Social robots should be encouraged and incorporated into daily life

RESEARCH ARTICLE

Expectations vs. actual behavior of a social

robot: An experimental investigation of the

effects of a social robot’s interaction skill level

and its expected future role on people’s

evaluations

Aike C. HorstmannID*, Nicole C. Krämer

Social Psychology: Media and Communication, University of Duisburg-Essen, Duisburg, Germany

* [email protected]

Abstract

Since social robots are increasingly entering areas of people’s personal lives, it is crucial to

examine what affects people’s perceptions and evaluations of these robots. In this study,

three potential influences are examined: 1) the robot’s level of interaction skills, 2) the

robot’s expected future role as a helpful assistant or a threatening competitor, and 3) peo-

ple’s individual background with regard to robots and technology in general. In an experi-

mental lab study with a 2x2 between-subjects-design (N = 162), people read a vignette

describing the social robot Nao either as assistant or competitor and subsequently inter-

acted with Nao, which either displayed high or low interaction skills. Results of a structural

equation model show that the robot’s interaction skill level had the strongest effect, with a

low level leading to a negative evaluation of the robot’s sociability and competence and sub-

sequently a negative general evaluation of the interaction with the robot. A robot which was

expected to become a competitor was also evaluated as less sociable than a robot expected

to become an assistant. Overall, in case of an actual interaction with a social robot, the

robot’s behavior is more decisive for people’s evaluations of it than their expectations or indi-

vidual backgrounds.

Introduction

Since social robots are taking over roles which are traditionally filled by humans and thus are

becoming a growing part of our society, it is crucial to examine what affects people’s evaluation

and acceptance of these robots. How a robot is evaluated depends on several factors like the

robot’s appearance [1–4], its nonverbal behavior [3, 5], other behavioral aspects (e.g., predict-

ability: [6, 7] and cooperativeness: [8]) and media portrayals [9–12]. This study focuses on

three aspects: the robot’s behavior, the user’s expectation, and the user’s individual

background.

With regard to behavior, the robot’s level of interaction skills should affect people’s evalua-

tions tremendously for several reasons. First of all, behavior in general appears to play a major

PLOS ONE

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 1 / 18

a1111111111

a1111111111

a1111111111

a1111111111

a1111111111

OPEN ACCESS

Citation: Horstmann AC, Krämer NC (2020)

Expectations vs. actual behavior of a social robot:

An experimental investigation of the effects of a

social robot’s interaction skill level and its expected

future role on people’s evaluations. PLoS ONE

15(8): e0238133. https://doi.org/10.1371/journal.

pone.0238133

Editor: Chen Lv, Nanyang Technological University,

SINGAPORE

Received: May 14, 2020

Accepted: August 9, 2020

Published: August 21, 2020

Copyright: © 2020 Horstmann, Krämer. This is an open access article distributed under the terms of

the Creative Commons Attribution License, which

permits unrestricted use, distribution, and

reproduction in any medium, provided the original

author and source are credited.

Data Availability Statement: All relevant data are

within the manuscript and its Supporting

Information files.

Funding: The Open Access Publication Fund of the

University of Duisburg-Essen supported the

publication of the article. No additional external

funding was received for this study.

Competing interests: The authors have declared

that no competing interests exist.

role in interactions with non-human entities [13]. Furthermore, interaction skills are essential

for a social robot to fulfill its function [14] and according to Davis [15], a technology’s per-

ceived usefulness is linked to people’s attitude towards and usage intentions of this technology.

When a robot displays low interaction skills, this robot and the interaction with it should be

evaluated differently compared to when the robot displays high interaction skills.

With regard to expectations, most people have not personally interacted with a social robot

yet and thus have to draw information from other sources, which are often mass media reports

[11]. Media representations of “good” as well as “bad” fictional robot characters form people’s

attitude towards and expectations about social robots and lead to double-minded feelings [10].

Since expectations serve as perceptual filters of the reality [16, 17], whether people expect a

social robot to become a helpful assistant or a threatening competitor should affect how people

evaluate the robot and the interaction with it. Moreover, when the robot is expected to become

a competitor, high interaction skills should be perceived more threatening and thus more neg-

atively than low skills.

A preliminary analysis of a small part of the data already indicated that the robot’s level of

interaction skills as well as people’s expectation regarding the robot’s future role both have an

impact on people’s evaluation of this robot as well as on the evaluation of the interaction with

the robot t [18]. However, when examining human-robot interaction processes, people’s indi-

vidual background needs be taken into account as well. In this work, we additionally look at

people’s technical affinity [19], their locus of control when using technologies [20], their previ-

ous experiences with real or fictional robots and their negative expectancies regarding robots.

To get a more comprehensive picture, the focus of this work is to look at these different influ-

ences (behavior, expectation, and individual background) in one structural equation model to

be able to examine how they relate to each other while affecting how a robot and consequently

the interaction with it are evaluated. Summing up, this study examines the main and interac-

tion effects of a social robot’s interaction skill level, the robot’s expected future role and users’

personality variables to find out which aspects play what kind of role in people’s perceptions

and evaluations of a robot with which they just interacted.

A social robot’s level of interaction skills

How a social robot behaves while interacting with people, should play a central role for peo-

ple’s evaluation of this robot. One reason for this assumption is that behavior in general was

found to have a tremendous impact in interactions with non-human entities [13]. According

to Rickenberg and Reeves [13], the evaluation of a character “depends on what the character

does, what it says, and how it presents itself” (p. 55). Against this background, this study looks

at the effects of different behaviors of a social robot, particularly whether this robot displays

poor or sophisticated social interaction skills.

