Conference PaperPDF Available

“The Chatbot is typing …” – The Role of Typing Indicators in Human-Chatbot Interaction

Authors:

Abstract

Chatbots have attracted considerable interest in recent years. A key design challenge to increase their adoption is to make the interaction with them feel natural and human-like. Therefore, it is suggested to incorporate social cues in the chatbot design. Drawing on the Computers are Social Actors paradigm and the "uncanny valley" hypothesis, we study the effect of one specific social cue (i.e., typing indicators) on social presence of chatbots. In an online experiment, we investigate the effect of two specific designs of typing indicators. Our preliminary results indicate that graphical typing indicators increase social presence of chatbots, but only for novice users. Therefore, our results suggest that the relationship between typing indicators and perceived social presence of chatbots depends on the design of these indicators and user's prior experience. We contribute with empirical insights and design knowledge that support researchers and practitioners in understanding and designing more natural human-chatbot interactions.
This is the author’s version of a work that was published in the following source
Gnewuch, U., Adam, M. T. P., Morana, S., and Maedche, A. 2018. “‘The Chatbot Is Typing …’ - The Role
of Typing Indicators in Human-Chatbot Interaction,” in Proceedings of the 17th Annual Pre-ICIS Workshop
on HCI Research in MIS, San Francisco, CA, USA, December 13th, 2018.
Please note: Copyright is owned by the author and / or the publisher.
Commercial use is not allowed.
Institute of Information Systems and Marketing (IISM)
Fritz-Erler-Strasse 23
76133 Karlsruhe - Germany
http://iism.kit.edu
Karlsruhe Service Research Institute (KSRI)
Kaiserstraße 89
76133 Karlsruhe Germany
http://ksri.kit.edu
© 2017. This manuscript version is made available under the CC-
BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-
nc-nd/4.0/
Gnewuch et al. Role of Typing Indicators in Human-Chatbot Interaction
Proceedings of the Sixteenth Annual Pre-ICIS Workshop on HCI Research in MIS, San Francisco, CA, December 13, 2018
1
“The Chatbot is typing …” – The Role of Typing Indicators
in Human-Chatbot Interaction
Ulrich Gnewuch
Karlsruhe Institute of Technology, Germany
ulrich.gnewuch@kit.edu
Stefan Morana
Karlsruhe Institute of Technology, Germany
stefan.morana@kit.edu
Marc T. P. Adam
The University of Newcastle, Australia
marc.adam@newcastle.edu.au
Alexander Maedche
Karlsruhe Institute of Technology, Germany
alexander.maedche@kit.edu
ABSTRACT
Chatbots have attracted considerable interest in recent years.
A key design challenge to increase their adoption is to make
the interaction with them feel natural and human-like.
Therefore, it is suggested to incorporate social cues in the
chatbot design. Drawing on the Computers are Social Actors
paradigm and the “uncanny valley” hypothesis, we study the
effect of one specific social cue (i.e., typing indicators) on
social presence of chatbots. In an online experiment, we
investigate the effect of two specific designs of typing
indicators. Our preliminary results indicate that graphical
typing indicators increase social presence of chatbots, but
only for novice users. Therefore, our results suggest that the
relationship between typing indicators and perceived social
presence of chatbots depends on the design of these
indicators and user’s prior experience. We contribute with
empirical insights and design knowledge that support
researchers and practitioners in understanding and designing
more natural human-chatbot interactions.
Keywords
Chatbot, conversational agent, typing indicator, social
presence, computers are social actors, online experiment.
INTRODUCTION
Text-based conversational agents, commonly referred to as
chatbots, have received a lot of attention in recent years.
They are designed to interact with humans using natural
language and are increasingly used on messaging platforms
and websites (Dale 2016). As advances in artificial
intelligence continue, many organizations are beginning to
implement chatbots to automate customer service and
reduce costs (Gartner 2018). Despite the large interest in
chatbots, their adoption and use is growing much slower
than expected (Liedtke 2017). One key barrier to the
adoption and use of chatbots is that the interaction with them
often does not feel natural and human-like (Schuetzler et al.
2014). However, established design principles for creating
chatbot interactions that feel natural to the user are scarce
(McTear 2017). Previous research on the design of websites
and recommendation agents suggests that incorporating
social cues, such as natural language and human-like
appearance, makes the interaction more natural and
enhances users’ perceived social presence (e.g., Qiu and
Benbasat 2009).
Much of this research builds on the Computers are Social
Actors (CASA) paradigm to explain why users apply social
rules and expectations in their interaction with information
systems (IS) that incorporate social cues, such as natural
language or human-like appearance (Nass et al. 1994). Many
studies have found that social cues can positively affect
users’ perceived social presence, trust, enjoyment, and usage
intentions (e.g., Qiu and Benbasat 2009; Wakefield et al.
2011). However, it has also been shown that social cues may
backfire, particularly when they irritate users or overplay the
system’s actual capabilities (Louwerse et al. 2005). In the
context of chatbots, a design feature that has been used by
both researchers and practitioners to make chatbot
interactions appear more natural and familiar to the user is
the so-called typing indicator (Appel et al. 2012;
Klopfenstein et al. 2017). This typing indicator (e.g., three
animated dots or the message “[Person X] is typing”) was
originally developed for text-based computer-mediated
communication (CMC) systems to create awareness that the
other person is typing in order to support turn-taking
(Auerbach 2014). Today, most messaging applications have
implemented different forms of these indicators.
Consequently, it is suggested to incorporate them in the
design of chatbots in order to make the interaction with them
feel more like interacting with a real human being (Appel et
al. 2012).
Existing research on typing indicators indicates that they
may also serve as a social cue as they increase the feeling of
being closer to another person (Shin et al. 2018) and are used
to make the interaction appear more natural (Appel et al.
