, pp. 177199.
Henseler, J., Ringle, C. M., and Sarstedt, M. 2015. “A new criterion for assessing discriminant validity in
variance-based structural equation modeling,” Journal of the Academy of Marketing Science (43:1),
pp. 115135.
HernandezOrtega, B., and Ferreira, I. 2021. “How smart experiences build service loyalty: The importance
of consumer love for smart voice assistants,” Psychology & Marketing (38:7), pp. 11221139.
Holtgraves, T., and Han, T.-L. 2007. “A procedure for studying online conversational processing using a
chat bot,” Behavior Research Methods (39:1), pp. 156163.
Horton, D., and Wohl, R. R. 1956. “Mass Communication and Para-Social Interaction,” Psychiatry (19:3),
pp. 215229.
Kim, H.-W., and Kankanhalli, A. 2009. “Investigating User Resistance to Information Systems
Implementation: A Status Quo Bias Perspective,” MIS Quarterly (33:3), pp. 567582.
Lee, K. M., and Nass, C. 2003. “Designing social presence of social actors in human computer interaction,”
in Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, Ft. Lauderdale,
Florida, USA, pp. 289296.
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
15
Lee, S. 2018. “Enhancing customers’ continued mobile app use in the service industry,” Journal of Services
Marketing (32:6), pp. 680691.
Lim, C. M., and Kim, Y.-K. 2011. “Older consumers Tv home shopping: Loneliness, parasocial interaction,
and perceived convenience,” Psychology & Marketing (28:8), pp. 763780.
Liu, X., Jiang, N., Fu, M., Cai, Z., Lim, E. T. K., and Tan, C.-W. 2022. “What Piques Users’ Curiosity on Open
Innovation Platforms? An Analysis Based on Mobile App Stores,” Information Systems Frontiers.
MacDorman, K. F., Green, R. D., Ho, C.-C., and Koch, C. T. 2009. “Too real for comfort? Uncanny responses
to computer generated faces,” Computers in Human Behavior (25:3), pp. 695710.
Maedche, A., Legner, C., Benlian, A., Berger, B., Gimpel, H., Hess, T., Hinz, O., Morana, S., and Söllner, M.
2019. “AI-Based Digital Assistants: Opportunities, Threats, and Research Perspectives,” Business &
Information Systems Engineering (61:4), pp. 535544.
Mari, A., and Algesheimer, R. 2021. “The Role of Trusting Beliefs in Voice Assistants during Voice Shopping,”
in Proceedings of the 54th Hawaii International Conference on System Sciences, pp. 40734082.
McKnight, D. H., Choudhury, V., and Kacmar, C. 2002. “Developing and Validating Trust Measures for e-
Commerce: An Integrative Typology,” Information Systems Research (13:3), pp. 334359.
McLean, G., and Osei-Frimpong, K. 2019. “Hey Alexa examine the variables influencing the use of
artificial intelligent in-home voice assistants,” Computers in Human Behavior (99), pp. 2837.
Moon, Y. 2000. “Intimate Exchanges: Using Computers to Elicit SelfDisclosure From Consumers,”
Journal of Consumer Research (26:4), pp. 323339.
Mori, M., MacDorman, K. F., and Kageki, N. 2012. “The Uncanny Valley [From the Field],” IEEE Robotics
& Automation Magazine (19:2), pp. 98100.
Moriuchi, E. 2021. “An empirical study on anthropomorphism and engagement with disembodied AIs and
consumers reuse behavior,” Psychology & Marketing (38:1), pp. 2142.
Nass, C., and Gong, L. 2000. “Speech interfaces from an evolutionary perspective,” Communications of the
ACM (43:9), pp. 3643.
Nass, C., and Moon, Y. 2000. “Machines and Mindlessness: Social Reponses to Computers,” Journal of
Social Issues (56:1), pp. 81103.
Nass, C., Steuer, J., and Tauber, E. R. 1994. “Computers are Social Actors,” in Proceedings of the ACM CHI
Conference on Human Factors in Computing Systems, Boston, USA, pp. 7278.
