Truly Social Robots

Understanding Human-Robot Interaction from the Perspective

of Social Psychology

Daniel Ullrich

and Sarah Diefenbach

Institute of Informatics, LMU Munich, Amalienstraße 17, 80333 Munich, Germany

Department of Psychology, LMU Munich, Leopoldstraße 13, 80802 Munich, Germany

Keywords: Social Robots, Human-Robot Interaction, HCI Theory, Social Psychology, Perception, Trust, Design Factors.

Abstract: Human-robot interaction (HRI) and especially social robots play an increasing role within the field of human-

computer interaction (HCI). Social robots are robots specifically designed to interact with humans, and

already entered different domains such as healthcare, transportation, or care of the elderly. However, research

and design still lack a profound theoretical basis, considering their role as social beings, and the psychological

rules that apply to the interaction between humans and robots. The present paper underlines this claim by a

list of central research questions and areas of relevance, and a summary of first results of own and others'

research. Finally, we suggest a research agenda and dimensions for a framework for social robot interaction,

which truly accounts for their social nature and relevant theory from social psychology.

1 INTRODUCTION

Social robots play an increasing role within the field

of human-computer interaction (HCI). In contrast to

industrial robots in the context of industry 4.0, social

robots are specifically designed to interact with

humans. Nowadays, the most popular areas of

application are healthcare (for an overview, see

Beasley, 2012), transportation, retail, care of the

elderly (e.g., Paro Robots, 2016), housekeeping, or

robots taking the role of a social companion or pet-

substitute (e.g., Robyn Robotics, 2016). With

technological advancements, further domains will

surely follow, so that the domain of social robots is

about to become one of the most important in human-

robot interaction (HRI). In this context, psychological

questions such as how a robot is perceived, whether

we trust or distrust it, accept or reject it, are of central

relevance (Taipale et al., 2015).

Already in the 90s, Nass and colleagues (1994)

coined the "Computers-Are-Social-Actors" (CASA)

paradigm, suggesting that people apply social rules

during their interaction with computers, which

naturally gains even more relevance in the particular

domain of social robots. Nevertheless, current

research and developments focus too much on

technological borders and possibilities, but disregard

social and psychological factors. Though HRI

researchers generally acknowledged social robots as

an important application domain, including studies on

anthropomorphism in social contexts (e.g., Fussel et

al., 2008), or specific relations between robot

behaviour and human perceptions (e.g., Hoffman et

al., 2014; Mok, 2016), an integrated view of these

findings is still missing. Relatively little attention is

paid to the essential nature of social robots as "social

beings among us", and mechanisms of social

perception and related phenomena of social

psychology in sum. Altogether, it seems that the

domain misses a theoretical grounding and

framework that fully accounts for the social nature of

social robots.

The present paper wants to make a first step towards

a better understanding of underlying mechanisms

related to the perception and interaction with social

robots, and a stronger integration of psychological

knowledge into research and design. We follow an

interdisciplinary approach, utilizing a combination of

theory and methods from HCI and psychology, in

order to provide a basis for successful and human-

centred robot design. More specifically, we want to

stress a dedicated perspective that understands social

robots "as a species" and highlights the psychological

rules that apply to the interaction between humans

and robots. We depict central research questions and

Ullrich D. and Diefenbach S.

Truly Social Robots - Understanding Human-Robot Interaction from the Perspective of Social Psychology.

DOI: 10.5220/0006155900390045

In Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2017), pages 39-45

ISBN: 978-989-758-229-5

areas of relevance, summarize first results of own and

others' research, and present a research agenda for the

domain of social robots, that accounts for their social

nature and relevant theory from social psychology.