2012). However, due to the fact that a chatbot is a machine,
which does not type responses on a keyboard, “faking
human responses” using these indicators may adversely
affect users’ perceptions of chatbots (Klopfenstein et al.
2017, p. 559). To the best of our knowledge, little empirical
research has examined the role of typing indicators in
Gnewuch et al. Role of Typing Indicators in Human-Chatbot Interaction
Proceedings of the Sixteenth Annual Pre-ICIS Workshop on HCI Research in MIS, San Francisco, CA, December 13, 2018
2
human-chatbot interaction. Moreover, there is a lack of
knowledge on whether different typing indicator designs
influence user perceptions of chatbots differently. To
address this gap, we focus on the concept of social presence,
which has been identified as a key factor in the design of IS
in online environments. Moreover, it is an important driver
of trust, enjoyment, and usage intentions (Cyr et al. 2007),
particularly in the domain of customer service in which
chatbots are increasingly used (Gartner 2018). Thus, we
address the following research question:
How do typing indicators influence users’ perceived social
presence of chatbots in customer service?
To address our research question, we conducted a three-
condition, between-subjects online experiment to investigate
the effect of typing indicators in human-chatbot interaction.
Drawing on the CASA paradigm (Nass et al. 1994) and the
uncanny valley hypothesis (Mori 1970), we examined how
the existence and two different designs of typing indicators
influence users’ perceived social presence of chatbots. The
results of our preliminary analysis show that typing
indicators positively influence novice users’ perceived social
presence of a chatbot, while they make no difference for
experienced users. Moreover, there is a significant
difference between the two designs of typing indicators for
novice users. By investigating the design and outcomes of a
specific social cue (i.e., typing indicators), our research
contributes to existing literature on the design of text-based
conversational agents.
RELATED WORK AND THEORETICAL FOUNDATION
Chatbots are Social Actors
Previous research has used a variety of terms to describe
systems that allow users to interact with them using natural
language (e.g., conversational agent, chatbot, or virtual
assistant). It has been shown that humans respond socially
to computers exhibiting human-like characteristics (Nass et
al. 1994). The CASA paradigm posits that, when users are
confronted with these so-called social cues from a computer
(e.g., natural language or human-like appearance), they
automatically apply social rules and expectations in their
interaction with it (Nass et al. 1994). Even rudimentary
social cues are sufficient to generate a wide range of social
responses from users. Following the CASA paradigm,
many studies have investigated how users react to various
social cues from computers, robots, and other technologies.
Examples in IS research include recommendation agents
(e.g., Qiu and Benbasat 2009) and websites (e.g.,
Wakefield et al. 2011). A few studies were also conducted
in the context of chatbots to examine the effect of visual
(e.g., Appel et al. 2012) or verbal cues (e.g., Schuetzler et
al. 2014). Across these studies, social cues have been found
to positively affect user perceptions of chatbots (e.g., social
presence, trust, or engagement). However, the provision of
inappropriate social cues to overly humanize chatbots may
also backfire when they too closely resemble human beings
(Gnewuch et al. 2017; Louwerse et al. 2005). The
“uncanny valley” hypothesis states that human-like
technologies are perceived as more agreeable up until they
become so human that people find their nonhuman
imperfections unsettling (Mori 1970). Therefore, design
features, such as typing indicators, that represent social
cues need to be designed carefully to limit possible
negative impacts (Fogg 2002).
Typing Indicators and Turn-Taking
Turn-taking is a fundamental mechanism for the
organization of turns in human-human interaction (Sacks
et al. 1974). It basically describes the rules by which
participants in a conversation manage who speaks when
and for how long in order to avoid overlaps and minimize
silence between turns (Sacks et al. 1974). In face-to-face
communication, turn-taking is facilitated by a multitude of
social cues such as gesticulations, eye contact, and facial
expressions (Wiemann and Knapp 1975). Since these cues
are missing in text-based CMC, such as instant messaging,
users are usually not aware of another person’s turn
because their messages do not appear on the users’ screen
until the person typing them hits the return key. To prevent
overlaps and increase turn-taking awareness, developers of
one of the first messaging applications, Microsoft’s MSN
Messenger, invented the typing indicator (Auerbach 2014).
Once a user started typing a message, this indicator
displayed “[Person X] is typing” on the other person’s
screen, which facilitated turn-taking and substituted the
missing social cues (e.g., facial expressions) used in
human-human interaction.
Nowadays, typing indicators have been implemented in all
major messaging applications. Looking at the most widely-
used messaging applications, primarily two different types
of designs can be identified: (1) a graphical typing
indicator (3 dots) and (2) a textual typing indicator
(Typing). Both typing indicators share some similarities
with filler interfaces found on websites (e.g., loading
screens, progress bars, or “Please wait” messages), which
have been found to influence users’ waiting experience and
perceived waiting time by directing attention away from
the wait (Lee et al. 2012).
As chatbots are often implemented in the same text-based
CMC channels (e.g., Facebook Messenger), they also
increasingly use typing indicators (Klopfenstein et al.
2017). Although practitioners frequently highlight their
potential benefits, empirical research on their role in
human-chatbot interaction is scarce. More specifically, it is
not clear whether typing indicators positively affect user
perceptions of chatbots (i.e., increase social presence) and
whether different typing indicator designs influence these
perceptions differently.
RESEARCH MODEL
Although typing indicators are implemented in most
messaging applications and also used by chatbots
(Klopfenstein et al. 2017), little empirical research has been
conducted to understand how they impact human-chatbot
interaction. Therefore, we decided to adopt a two-step
Gnewuch et al. Role of Typing Indicators in Human-Chatbot Interaction
Proceedings of the Sixteenth Annual Pre-ICIS Workshop on HCI Research in MIS, San Francisco, CA, December 13, 2018
3
approach to derive our hypotheses and develop our research
model (c.f., Lee et al. 2012). Before exploring different
typing indicator designs, it needs to be examined whether the
existence of the typing indicator itself influences users’
perceived social presence of a chatbot. Drawing on the
CASA paradigm and the uncanny valley hypothesis, we
formulate four hypotheses on the effect of the existence and
two designs of typing indicators.