Neslin, S. A., and Greenhalgh, L. 1983. “Nash’s Theory of Cooperative Games as a Predictor of the Outcomes
of Buyer-Seller Negotiations: An Experiment in Media Purchasing,” Journal of Marketing Research
(20:4), pp. 368379.
Nunnally, J. C., and Bernstein, I. H. 1994. “The Assessment of Reliability,” Psychometric Theory (3:1), pp.
248292.
Nwankpa, J. K., and Datta, P. 2022. “Leapfrogging Healthcare Service Quality in Sub-Saharan Africa: The
Utility-Trust Rationale of Mobile Payment Platforms,” European Journal of Information Systems
(31:1), pp. 4057.
Pfeuffer, N., Adam, M., Toutaoui, J., Hinz, O., and Benlian, A. 2019a. “Mr. and Mrs. Conversational Agent
- Gender Stereotyping in Judge-Advisor Systems and the Role of Egocentric Bias,” in Proceedings of
the 40th International Conference of Information Systems (ICIS), Munich, Germany.
Pfeuffer, N., Benlian, A., Gimpel, H., and Hinz, O. 2019b. “Anthropomorphic Information Systems,”
Business & Information Systems Engineering (61:4), pp. 523533.
Pitardi, V., and Marriott, H. R. 2021. “Alexa, she’s not human but… Unveiling the drivers of consumers’
trust in voicebased artificial intelligence,” Psychology & Marketing (38:4), pp. 626642.
Puntoni, S., Reczek, R. W., Giesler, M., and Botti, S. 2021. “Consumers and Artificial Intelligence: An
Experiential Perspective,” Journal of Marketing (85:1), pp. 131151.
Qiu, L., and Benbasat, I. 2009. “Evaluating Anthropomorphic Product Recommendation Agents: A Social
Relationship Perspective to Designing Information Systems,” Journal of Management Information
Systems (25:4), pp. 145181.
Rajaobelina, L., and Bergeron, J. 2009. “Antecedents and consequences of buyerseller relationship quality
in the financial services industry,” International Journal of Bank Marketing (27:5), pp. 359380.
Reeves, B., and Nass, C. 1996. The media equation: How people treat computers, television, and new media
like real people and places, Stanford, CA: CSLI Publications.
Reinkemeier, F., and Gnewuch, U. 2022. “Match or Mismatch? How Matching Personality and Gender
between Voice Assistants and Users Affects Trust in Voice Commerce,” in Proceedings of the 55th
Hawaii International Conference on System Sciences, pp. 43264335.
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
16
Ringle, C. M., Wende, S., and Becker, J.-M. 2015. SmartPLS 3, Boenningstedt: SmartPLS GmbH.
Riquel, J., Brendel, A. B., Hildebrandt, F., Greve, M., and Dennis, A. R. 2021. ““F*** You!” An
Investigation of Humanness, Frustration, and Aggression in Conversational Agent Communication,” in
Proceedings of the 42nd International Conference on Information Systems (ICIS), Austin, USA.
Rzepka, C., Berger, B., and Hess, T. 2020. “Why Another Customer Channel? Consumers’ Perceived
Benefits and Costs of Voice Commerce,” in Proceedings of the 53rd Hawaii International Conference
on System Sciences, pp. 40794088.
Sarstedt, M., Hair, J. F., Cheah, J.-H., Becker, J.-M., and Ringle, C. M. 2019. “How to specify, estimate, and
validate higher-order constructs in PLS-SEM,” Australasian Marketing Journal (27:3), pp. 197211.
Schanke, S., Burtch, G., and Ray, G. 2021. “Estimating the Impact of “Humanizing” Customer Service
Chatbots,” Information Systems Research (32:3), pp. 736751.
Schroeder, J., and Epley, N. 2016. “Mistaking minds and machines: How speech affects dehumanization
and anthropomorphism,” Journal of Experimental Psychology: General (145:11), pp. 14271437.