2 THE INTERMEDIATE

POSITION OF SOCIAL

ROBOTS

One specific characteristic of social robots is their

intermediate position between "usual" human-

computer interaction (HCI) and human,

interindividual interaction. When interacting with our

environment, our behaviour often relies on scripted

pattern. However, in the case of social robots, neither

general models of human-computer interaction nor

models about human interaction seem fully

transferrable. While being a piece of technology on

the one hand, their anthropomorphic shape, their

ability to speak and interaction capabilities suggest

robots to be “more” than technology and algorithms,

as also our own recent studies showed in impressing

ways (e.g., Männlein, 2016; Ullrich, 2017; Weber,

2016).

Thus, a central topic within the understanding of

social robots is that of projection and classification.

Humans classify objects with which they interact both

through bottom-up and top-down processes. The

former uses cues of the interaction artefact (it is small,

round, has digits 1-12 and watch hands) and the latter

prior learned knowledge (I know how a watch looks

like, I have seen it before). The particular feature of a

social robot is that it hold cues that qualifies for the

class “intelligent living being” and sometimes even

“human” (to an extent). Being perceived as a member

of such class results in specific user expectations and

behaviour (such as over-trust) which differs from

those of other classes of technology

A lack of understanding of the mechanisms affecting

our perception of intelligent technology can result in

flawed designs with yet unknown consequences. For

example, a less-than-ideal designed autopilot in the

automotive context lead to over-optimistic

expectations regarding its actual capabilities. The

driver developed over-trust in the system and used it

in situations that the system could not handle, which

ultimately resulted in a fatal accident. In sum, with

the current state of research, we seem to be

unprepared for this challenge and are not yet

exploiting the possibilities of social robots to full

potential.

In order to design and act responsibly, we need to

seriously acknowledge the social nature of social

robots and relations to the general mechanisms of

social psychology (e.g., responsibility attributions,

judgments and decision making in social contexts).

Only a thorough understanding of the social

perception and reactions towards robots can enable

adequate design decisions and exploit social robots to

full potential. Otherwise, unintended, sometimes

dramatic consequences, may occur, e.g., accidents

due to over-trust in "smart" technology as outlined

above. It needs a deeper exploration of the principles

that determine what people perceive and how they

behave when being confronted with social robots.

Naturally, such questions gain even more complexity

when we think of settings where more than one robot

is involved. To foresee peoples' reactions and

perceptions when being confronted with social

robots, a thorough understanding of social robots as a

"species" and the unfurling human-robot relation is

required. To translate such insights into adequate

design solutions, it needs insights about the

consequences of specific robot properties, and the

kind of reactions they afford.

With an interdisciplinary background in psychology

and computer science, our vision is to bridge

knowledge from the two areas towards a human-

centred design of social robots, with an emphasis on

the effects of social context. Both fields provide

manifold theories and insights of interaction with

technology or humans respectively that may be

fruitfully combined. For example, in the HCI context,

this may be general models of user experience and

evaluation of interactive technology (Law and van

Schaik, 2010) or approaches to model artificial

intelligence (e.g., Cohen and Feigenbaum, 2014). In

social psychology, theories about social roles, social

identity, group dynamics and attribution mechanisms

(Smith et al., 2014) could support the shaping of

social robot behaviour and task suitability.

3 CENTRAL RESEARCH

QUESTIONS AND FIRST

INSIGHTS INTO SOCIAL

ROBOT INTERACTION

The present section depicts exemplary research

questions and first insights from own and others'

studies, underlining the need for a stronger

integration of (social) psychological research within

the domain of social robots. After that, we extract

three general dimensions of interest in social robot

HUCAPP 2017 - International Conference on Human Computer Interaction Theory and Applications

interaction, forming a basis for future research and a

systematic link between design factors and

psychological consequences.

3.1 Personality: What Character Do

Humans Appreciate in One

Situation or the Other?