Social presence has been identified as a key factor in the
design of IS in online environments and an important driver
of trust, enjoyment, and usage intentions (Cyr et al. 2007).
The concept of social presence is used to understand how
feelings of warmth, sociability, and human contact can be
created without actual human contact (Gefen and Straub
2004). Previous research has shown that many social cues
incorporated in websites, recommendations agents, and
other technologies, create perceptions of social presence
(e.g., Qiu and Benbasat 2009). According to the CASA
paradigm, these perceptions are the result of an unconscious
process, in which users respond to technologies as though
they were human, despite knowing that they are interacting
with a machine (Nass et al. 1994). Drawing on CASA, we
argue that typing indicators serve as a social cue in the
interaction with a chatbot. More specifically, we believe that
when users interact with a chatbot with a typing indicator,
their perceptions of the chatbot are shaped by their social
expectations from interacting with other human beings (i.e.,
using messaging applications such as WhatsApp). Although
they know that the chatbot does not type responses on a
keyboard, they will subconsciously apply the social rules
practiced in their daily life, which in turn generates
perceptions of social presence similar to those that would be
generated if the user were interacting with another human.
Therefore, we propose that in human-chatbot interaction, the
mere existence of a typing indicator, regardless of its design
(i.e., graphical/3 Dots or textual/Typing), will lead to higher
levels of perceived social presence. Hence, we argue that:
H1a,b: Users exhibit higher levels of perceived social
presence when interacting with a chatbot with (a) a
graphical typing indicator and (b) a textual typing indicator,
compared to the same chatbot without a typing indicator.
As mentioned above, there are primarily two different
designs of typing indicators used in major messaging
applications: a graphical typing indicator (3 Dots) and a
textual typing indicator (Typing). Although we argue that the
existence of typing indicators itself influences users’
perceived social presence of chatbots, we assume that there
are differences in user perceptions between the two
identified designs. However, based on theory, it is not clear
how these differences will manifest themselves. Therefore,
we formulate two contrasting hypotheses on the effect of
different typing indicator designs (graphical vs. textual) on
social presence.
First, the graphical typing indicator has a rather functional
design, similar to loading screens or progress bars of
websites. Thus, while it indicates that “something” is
happening, it does not explicitly state what happens. Since
users are familiar with the graphical typing indicator from
their use of messaging applications, they subconsciously
associate a human action (i.e., typing) with it. According to
the CASA paradigm, even such minimal social cues can
trigger a wide range of social responses (Nass et al. 1994),
resulting in increased social presence as compared to no
typing indicators. However, the textual typing indicator
explicitly states that the chatbot is “typing” while users are
waiting for a response. Therefore, it can be argued that this
explicit message is a more salient social cue because it
imitates human action. Research based on CASA has found
that when users are exposed to more or stronger social cues,
their social responses become stronger as well (e.g., von der
Pütten et al. 2010). Therefore, we propose that a textual
typing indicator generates a higher level of perceived social
presence than a graphical typing indicator:
H2a: Users exhibit higher levels of perceived social presence
when interacting with a chatbot with a textual typing
indicator compared to a graphical typing indicator.
However, previous research also points out that “turn[ing]
up the volume on the social element” of technologies can
have some undesirable side effects (Fogg 2002, p. 114) and
may even backfire (Louwerse et al. 2005). For example,
Groom et al. (2009) found that users were less comfortable
interacting with a very realistic and human-like
conversational agent than with an agent with lower realism.
The uncanny valley hypothesis states that when technologies
become more human-like (e.g., by incorporating stronger
social cues), they are perceived as more agreeable and social,
until a point beyond which the reaction is reversed (Mori
1970). Therefore, we argue that textual typing indicators
explicitly stating that the chatbot is “typing” approach the
edge of the uncanny valley. Instead of increasing users’
perceptions of social presence, these indicators might
backfire because users feel like they are “faking” human
action. Thus, we propose that a textual typing indicator
generates a lower level of social presence than a graphical
typing indicator:
H2b: Users exhibit lower levels of perceived social presence
when interacting with a chatbot with a textual typing
indicator compared to a graphical typing indicator.
Figure 1. Research Model
METHOD
To test our research model, we conducted an online
experiment, in which participants interacted with a chatbot
in a customer service context. In the experiment, participants
were provided with a hypothetical scenario of using the
chatbot to search for a cheaper mobile phone plan that fit
their individual needs. After they received a
recommendation by the chatbot and ended their
conversation, they were asked to complete a survey about
Gnewuch et al. Role of Typing Indicators in Human-Chatbot Interaction
Proceedings of the Sixteenth Annual Pre-ICIS Workshop on HCI Research in MIS, San Francisco, CA, December 13, 2018
4
their perceptions of the chatbot and their interaction with it.
Overall, 256 student subjects participated in the experiment.
They were randomly assigned to one of the experimental
conditions.
Treatment Configuration and Chatbot Scenario
The online experiment employed a between-subjects design
with three conditions (typing indicator: none, graphical,
textual) to avoid potential carry-over effects. In all
conditions, participants were told that they were interacting
with a chatbot. In the control condition (CTRL) condition,
participants interacted with a chatbot that did not display a
typing indicator before sending a message. In the first
treatment condition (3DOTS), the chatbot displayed a
graphical typing indicator before sending a message (see
Table 1). In the second treatment condition (TYPING), the
chatbot displayed a textual typing indicator before sending a
message. In each condition, the chatbots’ responses were
delayed by 2.3 seconds to ensure that participants were
sufficiently exposed to the typing indicators. Since response
time can also serve as a social cue (Gnewuch et al. 2018),
responses of the chatbot without typing indicators were also
delayed to keep the chatbots’ response time identical and
hence comparable across all conditions.