Seeger, A.-M., Pfeiffer, J., and Heinzl, A. 2018. “Designing Anthropomorphic Conversational Agents:
Development and Empirical Evaluation of a Design Framework,” in Proceedings of the 39th
International Conference of Information Systems (ICIS), San Francisco, CA, USA.
Seeger, A.-M., Pfeiffer, J., and Heinzl, A. 2021. “Texting with Human-like Conversational Agents: Designing
for Anthropomorphism,” Journal of the Association for Information Systems (4:22).
Skjuve, M., Følstad, A., Fostervold, K. I., and Brandtzæg, P. B. 2021. “My Chatbot Companion - a Study of
Human-Chatbot Relationships,” International Journal of Human-Computer Studies (149).
Son, Y., and Oh, W. 2018. Alexa, Buy Me a Movie!: How AI Speakers Reshape Digital Content
Consumption and Preference,” in Proceedings of the 39th International Conference of Information
Systems (ICIS), San Francisco, CA, USA.
Sun, C., Shi, Z., Liu, X., Ghose, A., Li, X., and Xiong, F. 2019. “The Effect of Voice AI on Consumer Purchase
and Search Behavior,” NYU Stern School of Business, pp. 143.
Teo, T. S. H., Srivastava, S. C., and Jiang, L. 2009. “Trust and Electronic Government Success: An Empirical
Study,” Journal of Management Information Systems (25:3), pp. 99132.
Venkatesh, V., and Davis, F. D. 2000. “A Theoretical Extension of the Technology Acceptance Model: Four
Longitudinal Field Studies,” Management Science (46:2), pp. 186204.
Verhagen, T., van Nes, J., Feldberg, F., and van Dolen, W. 2014. “Virtual Customer Service Agents: Using
Social Presence and Personalization to Shape Online Service Encounters,” Journal of Computer-
Mediated Communication (19:3), pp. 529545.
Voicebot. 2020. Smart Speaker Consumer Adoption Report. https://research.voicebot.ai/download-
smart-speaker-consumer-adoption-2020-executive-summary. Accessed 5 September 2022.
Voicebot, and Voysis. 2018. Voice Shopping Consumer Report. https://voicebot.ai/voice-shopping-report-
2018. Accessed 5 September 2022.
Whang, C., and Im, H. 2021. “”I Like Your Suggestion! the role of humanlikeness and parasocial
relationship on the website versus voice shopper’s perception of recommendations,” Psychology &
Marketing (38:4), pp. 581595.
Wolf, T. 2019. “Intensifying User Loyalty Through Service Gamification: Motivational Experiences and
Their Impact on Hedonic and Utilitarian Value,” in Proceedings of the 40th International Conference
of Information Systems (ICIS), Munich, Germany.
Wong, Y. H., Hung, H., and Chow, W. 2007. “Mediating effects of relationship quality on customer
relationships: An empirical study in Hong Kong,” Marketing Intelligence & Planning (25:6), pp. 581
596.
Wuenderlich, N. V., and Paluch, S. 2017. “A Nice and Friendly Chat with a Bot: User Perceptions of AI-
Based Service Agents,” in Proceeding of the 38th International Conference on Information Systems
(ICIS), Seoul, South Korea, pp. 111.
Xie, T., and Pentina, I. 2022. “Attachment Theory as a Framework to Understand Relationships with Social
Chatbots: A Case Study of Replika,” in Proceedings of the 55th Hawaii International Conference on
System Sciences.
Zhao, X., Lynch, J. G., and Chen, Q. 2010. “Reconsidering Baron and Kenny: Myths and Truths about
Mediation Analysis,” Journal of Consumer Research (37:2), pp. 197206.
Zhou, F., and Jia, W. 2018. “How a Retailer’s Website Quality Fosters Relationship Quality: The Mediating
Effects of Parasocial Interaction and Psychological Distance,” International Journal of Human-
Computer Interaction (34:1), pp. 7383.
Humanizing Voice Assistants in Voice Commerce
Forty-Third International Conference on Information Systems, Copenhagen 2022
17
Zierau, N., Elshan, E., Visini, C., and Janson, A. 2020. “A Review of the Empirical Literature on
Conversational Agents and Future Research Directions,” in Proceedings of the 41st International
Conference of Information Systems (ICIS).