One central question within any social situation is the

perception of the others' character and consequences

for liking and reactions. This of course, also applies

to the interaction with social robots. HRI research

already showed that robot personality is relevant, that

subtle changes in a robot’s appearance can lead

differences in perceived robot personality, and further

effects on social aspects like trust, acceptance or

compliance (e.g., Goetz et al., 2003; Kim et al., 2008;

Salem et al., 2015; Walters et al., 2008). However, a

systematic view on these findings is still missing,

leaving unclear whether there is a general kind of

robot personality that promotes or diminishes liking

and acceptance. There actually are two common,

contradicting theories, with little empirical evidence

in HRI research for both of them. The first is

similarity attraction, i.e., a person chooses and prefers

to interact with other people/robots similar to them

(e.g., Byrne, 1971; Lee et al., 2006; Tapus et al.,

2008). The second is the complementary principle,

stating that a person is more attracted to people with

personality traits that are contrary to their own (e.g.,

Leary, 2004; Lee et al., 2006; Sullivan, 2013).

While previous research explored robot personality

and effects on liking as an isolated factor, also the task

context could be of relevance, and personality and

task context may interact with each other. Just like we

expect different behaviours/shades of personality

(e.g., encouraging, critical) from a friend between

different situations, we may also judge different robot

personalities as more or less appropriate from one

situation to the other. Thus, design recommendations

for robot personality may vary depending on the

specific area of application. An own study found first

evidence for this assumption (Männlein, 2016). We

explored effects of three different robot personalities

in four different usage scenarios. While in some

scenarios, a neutral, conservative personality was

preferred, in others, participants wanted a robot with

strong character, which could be a positive (nice,

friendly) or even a negative (stubborn, grumbling)

personality. As a general tendency, differences in

robot personality were more relevant in exploration-

oriented scenarios (e.g., a social robot as house mate)

and less relevant in goal-oriented scenarios (e.g., a

social robot selling a train ticket).

3.2 How Much Do We Rely on Robots'

Judgments - Compared to Human

Judgments?

As already outlined above, trust in robots' judgments

and capabilities is a central factor to foresee the

reactions towards robots and to design responsibly.

For a first exploration of the basis level of trust

towards robots (compared to humans) we ran a

replication of the famous Asch (1951) paradigm in

the context of social robots (Ullrich et al., 2017).

Asch explored peoples' reactions to majority opinions

on their own perceptions and judgments. The

experimental setting poses a simple task: Identifying

a line out of three lines that matches a reference line.

In the control condition, nearly all participants are

able to perform the task correctly and pick the right

line. Variations of social context then demonstrate the

influence of group opinions on individual

judgements. It showed that people begin to mistrust

their own perceptions when their social environments

comes to other "perceptions" then their own. If

surrounded by confederates, instructed to pick a

wrong line, people tend to adjust their judgments as

well and pick a wrong line then, even if their

perception probably tells them otherwise. However,

in another experimental condition, already one among

the many confederates who picks the right line could

induce positive encouragement and a trend towards

more correct judgments.

In our replication study, one of the confederates was

a social robot, who participated in the experiment as

well (see Figure 1). Participants entered their

judgments through a computer interface, and were

also displayed the (seeming) judgments of all other

participants, including the robots’. The general trend

of results was that the social impact of the robot on

individual judgments was even higher than that of the

other participants. Especially the effect of positive

encouragement was more pronounced than if a human

participant was the only one giving the correct

answer. This shows, the generally high level of trust

towards social robots, and, as in the present case, how

this effect may be used for positive encouragement

(e.g., in the field of therapy/rehabilitation. If the robot

believes in my skills, I will do the same, and the

robot's optimistic judgments may be even more

powerful than what the doctor says.) On the other

hand, it also hints at the high sensibility and

responsibility related to the design of social robots. If

there is such a high potential for trust in social robots,

it is essential that such trust is used in an adequate

way, and to avoid over-trust.

Truly Social Robots - Understanding Human-Robot Interaction from the Perspective of Social Psychology

Figure 1: Replication study of the Asch paradigm on

conformity and perception with a social robot as

participant. (A) shows the robots’ cubicle and custom user

interface, (b) the experimental situation and (c) the on

screen user interface for participants.