Condition
Typing
Indicator
Description
CTRL
None
-
3DOTS
Graphical
Three animated dots fading in one after
another and then fading out, placed
above the user input field.
TYPING
Textual
A textual status message “typing…
placed above the user input field.
Table 1. Treatment Configuration
Measures and Manipulation Check
All measures used in the survey were adapted from
established scales. Social presence was assessed using the
items from Gefen and Straub (1997). Moreover, we
measured control variables, such as disposition to trust
technology (Lankton et al. 2015) and prior experience with
chatbots (ranging from “never” to “daily”), as well as
collected demographic information (e.g., age, gender).
Condition
Social Presence
Mean
SD
SE
CTRL (n=63)
3.448
1.442
0.182
3DOTS (n=63)
3.902
1.451
0.183
TYPING (n=63)
3.663
1.408
0.177
SD = standard deviation | SE = standard error
Table 2. Descriptive Statistics
We included a manipulation check to test whether the typing
indicator manipulation was successful. Participants were
asked to rate whether the chatbot indicated that a response
was being prepared/generated, using a 7-point Likert scale
(1 = “strongly disagree”; 7 = “strongly agree). A one-way
ANOVA showed a significant influence of the experimental
conditions on perceived indication that a response was
prepared/generated (F(2, 186)=439.3, p<.001).
PRELIMINARY RESULTS
In our preliminary analysis conducted so far, we estimated
two regression models to analyze the effect of the treatment
conditions on perceived social presence and the interaction
between users’ prior experience with chatbots and the
treatment condition along with two user-related factors (i.e.,
gender, disposition to trust technology). We differentiated
between novice users who have never used a chatbot and
experienced users who use a chatbot at least 1-2 times per
year. Our preliminary analysis shows mixed results. For
novice users, we find that the existence of a graphical typing
indicator (i.e., three animated dots) significantly impacts
their perceived social presence of a chatbot. However, this
relationship is not significant for experienced users.
Contrary to our expectations, we did not observe a
significant effect for the textual typing indicators (i.e.,
“typing…”), neither for novice nor experienced users.
However, novice users perceived a chatbot with a graphical
typing indicator significantly more socially present than a
chatbot with a textual typing indicator. Our results indicate
that the relationship between typing indicators and user
perceptions of chatbots is more complex than assumed.
Moreover, individual user characteristics, such as
experience with chatbots, seem to play an important role in
this relationship.
CONCLUSION AND EXPECTED CONTRIBUTIONS
This paper provides first insights on the role of different
typing indicators in human-chatbot interaction. In particular,
our findings suggest that the relationship between typing
indicators and perceived social presence of chatbots depends
on the design of these indicators and user’s prior experience
with chatbots. Therefore, our findings contribute to existing
literature on the design and evaluation of text-based
conversational agents. In our future research, we intend to
continue our data analysis and plan to supplement our survey
responses with findings from a qualitative analysis of the
chatbot conversations.
REFERENCES
1. Appel, J., von der Pütten, A., Krämer, N. C., and Gratch,
J. 2012. “Does Humanity Matter? Analyzing the
Importance of Social Cues and Perceived Agency of a
Computer System for the Emergence of Social Reactions
during Human-Computer Interaction,” Advances in
Human-Computer Interaction (2012), pp. 110.
2. Auerbach, D. 2014. “I Built That ‘So-and-So Is Typing’
Feature in Chat.”
(http://www.slate.com/articles/technology/bitwise/2014/0
2/typing_indicator_in_chat_i_built_it_and_i_m_not_sorr
y.html, accessed August 8, 2018).
3. Cyr, D., Hassanein, K., Head, M., and Ivanov, A. 2007.
“The Role of Social Presence in Establishing Loyalty in
Gnewuch et al. Role of Typing Indicators in Human-Chatbot Interaction
Proceedings of the Sixteenth Annual Pre-ICIS Workshop on HCI Research in MIS, San Francisco, CA, December 13, 2018
5
E-Service Environments,” Interacting with Computers
(19:1), pp. 4356.
4. Dale, R. 2016. “The Return of the Chatbots,” Natural
Language Engineering (22:05), pp. 811817.
5. Fogg, B. J. 2002. “Computers as Persuasive Social
Actors,” in Persuasive Technology: Using Computers to
Change What We Think and Do, San Francisco, CA,
USA: Morgan Kaufmann Publishers, pp. 89120.
6. Gartner. 2018. “Gartner Says 25 Percent of Customer
Service Operations Will Use Virtual Customer Assistants
by 2020.
(https://www.gartner.com/newsroom/id/3858564,
accessed September 11, 2018).
7. Gefen, D., and Straub, D. W. 1997. “Gender Differences
in the Perception and Use of E-Mail: An Extension to the
Technology Acceptance Model,” MIS Quarterly (21:4),
pp. 389400.
8. Gefen, D., and Straub, D. W. 2004. “Consumer Trust in
B2C E-Commerce and the Importance of Social Presence:
Experiments in e-Products and e-Services, Omega
(32:6), pp. 407424.
9. Gnewuch, U., Morana, S., Adam, M. T. P., and Maedche,
A. 2018. “Faster Is Not Always Better: Understanding the
Effect of Dynamic Response Delays in Human-Chatbot
Interaction,” in Proceedings of the European Conference
on Information Systems (ECIS), Portsmouth, UK.