Appendix
Construct
Item
Loading
Source(s)
Perceived
humanness
How humanlike did you perceive the voice assistant to be
during the interaction?
.719
Diederich
et al.
(2019);
Holtgraves
and Han
(2007)
How skilled do you perceive the voice assistant to be?
.789
How thoughtful do you perceive the voice assistant to be?
.774
How polite do you perceive the voice assistant to be?
.634
How responsive do you perceive the voice assistant to be?
.752
How engaging do you perceive the voice assistant to be?
.681
Social
presence
While interacting with the voice assistant, I felt a sense of …
Gefen and
Straub
(2003)
… human contact.
.824
… personalness.
.846
… human warmth.
.863
… sociability.
.803
… human sensitivity.
.880
Parasocial
interaction
Interacting with the voice assistant made me comfortable,
as if I were with a friend.
.794
Lee (2018);
Zhou and
Jia (2018)
When I interact with the voice assistant, I feel included.
.812
Interacting with the voice assistant made me relax.
.813
Relationship
quality
Satisfaction:
Collier and
Sherrell
(2010);
Han and
Yang
(2018)
Overall, I am very satisfied with the voice assistant.
.907
I am very pleased with the voice assistant.
.928
I am pleased with the quality of this voice assistant’s
purchasing support.
.860
Trust:
McKnight
et al.
(2002);
Qiu and
Benbasat
(2009)
In the purchase process, the voice assistant appeared ...
... honest with me.
.646
... sincere and genuine.
.672
... objective in its product recommendations. (dropped)
.492
In the purchase process, the voice assistant was ...
… competent.
.763
… very knowledgeable about the products.
.723
good at satisfying my needs.
.755
The voice assistant gave the impression of
... being interested in making sure I'm okay. (dropped)
.579
... doing its best to help me with the purchase.
.719
... acting in my best interest.
.689
Intention to
use the voice
assistant for
shopping
Assuming I had access to the voice assistant in the future, I would ...
Arnett et
al. (2003);
Venkatesh
and Davis
(2000)
... use it to buy products.
.949
... frequently use it to buy products.
.949
… recommend friends to buy products through a similar
voice assistant.
.936
Table 3. Constructs, Items, and Factor Loadings
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Curiosity, a motivational state of exploratory behavior, is conducive to innovation diffusion by encouraging users’ exploration on open innovation platforms. Yet, despite its importance, there is a scarcity of research investigating the mechanism for piquing users’ curiosity. Accordingly, we advance a research model to unravel how platform service quality, in the form of service content quality and service delivery quality, affects users’ epistemic and perceptual curiosity via inducing their trust and distrust in a platform. Taking mobile app stores as our empirical context, we collected data from 431 users to validate our hypothesized relationships. Analytical results indicate that both dimensions of platform service quality positively influence users’ trust in platform, whereas only service delivery quality negatively influences users’ distrust in platform. Furthermore, trust in platform directly triggers curiosity whereas distrust in platform positively influences users’ feeling-of-deprivation, which in turn triggers curiosity. In this sense, our analytical results reveal the mediating roles of distrust in platform and feeling-of-deprivation in the relationship between service delivery quality and curiosity.
Article
Full-text available
Research has shown that employing social cues (e.g., name, human-like avatar) in chatbot design enhances users’ social presence perceptions and their chatbot usage intentions. However, the picture is less clear for the social cue of chatbot response time. While some researchers argue that instant responses make chatbots appear unhuman-like, others suggest that delayed responses are perceived less positively. Drawing on social response theory and expectancy violations theory, this study investigates whether users’ prior experience with chatbots clarifies the inconsistencies in the literature. In a lab experiment ( N = 202), participants interacted with a chatbot that responded either instantly or with a delay. The results reveal that a delayed response time has opposing effects on social presence and usage intentions and shed light on the differences between novice users and experienced users – that is, those who have not interacted with a chatbot before vs. those who have. This study contributes to information systems literature by identifying prior experience as a key moderating factor that shapes users’ social responses to chatbots and by reconciling inconsistencies in the literature regarding the role of chatbot response time. For practitioners, this study points out a drawback of the widely adopted “one-design-fits-all” approach to chatbot design.