3.3 Responsibility: How Much

Accountability Do We Assign to

Robots – or Ourselves?

Closely related to the issue of trust and distrust is the

topic of responsibility, and how much accountability

humans assign to robots, compared to other humans.

Again, mechanisms from social psychology appear as

a helpful start to understand in which situations what

level of accountability is assigned. Though trusting

robots in general, human's attributions also reflect the

concern for self-protection, and making others'

accountable for mistakes. This effect has already been

demonstrated in the HCI domain in various fields, but

gains increasing importance in the domain of social

robots, where accountability attributions have severe

consequences for the following reactions towards

robots as social agents. For example, a study by Moon

(2003) in the field of consumer psychology explored

responsibility attributions in the context of computer

aided purchase decisions. In general, the results

reflect a self-serving bias, where consumers tend to

blame computers for negative outcomes but take

personal credit for positive ones. However, this effect

is also moderated by the personal history of self-

disclosure ´between human and computer. In a more

intimate relationship, consumers are more willing to

credit the computer for positive outcomes, and more

willing to accept responsibility for negative

outcomes. Such effects, of course, are also highly

relevant in the domain of social robots that even

provides more room for relationship building than

just "usual" human-computer interaction.

3.4 In- or Outgroup: What Makes

Robots Being One of Us? What Are

the General Dimensions of Social

Robot Perception?

Finally, central to all the matters about trust,

responsibility, and characterization, and the question

to what degree mechanisms of social interaction may

apply to the domain of social robot interaction,

appears the question about what makes robots being

one of us, and the general dimensions of social robot

perception. As outlined in the introduction section,

the interaction with social robots can be positioned

somewhere between normal human-human and

human-computer interaction. Subtle differences in

their design may decide about mechanisms of

projection and classification in one or the other

direction, and in consequence, the activated

psychological processes when entering the

interaction. To consider this in design, an important

prerequisite is to know the general dimensions along

which we classify a robot as social being or not, and

which design factors are relevant for the overall

perceived human-likeness.

In an experimental study (Weber, 2016) we explored

the relative impact on perceived human-likeness for

two central factors in social robot design, namely,

motion and speech. In our study, the role of the social

robot was applied in the sports context, more

specifically, being a karate teacher, giving

instructions for specific karate moves (see Figure 2).

Each factor (motion, speech) was realized in three

degrees of differing fidelity by help of a Nao robot,

and through systematic combination, the relative

impact of these factors was tested. Overall, speech

was found to be more relevant than motion for

perceived human-likeness, global impression, and

HUCAPP 2017 - International Conference on Human Computer Interaction Theory and Applications

general preference. Of course, this finding cannot be

generalized yet and further research with a wide range

of settings and other design factors and robot-types is

necessary. However, it already reveals the importance

of dedicated knowledge on the specific effects of

single design factors and their relative importance.

Such insights allow concentrating design efforts on

the most relevant parts and consequences from a

psychological perspective.

Figure 2: A social robot teaching karate moves.

4 OUTLOOK AND RESEARCH

AGENDA

As exemplified by the questions raised in the

preceding paragraphs, the overall aim of future

research around social robot interaction must be a

better understanding of the underlying psychological

mechanisms and exploration of its impact on robot

properties, design fundamentals, and dynamics in

social contexts. More specifically, our research

agenda suggests three fundamental directions. First, a

thorough exploration of psychological mechanisms

and dynamics of social interaction, through a series of

experiments, with varying independent (e.g.,

personality, anthropomorphism) and dependent

variables (e.g., trust, human-likeness, perceived will,

behaviour correlates of over-/under-trust). In our

experiments, we used a NAO robot as representative

for the class of social robots. Although our own

research as well as others’ show that a high fidelity

humanoid robot like Sophia (Hanson Robotics, 2016)

is not necessarily needed to evoke social effects (e.g.,

social presence, Hoffman et al, 2015), a broader

variation of fidelity within the same experimental

settings is preferable to explore the range of effects.