10. Gnewuch, U., Morana, S., and Maedche, A. 2017.
“Towards Designing Cooperative and Social
Conversational Agents for Customer Service,” in
Proceedings of the 38th International Conference on
Information Systems (ICIS), Seoul, South Korea.
11. Groom, V., Nass, C., Chen, T., Nielsen, A., Scarborough,
J. K., and Robles, E. 2009. “Evaluating the Effects of
Behavioral Realism in Embodied Agents,” International
Journal of Human-Computer Studies (67:10), pp. 842
849.
12. Klopfenstein, L. C., Delpriori, S., Malatini, S., and
Bogliolo, A. 2017. “The Rise of Bots: A Survey of
Conversational Interfaces, Patterns, and Paradigms,” in
Proceedings of the 2017 Conference on Designing
Interactive Systems, pp. 555565.
13. Lankton, N. K., Mcknight, D. H., and Tripp, J. 2015.
“Technology, Humanness, and Trust: Rethinking Trust in
Technology,” Journal of the Association for Information
Systems (16:10), pp. 880918.
14. Lee, Y., Chen, A., and Ilie, V. 2012. “Can Online Wait Be
Managed? The Effect of Filler Interfaces and Presentation
Modes on Perceived Waiting Time Online.,”
Management Information Systems Quarterly (36:2), pp.
365394.
15. Liedtke, T. 2017. “Facebook’s Perfect, Impossible
Chatbot,” MIT Technology Review.
(https://www.technologyreview.com/s/604117/facebooks
-perfect-impossible-chatbot/, accessed August 9, 2018).
16. Louwerse, M. M., Graesser, A. C., Lu, S., and Mitchell,
H. H. 2005. “Social Cues in Animated Conversational
Agents,” Applied Cognitive Psychology (19:6), pp. 112.
17. McTear, M. F. 2017.The Rise of the Conversational
Interface: A New Kid on the Block?,” in Future and
Emerging Trends in Language Technology. Machine
Learning and Big Data. FETLT 2016, pp. 3849.
18. Mori, M. 1970. “The Uncanny Valley,” Energy (7:4), pp.
3335.
19. Nass, C., Steuer, J., and Tauber, E. R. 1994. “Computers
Are Social Actors,” in Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems,
Boston, MA, USA, pp. 7278.
20. von der Pütten, A. M., Krämer, N. C., Gratch, J., and
Kang, S.-H. 2010. ““It Doesn’t Matter What You Are!”
Explaining Social Effects of Agents and Avatars,”
Computers in Human Behavior (26:6), pp. 16411650.
21. 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. 145182.
22. Sacks, H., Schegloff, E. A., and Jefferson, G. 1974. “A
Simplest Systematics for the Organization of Turn-Taking
for Conversation,” Language (50:4), p. 696.
23. Schuetzler, R. M., Grimes, G. M., Giboney, J. S., and
Buckman, J. 2014. “Facilitating Natural Conversational
Agent Interactions: Lessons from a Deception
Experiment,” in Proceedings of the 35th International
Conference on Information Systems (ICIS 14), Auckland,
NZ, pp. 116.
24. Shin, I.-G., Seok, J.-M., and Lim, Y.-K. 2018. “Too Close
and Crowded: Understanding Stress on Mobile Instant
Messengers Based on Proxemics,” in Proceedings of the
2018 CHI Conference on Human Factors in Computing
Systems (CHI ’18), Montreal, Canada, pp. 112.
25. Wakefield, R. L., Wakefield, K. L., Baker, J., and Wang,
L. C. 2011. “How Website Socialness Leads to Website
Use,European Journal of Information Systems (20:1),
pp. 118132.
26. Wiemann, J. M., and Knapp, M. L. 1975. Turn‐taking in
Conversations,” Journal of Communication (25:2), pp.
7592.
... Thus, social cues assigned to an agent trigger humans to apply social behaviors and rules towards the agent [11]. Such social cues could be a name, emotions [6] or also typing indicators [42]. The latter also addresses the concept of social presence [42], i.e. an agent is perceived as socially present, aware and conscious [32]. ...
... Such social cues could be a name, emotions [6] or also typing indicators [42]. The latter also addresses the concept of social presence [42], i.e. an agent is perceived as socially present, aware and conscious [32]. Still, next to the Social Response Theory, researchers also refer to the concept of the Uncanny Valley by Mori [23] reasoning the application of less social cues in order to match the human likeness with competence for maximum affinity [23,33,42]. ...
... The latter also addresses the concept of social presence [42], i.e. an agent is perceived as socially present, aware and conscious [32]. Still, next to the Social Response Theory, researchers also refer to the concept of the Uncanny Valley by Mori [23] reasoning the application of less social cues in order to match the human likeness with competence for maximum affinity [23,33,42]. As it is quite easy to generate a social relationship between humans and computers, it is recommended to make use of rudimentary but powerful cues instead of developing highly complex agents [11]. ...
Conference Paper
Full-text available
This paper aims at implementing a hybrid form of group work through the incorporation of an intelligent collaborative agent into a Collaborative Writing process. With that it contributes to the overall research gap establishing acceptance of AI towards complementary hybrid work. To approach this aim, we follow a Design Science Research process. We identify requirements for the agent to be considered a teammate based on expert interviews in the light of Social Response Theory and the concept of the Uncanny Valley. Next, we derive design principles for the implementation of an agent as teammate from the collected requirements. For the evaluation of the design principles and the human teammates' perception of the agent, we instantiate a Collaborative Writing process via a web-application incorporating the agent. The evaluation reveals the partly successful implementation of the developed design principles. Additionally, the results show the potential of hybrid collaboration teams accepting non-human teammates.