Conference Paper
Full-text available
With increasing adoption of AI social chatbots, especially during the pandemic-related lockdowns, when people lack social companionship, there emerges a need for in-depth understanding and theorizing of relationship formation with digital conversational agents. Following the grounded theory approach, we analyzed in-depth interview transcripts obtained from 14 existing users of AI companion chatbot Replika. The emerging themes were interpreted through the lens of the attachment theory. Our results show that under conditions of distress and lack of human companionship, individuals can develop an attachment to social chatbots if they perceive the chatbots’ responses to offer emotional support, encouragement, and psychological security. These findings suggest that social chatbots can be used for mental health and therapeutic purposes but have the potential to cause addiction and harm real-life intimate relationships.
Conference Paper
Full-text available
Despite the ubiquity of voice assistants (VAs), they see limited adoption in the form of voice commerce, an online sales channel using natural language. A key barrier to the widespread use of voice commerce is the lack of user trust. To address this problem, we draw on similarity-attraction theory to investigate how trust is affected when VAs match the user's personality and gender. We conducted a scenario-based experiment (N = 380) with four VAs designed to have different personalities and genders by customizing only the auditory cues in their voices. The results indicate that a personality match increases trust, while the effect of a gender match on trust is non-significant. Our findings contribute to research by demonstrating that some types of matches between VAs and users are more effective than others. Moreover, we reveal that it is important for practitioners to consider auditory cues when designing VAs for voice commerce.
Conference Paper
Full-text available
Conversational Agents (CAs) are becoming part of our everyday lives, whether in the form of voice assistants (such as Siri or Alexa) or as chatbots (for instance, on Facebook). When looking at CAs, users frequently start to display aggressive behavior towards CAs, such as insulting it. However, why some users tend to harass CAs and how this behavior can be explained remains unclear. We conducted a two-conditions online experiment with 201 participants on the interrelation of human-like design, dissatisfaction, frustration, and aggression to address this circumstance. The results reveal that frustration drives aggression. Specifically, users with high impulsivity tend to severely insult a CA when it makes an error. To prevent such behavior, our results indicate that human-like design of CA reduces dissatisfaction, which is a driver of frustration. However, we also discovered that human-like design directly increases frustration, making it a double-edged sword that has to be further investigated.
Article
Full-text available
Conversational agents (CAs), described as software with which humans interact through natural language, have increasingly attracted interest in both academia and practice, due to improved capabilities driven by advances in artificial intelligence and, specifically, natural language processing. CAs are used in contexts like people's private life, education, and healthcare, as well as in organizations, to innovate and automate tasks, for example in marketing and sales or customer service. In addition to these application contexts, such agents take on different forms concerning their embodiment, the communication mode, and their (often human-like) design. Despite their popularity, many CAs are not able to fulfill expectations and to foster a positive user experience is a challenging endeavor. To better understand how CAs can be designed to fulfill their intended purpose, and how humans interact with them, a multitude of studies focusing on human-computer interaction have been carried out. These have contributed to our understanding of this technology. However, currently a structured overview of this research is missing, which impedes the systematic identification of research gaps and knowledge on which to build on in future studies. To address this issue, we have conducted an organizing and assessing review of 262 studies, applying a socio-technical lens to analyze CA research regarding the user interaction, context, agent design, as well as perception and outcome. We contribute an overview of the status quo of CA research, identify four research streams through a cluster analysis, and propose a research agenda comprising six avenues and sixteen directions to move the field forward.