Second, a systematic exploration of the design space

and the relevance of single design factors for

perceptions, perceived character, trust, and

acceptance, with the goal to derive design pattern for

an intended robot experience in different scenarios,

areas of application, and contextual requirements

(e.g., security-related issues).

Third, an exploration of group dynamics in settings

with multiple social robots. As already noted above,

designing for social robot interaction gains even more

complexity in settings where more than one robot is

involved. This for example, is already the case in the

Japanese Henn na Hotel, where the human staff was

almost fully replaced by social robots, who are now

running the reception, doing cleaning services etc.

(see Figure 3). In order to foresee the emerging

dynamics in such settings, knowledge about the

special characteristics in multi-robot interaction is

crucial. This includes, for example to develop

paradigms for multi-robot-collaboration studies, and

to explore how findings from studies on single robot-

human interaction might change when robots

constitute the majority.

Figure 3: Social robots running the reception at the

Japanese Henn na hotel (www.h-n-h.jp).

Truly Social Robots - Understanding Human-Robot Interaction from the Perspective of Social Psychology

Finally, knowledge from all three research directions

must be synthesized in an integrative model on social

robots "as a species", providing an overview of

relevant mechanisms and variables of social robot

interaction, and their interrelations. Such knowledge

will then allow design recommendations for specific

domains and use cases.

5 CONCLUSION

As exemplified above, entering the domain of social

robots, means entering a domain that asks for other,

possible even more sensible and complex

considerations, than HCI design per se. While social

robots form a great potential to enrich our society,

profound knowledge about the peculiarities of their

species is needed, to bring them into our world with

best effect, and support a fruitful collaboration

between research and practice. We hope the present

considerations may help to outline the importance of

this endeavour, and that our studies will provide a

basis to create better, trusted, and accepted social

robots, in a way that positively contributes to human

(robot) society.

ACKNOWLEDGEMENTS

We thank Simon Männlein, Thomas Weber, and

Valentin Zieglmeier for their effort of planning and

conducting experiments, as their help was crucial for

our research.

REFERENCES

Asch, S. E., 1951. Effects of group pressure upon the

modification and distortion of judgments. Groups,

leadership, and men, 222-236.

Bartneck, C., Kulić, D., Croft, E., Zoghbi, S., 2009.

Measurement instruments for the anthropomorphism,

animacy, likeability, perceived intelligence, and

perceived safety of robots. International journal of

social robotics, 1(1), 71-81.

Beasley, R. A., 2012. Medical Robots: Current Systems and

Research Directions. Journal of Robotics 2012, 14.

Byrne, D. E., 1971. The attraction paradigm, volume 11.

Academic Press.

Cohen, P. R., Feigenbaum, E. A. (Eds.), 2014. The

handbook of artificial intelligence (Vol. 3).

Butterworth-Heinemann.

Fussell, S. R., Kiesler, S., Setlock, L. D., Yew, V., 2008.

How people anthropomorphize robots. In Proceedings

of the 3rd ACM/IEEE international conference on

Human robot interaction, 145-152. ACM.

Goetz, J., Kiesler, S., Powers, A., 2003. Matching robot

appearance and behavior to tasks to improve human-

robot cooperation. In The 12th IEEE International

Workshop on Robot and Human Interactive

Communication, 2003. Proceedings. ROMAN 2003,

55–60. IEEE.

Hanson Robotics, 2016. Retrieved November 30, 2016

from http://www.hansonrobotics.com/robot/sophia/

Hoffman, G., Birnbaum, G. E., Vanunu, K., Sass, O., Reis,

H. T., 2014. Robot responsiveness to human disclosure

affects social impression and appeal. In Proceedings of

the 2014 ACM/IEEE international conference on

Human-robot interaction, 1-8. ACM.