... If a chatbot interface can trigger the "social presence heuristic" instead of the "machine heuristic," it can have positive effects on interaction outcomes. Theory and research overwhelmingly support the idea that higher levels of perceived social presence can lead to better evaluation of the overall interaction (Venkatesh, Morris, and Ackerman 2000;Gnewuch et al. 2018) and higher intentions to purchase on an e-commerce website (Lu, Fan, and Zhou 2016). ...
... In the LOW and MED group, we decided to not delay the RACs' response time. Moreover, the RAC in the HIGH group provides a graphical indication that the message is prepared (i.e. during the dynamic delay), while LOW and MED provided no such graphical indication (Gnewuch et al. 2018b). In addition to these social cues addressing the visual appearance and chronometric behavior, we also decided to vary the way the RAC is texting. ...
Conference Paper
Full-text available
Robo-advisors promise to offer professional financial advice to private households at low cost. However, the automation of financial advisory processes to fully-digitalized services often goes hand in hand with a loss of human contact and lack of trust. Especially in an investment context, customers demand a human involved to ensure the trustworthiness of digitalized financial services. In this paper, we report the findings of a lab experiment (N=183) investigating how different levels of anthropomorphic design of a robo-advisor chatbot might compensate for the lack of human involvement and positively affect users’ trusting beliefs and likeliness to follow its recommendations. We found that the anthropomorphic design influences users’ perceived social presence of the chatbot. While trusting beliefs in the chatbot enhance users’ likeliness to follow the recommendation by the chatbot, we do not find evidence for a direct effect of social presence on likeliness to follow its recommendation. However, social presence has a positive indirect effect on likeliness to follow the recommendation via trusting beliefs. Our research contributes to the body of knowledge on the design of robo-advisors and provides more insight into the factors that influence users’ investment decisions when interacting with a robo-advisor chatbot.
... Diese Eingabeanzeigen, die dem Nutzer aus Messaging-Plattformen bekannt sind (bspw. drei animierte Punkte in Facebook Messenger), werden ebenfalls als ein soziales Signal wahrgenommen ( Gnewuch et al. 2018b). Die soziotechnische Gestaltung von Chatbots sollte daher auch die Abhängigkeiten zwischen verschiedenen sozialen Signalen berücksichtigen. ...
Chapter
Full-text available
Chatbots sind softwarebasierte Systeme, die mittels natürlicher Sprache mit Menschen interagieren. Viele Unternehmen setzen zunehmend auf Chatbots, um Kunden bei der Suche nach Informationen über Produkte oder Dienstleistungen und bei der Durchführung einfacher Prozesse zu unterstützen. Nichtsdestotrotz ist die Akzeptanz von Chatbots bei vielen Nutzern derzeit noch gering. Ein Grund dafür ist, dass sich die Interaktion mit Chatbots nur selten natürlich und menschlich anfühlt. Es wächst deshalb die Erkenntnis, dass neben einer guten technischen Plattform auch weitere Faktoren bei der Gestaltung von Chatbots beachtet werden sollten. Die soziotechnische Gestaltung von Chatbots fokussiert sich daher auf die sozialen Signale eines Chatbots. Diese sozialen Signale (z. B. Lächeln, Sprachstil oder Antwortgeschwindigkeit) spielen nicht nur in der zwischenmenschlichen Kommunikation, sondern auch in der Interaktion mit Chatbots eine grosse Rolle. Dieser Artikel erläutert die Grundlagen der soziotechnischen Gestaltung von Chatbots, verdeutlicht die Wirkung sozialer Signale anhand eines Forschungsbeispiels und diskutiert kritisch die Vermenschlichung von Chatbots.
... "a key design challenge to increase their adoption is to make the interaction with them feel natural and human-like" [GM18]. ...
Thesis
Full-text available
Goal of this research was to improve the user experience of chatbot systems by using tone analysis - emotion recognition in text. Therefore, the author investigated whether tone analysis utilised for a chatbot has a significant impact on the resulting UX. The idea was detecting the tone in a user utterance at certain steps of a conversation between chatbot and user. The detected tone was used to determine an appropriate response to the user. This way, the chatbot should appear as it would be able to express and apply empathy. The aforementioned approach was applied on a use case of comforting emotional support in cases of bullying at the workplace. A process has been developed to help people with their bullying situation. At certain steps in the conversation, user utterances were sent to a tone analyser service, which sent back the dominant tone determining the following response of the chatbot to the user. To investigate the research question, the chatbot Mia AI has been devel- oped, which provides comforting emotional support giving advice and aiming for relief. The chatbot system was tested through an A/B testing setting in which 16 participants interacted with two different versions of Mia AI. Chatbot A was developed with tone analysis; chatbot B was developed with- out tone analysis. After each interaction, users rated the chatbot through a BLRI and an AttrakDiff questionnaire assessing the empathy and the UX of the chatbot. After the activity, the data was statistically analysed. The results suggest that there was a significant difference between the empathy of chatbot A and chatbot B, as well as between the UX of chatbot A and chatbot B. Further, chatbot A received a higher rating in both, empathy and UX, than chatbot B. Based on the research results, tone analysis seemed to improve the em- pathy of the chatbot, and the empathy seemed to improve the UX of the chatbot, the researcher concludes that tone analysis improves the UX of chatbots. As developing a chatbot with tone analysis comes along with major efforts for creating dialogue strategies and conversation flows to handle each possible tone, the decision on whether to use tone analysis should be made weighing costs and benefits. Only use cases considered as emotion- ally charged, as the implemented use case, may benefit from tone analysis. For use cases without emotionally charged conversation steps, tone analysis possibly would make no significant difference while taking high costs into account. Developers should apply use-case-specific checking of applicability; in case their use case may include emotionally charged conversation steps, the empathy and UX of the chatbot may benefit significantly.