Article
Full-text available
Smart voice assistants (SVAs) have emerged as new artificial intelligence service platforms. They have the capacity to act like actual human assistants and modify traditional forms of human–computer interactions. So, consumers relate to their SVA as though it was a person, despite knowing that they are interacting with a machine. Based on the stimulus‐organism‐response framework, this paper examines feelings of love that consumers develop for SVAs when they are interacting. It proposes that these feelings act as the psychological mechanism to transmit the effect of consumers' experiences with the technology (i.e., smart experiences) on their service loyalty. Feelings of love are conceptualized following the triangular theory of love that considers three components: passion, intimacy, and commitment. Service loyalty refers to consumers' intentions to continue using SVAs in the future and to recommend SVAs to other people through physical and/or electronic word of mouth communication. The results show that smart experiences influence consumers' passion for technology, while passion explains their intimacy and commitment. Consumer intimacy and commitment for SVAs lead to service loyalty. Therefore, this paper contributes to research focusing on the importance of consumers' feelings for SVAs and shedding light on the process that drives to service loyalty.
Article
Conversational agents (CAs) are natural language user interfaces that emulate human-to-human communication. Because of this emulation, research on CAs is inseparably linked to questions about anthropomorphism—the attribution of human qualities, including consciousness, intentions, and emotions, to nonhuman agents. Past research has demonstrated that anthropomorphism affects human perception and behavior in human-computer interactions by, for example, increasing trust and connectedness or stimulating social response behaviors. Based on the psychological theory of anthropomorphism and related research on computer interface design, we develop a theoretical framework for designing anthropomorphic CAs. We identify three groups of factors that stimulate anthropomorphism: technology design-related factors, task-related factors, and individual factors. Our findings from an online experiment support the derived framework but also reveal novel yet counterintuitive insights. In particular, we demonstrate that not all combinations of anthropomorphic technology design cues increase perceived anthropomorphism. For example, we find that using only nonverbal cues harms anthropomorphism; however, this effect becomes positive when nonverbal cues are complemented with verbal or human identity cues. We also find that CAs’ disposition to complete computerlike versus humanlike tasks and individuals’ disposition to anthropomorphize greatly affect perceived anthropomorphism. This work advances our understanding of anthropomorphism and contextualizes the theory of anthropomorphism within the IS discipline. We advise on the directions that research and practice should take to find the sweet spot for anthropomorphic CA design.
Article
Can mobile payment platforms (MPPs) elevate Sub-Saharan Africa’s (SSA) healthcare service quality (HSQ)? In a region home to more than one billion people yet beset by a paucity of services and infrastructure, mobile devices and platforms are rapidly gaining ground. Rising healthcare needs and costs along with the fact that more than 60% of SSA lives in areas nearly 1.5 miles away from paved roads that lead to hospitals and banks, accessibility to traditional financial services remains a serious concern. As Sub-Saharan hospitals increase billing access via mobile payment platforms, there is little understanding of whether and how mobile payment platforms influence healthcare service quality (HSQ). Drawing upon economic utility and trust perspectives, this study develops an integrative model to examine mobile payment platform utility’s impact on patient trust, commitment, and HSQ perceptions. The model is empirically tested using a survey of 527 patients from the West African country of Nigeria. Results reveal a distinct amalgam of MPP utility, trust, and commitment as antecedents to HSQ perceptions, thus generating important insights and contributing to the ongoing inquiry of how digital platforms in developing countries can influence socio-economic development. Overall, the results support the study’s hypotheses and underscore the important role of MPPs.
Article
In this work, we investigate how applying human-like characteristics to customer service chatbots can influence retail outcomes. This is an important managerial question as creating effective chatbot experiences through messaging platforms has proven difficult for organizations. Often, chatbot developers apply characteristics such as giving a chatbot a human name, adding humor, and so on, without knowing how these features influence end user behavior. Implementing a field experiment in collaboration with a dual channel clothing retailer based in the United States, we automate a used clothing buy-back process, such that individuals engage with the retailer's autonomous chatbot to describe the used clothes they wish to sell, obtain a cash offer, and (if they accept) print a shipping label to finalize the transaction. We provide evidence that, in this retail setting, anthropomorphism is beneficial for transaction outcomes, but that it also leads to significant increases in consumers’ sensitivity to the offer amount. We argue that the latter effect occurs because, as a chatbot becomes more human-like, consumers shift to a fairness evaluation or negotiating mindset.