Hoffman, G., Forlizzi, J., Ayal, S., Steinfeld, A., Antanitis,

J., Hochman, G., ... & Finkenaur, J., 2015. Robot

presence and human honesty: Experimental evidence.

In Proceedings of the Tenth Annual ACM/IEEE

International Conference on Human-Robot Interaction,

181-188. ACM.

Kim, H., Kwak, S. S., Kim, M., 2008. Personality design of

sociable robots by control of gesture design factors. In

ROMAN 2008 - The 17th IEEE International

Symposium on Robot and Human Interactive

Communication, 494–499. IEEE.

Law, E. L. C., van Schaik, P., 2010. Modelling user

experience–An agenda for research and practice.

Interacting with computers, 22(5), 313-322.

Leary, T., 2004. Interpersonal diagnosis of personality: A

functional theory and methodology for personality

evaluation. Wipf and Stock Publishers.

Lee, K.M., Peng, W., Jin, S.-A., Yan, C., 2006. Can Robots

Manifest Personality?: An Empirical Test of

Personality Recognition, Social Responses, and Social

Presence in Human-Robot Interaction. Journal of

Communication, 56(4), 754–772.

Männlein, S., 2016. Exploring robot-personalities Design

and measurement of robot-personalities for different

areas of application. Master-Thesis, LMU Munich.

Mok, B., 2016. Effects of proactivity and expressivity on

collaboration with interactive robotic drawers. In 2016

11th ACM/IEEE International Conference on Human-

Robot Interaction (HRI),

633-634. IEEE.

Moon, Y., 2003. Don’t blame the computer: When self-

disclosure moderates the self-serving bias. Journal of

Consumer Psychology, 13(1), 125-137.

Nass C., Steuer, J., Tauber, E.R., 1994. Computers are

social actors. Computer Human Interaction (CHI)

Conference: Celebrating Interdependence 1994, 72–

78.

Paro Robots USA. 2014. PARO Therapeutic Robot.

Retrieved October 20, 2016 from

http://www.parorobots.com/

Robyn Robotics Ab. 2015. JustoCat. Retrieved October 20,

2016 from http://www.justocat.com/

Salem, M., Lakatos, G., Amirabdollahian, F., Dautenhahn,

K., 2015. Would You Trust a (Faulty) Robot?: Effects

of Error, Task Type and Personality on Human-Robot

Cooperation and Trust. Proceedings of the Tenth

HUCAPP 2017 - International Conference on Human Computer Interaction Theory and Applications

Annual ACM/IEEE International Conference on

Human-Robot Interaction, 141–148. ACM.

Smith, E. R., Mackie, D. M., Claypool, H. M., 2014. Social

psychology. Psychology Press.

Sullivan, H. S., 2013. The interpersonal theory of

psychiatry. Routledge.

Taipale, S., Luca, F. D., Sarrica, M., Fortunati, L., 2015.

Social Robots from a Human Perspective. In Springer

International Publishing.

Tapus, A., Tapus, C., Mataric, M. J., 2008. User—robot

personality matching and assistive robot behavior

adaptation for post-stroke rehabilitation therapy.

Intelligent Service Robotics, 1(2), 169–183.

Ullrich, D., 2017, accepted. Robot personality insights.

Designing suitable robot personalities for different

domains. I-com Journal of Interactive Media.

Ullrich, D., Butz, A., Diefenbach, S., 2017, under review.

More than just human. The psychology of trust in social

robots.

Weber, T., 2016. Show me your moves, Robot-sensei! The

influence of motion and speech on perceived human-

likeness of robotic teachers. Bachelor-Thesis, LMU

Munich.

Walters, W. L., Syrdal, D. S., Dautenhahn, K., Boekhorst,

R., Koay, K. L., 2008. Avoiding the uncanny valley:

Robot appearance, personality and consistency of

behavior in an attention-seeking home scenario for a

robot companion. Autonomous Robots, 24(2), 159–178.

Truly Social Robots - Understanding Human-Robot Interaction from the Perspective of Social Psychology