Conference Paper
Full-text available
Conversational agents (CAs)—software systems emulating conversations with humans through natural language—reshape our communication environment. As CAs have been widely used for applications requiring human-like interactions, a key goal in information systems (IS) research and practice is to be able to create CAs that exhibit a particular personality. However, existing research on CA personality is scattered across different fields and researchers and practitioners face difficulty in understanding the current state of the art on the design of CA personality. To address this gap, we systematically analyze existing studies and develop a framework on how to imbue CAs with personality cues and how to organize the underlying range of expressive variation regarding the Big Five personality traits. Our framework contributes to IS research by providing an overview of CA personality cues in verbal and non-verbal language and supports practitioners in designing CAs with a particular personality.
Conference Paper
Full-text available
The increasing application of Conversational Agents (CAs) is rapidly changing the way customers access services. In practice, however, CAs are frequently flawed, such as they are often misinterpreting users' requests. Even through future advancements in technology, it remains unlikely that a CA will be flawless and error-free. Thus, understanding how to design CAs so that users are more willing to overlook errors constitutes an important area of research. In this regard, our study positions the human-like design of CAs as a potential way to ensure service satisfaction when errors occur. We examined the effect of human-like designed flawed CAs by conducting a two-condition experiment with 421 participants, analyzing its effect on users' emotional state and service satisfaction. Our results show that the human-like design of flawed CAs leads to more positive emotions, which improve users' service satisfaction. Therefore, designers should adopt a human-like design to prevent dissatisfaction with flawed CAs.
Article
Full-text available
Zusammenfassung Immer mehr Unternehmen bieten ihren Kunden verschiedene Chatbots für unterschiedliche Geschäftsfälle an. Damit ein Kunde immer den passenden Chatbot für sein Anliegen findet, sollten die einzelne domänen-spezifischen, also auf bestimmte Geschäftsfelder spezialisierten Chatbots in korrekter Art und Weise miteinander interagieren können. Zudem sollte die Implementierung, Wartung sowie Koordination der verschiedenen Chatbots effizient bleiben. Ein Unternehmen muss folglich die geeigneten technischen Strukturen schaffen. Dieser Artikel stellt daher verschiedene Architekturmuster für Multi-Chatbot-Landschaften vor, diskutiert ihre Vor- und Nachteile und präsentiert eines davon, der Bot-Orchestrator, im Detail.
Article
Human beings express affinity (Shinwa‐kan in Japanese language) in communicating transactive engagements among healthcare providers, patients and healthcare robots. The appearance of healthcare robots and their language capabilities often feature characteristic and appropriate compassionate dialogical functions in human–robot interactions. Elements of healthcare robot configurations comprising its physiognomy and communication properties are founded on the positivist philosophical perspective of being the summation of composite parts, thereby mimicking human persons. This article reviews Mori's theory of the Uncanny Valley and its consequent debates, and examines "Uncanny" relations with generating healthcare robot conversational content with artificial affective communication (AAC) using natural language processing. With healthcare robots provoking influential physical composition and sensory expressions, the relations in human–healthcare robot transactive engagements are argued as supportive of the design and development in natural language processing. This implies that maintaining human–healthcare robot interaction and assessing the eeriness situations explained in the Uncanny Valley theory are crucial positions for healthcare robot functioning as a valuable commodity in health care. As such, physical features, language capabilities and mobility of healthcare robots establish the primacy of the AAC with natural language processing as integral to healthcare robot–human healthcare practice.
Article
Full-text available
Several domain-specific assistants in the form of chatbots have conquered many commercial and private areas. However, there is still a limited level of systematic knowledge of the distinctive characteristics of design ele- ments for chatbots to facilitate development, adoption, implementation, and further research. To close this gap, the paper outlines a taxonomy of design elements for chatbots with 17 dimensions organized into the perspectives intel- ligence, interaction and context. The conceptually groun- ded design elements of the taxonomy are used to analyze 103 chatbots from 23 different application domains. Through a clustering-based approach, five chatbot arche- types that currently exist for domain-specific chatbots are identified. The developed taxonomy provides a structure to differentiate and categorize domain-specific chatbots according to archetypal qualities that guide practitioners when taking design decisions. Moreover, the taxonomy serves academics as a foundation for conducting further research on chatbot design while integrating scientific and practical knowledge.
Conference Paper
Full-text available
A key challenge in designing conversational user interfaces is to make the conversation between the user and the system feel natural and human-like. In order to increase perceived humanness, many systems with conversational user interfaces (e.g., chatbots) use response delays to simu-late the time it would take humans to respond to a message. However, delayed responses may also negatively impact user satisfaction, particularly in situations where fast response times are expected, such as in customer service. This paper reports the findings of an online experiment in a customer service context that investigates how user perceptions differ when interacting with a chatbot that sends dynamically delayed responses compared to a chatbot that sends near-instant responses. The dynamic delay length was calculated based on the complexity of the re-sponse and complexity of the previous message. Our results indicate that dynamic response de-lays not only increase users’ perception of humanness and social presence, but also lead to greater satisfaction with the overall chatbot interaction. Building on social response theory, we provide evidence that a chatbot’s response time represents a social cue that triggers social re-sponses shaped by social expectations. Our findings support researchers and practitioners in understanding and designing more natural human-chatbot interactions.
Conference Paper
Full-text available
This study reports the results of a laboratory experiment exploring interactions between humans and a conversational agent. Using the ChatScript language, we created a chat bot that asked participants to describe a series of images. The two objectives of this study were (1) to analyze the impact of dynamic responses on participants' perceptions of the conversational agent, and (2) to explore behavioral changes in interactions with the chat bot (i.e. response latency and pauses) when participants engaged in deception. We discovered that a chat bot that provides adaptive responses based on the participant's input dramatically increases the perceived humanness and engagement of the conversational agent. Deceivers interacting with a dynamic chat bot exhibited consistent response latencies and pause lengths while deceivers with a static chat bot exhibited longer response latencies and pause lengths. These results give new insights on social interactions with computer agents during truthful and deceptive interactions.
Conference Paper
Full-text available
The idea of interacting with computers through natural language dates back to the 1960s, but recent technological advances have led to a renewed interest in conversational agents such as chatbots or digital assistants. In the customer service context, conversational agents promise to create a fast, convenient, and cost-effective channel for communicating with customers. Although numerous agents have been implemented in the past, most of them could not meet the expectations and disappeared. In this paper, we present our design science research project on how to design cooperative and social conversational agents to increase service quality in customer service. We discuss several issues that hinder the success of current conversational agents in customer service. Drawing on the cooperative principle of conversation and social response theory, we propose preliminary meta-requirements and design principles for cooperative and social conversational agents. Next, we will develop a prototype based on these design principles.
Conference Paper
Full-text available
This work documents the recent rise in popularity of messaging bots: chatterbot-like agents with simple, textual interfaces that allow users to access information, make use of services, or provide entertainment through online messaging platforms. Conversational interfaces have been often studied in their many facets, including natural language processing, artificial intelligence, human-computer interaction, and usability. In this work we analyze the recent trends in chatterbots and provide a survey of major messaging platforms, reviewing their support for bots and their distinguishing features. We then argue for what we call "Botplication", a bot interface paradigm that makes use of context, history, and structured conversation elements for input and output in order to provide a conversational user experience while overcoming the limitations of text-only interfaces.
Article
Full-text available
By all accounts, 2016 is the year of the chatbot. Some commentators take the view that chatbot technology will be so disruptive that it will eliminate the need for websites and apps. But chatbots have a long history. So what's new, and what's different this time? And is there an opportunity here to improve how our industry does technology transfer?
Article
Full-text available
Information systems (IS) research has demonstrated that humans can and do trust technology. The current trust in technology literature employs two different types of trust in technology constructs. Some researchers use human-like trust constructs (e.g., benevolence, integrity, and ability), while other researchers use system-like trust constructs (e.g., helpfulness, reliability, and functionality). Interestingly, past research shows that both sets of measures influence important dependent variables, but the literature does not explain when one type should be used instead of the other type. In this paper, we use trust, social presence, and affordance theories to shed light on this research problem. We report on two studies. In study 1, we argue first that technologies vary in their perceived “humanness”. Second, we argue that, because users perceive two technologies to differ in humanness, they will develop trust in each technology differently (i.e., along more human-like criteria or more system-like criteria). We study two technologies that vary in humanness to explore these differences theoretically and empirically. We demonstrate that, when the trust construct used aligns well with how human the technology is, it produces stronger effects on selected outcome variables than does a misaligned trust construct. In study 2, we assess whether these technologies differ in humanness based on social presence, social affordances, and affordances for sociality. We find that these factors do distinguish whether technology is more human-like or system-like. We provide implications for trust-in-technology research. © 2015, Association for Information Systems. All rights reserved.
Conference Paper
Nowadays, mobile instant messaging (MIM) is a necessity for our private and public lives, but it has also been the cause of stress. In South Korea, MIM stress has become a serious social problem. To understand this stress, we conducted four focus groups with 20 participants under MIM stress. We initially discovered that MIM stress relates to how people perceive the territory in MIM. We then applied proxemics-the theory of human use of space-to the thematic analysis as the rationale. The data revealed two main themes: too close and too crowded. The participants were stressed due to design features that let strangers or crowds into their MIM applications and forced them to interact and share their status with them. Based on this finding, we propose a set of implications for designing anti-stress MIM applications.
Conference Paper
The conversational interface has become a hot topic in the past year or so, providing the primary means of interaction with chatbots, messaging apps, and virtual personal assistants. Major tech companies have been making huge investments in the supporting technologies of artificial intelligence, such as deep learning and natural language processing, with the aim of creating systems that will enable users of smartphones and other devices to obtain information and access services in a natural, conversational way. Yet the vision of the conversational interface is not new, and indeed there is a history of research in dialogue systems, voice user interfaces, embodied conversational agents, and chatbots that goes back more than fifty years. This chapter explores what has changed to make the conversational interface particularly relevant today, examines some key issues from earlier work that could inform the next generation of conversational systems, and highlights some challenges for future work.
Article
Long waits online undermine users' evaluations of Web sites and their providers, triggering abandonment behaviors. Yet e-business researchers and practitioners have not perfected mechanisms to respond to online wait issues. A filler interface that runs during the wait for search results may influence online users' perceived waiting time (PWT); however, no scientific investigation has attempted to design effective filler interfaces for managing online waits. By adopting resource allocation theory, cognitive absorption theory, and human computer interaction (HCI) theories (competition for attention, visual search, and motion effect), we design diverse filler interfaces and investigate their effects on antecedents of PWT. The proposed research model considers cognitive absorption factors such as temporal dissociation, focused immersion, and heightened enjoyment as antecedents of PWT, which in turn triggers three outcomes: affective appraisals, cognitive appraisals, and Web site use intention. A multistage, multimethod approach is used to test the research hypotheses. In the first stage, we compare a filler interface condition with a no-filler interface condition, and find the superiority of a filler interface with respect to inducing focused immersion and temporal dissociation. In the second stage, we conduct two controlled experiments to examine whether filler interfaces with various designs (varying the presence and relevance of image, text, and image motion) distinctly influence antecedents of PWT and confirm their distinctive effects on focused immersion, temporal dissociation, and heightened enjoyment. In addition, by conducting a structural equation modeling analysis, we find that our research model explains 51 percent, 51 percent, 44 percent, and 45 percent of the variance in PWT, affective appraisals, cognitive appraisals, and Web site use intention respectively. Theoretical and practical implications of these findings are provided.