Since the media’s focus is often predominantly on the success and progress of a technology

and neglects problems and setbacks, this leads people to form exorbitantly high expectations

about social robots and their skills [11]. In this vein, robots are generally expected to be perfor-

mance-oriented, i.e., efficient, reliable, precise, rational, and perfectionist [21–23]. Likewise,

research by Kwon et al. [24] showed that people tend to generalize social capabilities for

humanoid robots, which also results in very high expectations for these robots. Looking at

social robots specifically, people expect them to be able to talk to them, to understand them

and to react to them in a sophisticated manner [11]. One of the main characteristics of socially

interactive robots is the ability to communicate with high-level dialogue [25]. Even though

there are several definitions of social robots, the ability to interact with its environment to

some degree (or at least to give the impression to do so) resembles the core definition of a

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 2 / 18

social robot [14]. Without adequate interaction skills, social robots would not be able to fulfill

their purpose of interacting with people in a social way and, thus, would not be very useful in

their common areas of application like home companionship, nursing care, entertainment,

and office/hotel assistance [14]. According to the findings obtained while developing the well-

established Technology Acceptance Model, perceived usefulness has a powerful effect on the

attitude towards using as well as on the actual use of a technology [15]. If a social robot has low

interaction skills, this should reduce its perceived usefulness and consequently lead to detri-

mental consequences for the evaluation of this robot. Research showed that when a robot’s

functions fall behind people’s expectations this leads to negative communication outcomes

like disappointment, mistrust, and rejection [26]. Thus, a robot displaying low levels of inter-

action skills should result in a negative evaluation of the robot’s sociability and competence.

How the robot is evaluated should then also affect the general evaluation of the interaction

with the robot. Against this background, the following hypotheses were postulated:

H1: A robot with a low level of interaction skills leads to a more negative evaluation of a) the

robot’s sociability and b) its competence, which consequently leads to c) a more negative

general evaluation of the interaction with the robot, compared to a robot with a high inter-

action skill level.

A social robot’s expected future role

In general, there are mainly two prominent prospects for social robots–one which is feared

and one which is desired. Since both images are promoted in media, this leads people to have

double-minded feelings towards robots [10].

On the one hand, there is this negative view on social robots becoming competitors. People

are worried about autonomous robots and loss of control, which is often accompanied by a

fear that humans will be either replaced or dominated by robots (known as “Frankenstein Syn-

drom”; [27–29]). Mass media, particularly science fiction formats, have a great influence on

people’s image of social robots and promote the idea of robots developing their own agenda

and revolting against humans [30]. In this vein, a study by Horstmann and Krämer [11]

showed that the more “bad” fictional robot characters people know, the more do they fear

robots to become superior and a threat to humans. To extend these survey-based findings, this

study aims to examine in a systematic-experimental way whether a negative expectation about

a robot affects how people evaluate this robot after they interacted with it. It is assumed that

interacting with a robot which is expected to learn from humans to become better and more

efficient than them and to take away tasks from them is undesirable and that this robot as well

as the interaction with it are evaluated poorly.

On the other hand, there is also this positive image of social robots functioning as assistants,

either in domestic, public, or work environments [11]. This idea of having an electronic help,

which makes life easier by carrying out tasks that are unpleasant or strenuous, is very appealing

to most people [11, 28, 31]. Thus, people should evaluate a robot which is expected to assist

humans with various tasks in the future positively and should also enjoy the interaction with

this robot.

In order to examine whether manipulating people’s expectations by framing the robot in a

negative light (by describing it as competitor working against humans) or in a positive light

(by describing it as assistant working for humans) influences how they evaluate the robot and

consequently the interaction with it, the following hypotheses were formulated:

H2: When a robot is expected to take over the role of a competitor, this leads to a more nega-

tive evaluation of a) the robot’s sociability and b) its competence, which consequently leads

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 3 / 18

to c) a more negative general evaluation of the interaction with the robot, compared to

when the robot is expected to become an assistant.

A social robot’s expected future role influencing the perception of its

interaction skill level

It is crucial for the acceptance and successful employment of social robots to find out which

influencing variables may interact with each other while affecting the evaluation of a robot. It

is known from interpersonal studies, that how a person’s behavior is interpreted is heavily

influenced by how desirable or rewarding it is to interact with the person [16]. It is a function

of all pre-interactional relationship and communicator characteristics (e.g., personality, repu-

tation, and nature of the relationship) and all interactional behaviors (e.g., an amusing com-

munication style or positive feedback; [32]). Transferring this to human-robot interaction,

how a robot’s behavior is perceived should also be affected by how desirable it is for people to

interact with this robot. Since there are ambiguous views on social robots, whether the robot is

expected to become a threatening competitor or a helpful assistant in the future should affect

how the robot’s level of interaction skills are perceived.

More specifically, a social robot’s skills might be perceived differently depending on how

people think this robot will use these skills in the future. In case of a robot which is expected to

become a competitor working against humans, high interaction skills could be perceived as

daunting. In science fiction, robots are often portrayed as highly skilled and extremely intelli-

gent, often surpassing humans and threatening to obtain world domination [27, 33–35].

Research has shown that these are scenarios that people are also afraid of in real life [28, 29],

especially when they are able to recall a lot of these negative fictional robot characters [11].

Thus, in case of a threatening competitor robot, people would most likely prefer this robot to

have low interaction skills. However, if a robot is expected to be a helpful assistant working for

humans, people would probably like to have this robot to be equipped with sophisticated inter-

action skills. Against this background, this study aims to examine how different levels of a

social robot’s interaction skills affect people’s evaluation of the robot and subsequently the

interaction with it when the robot is framed as undesirable competitor compared to when it is

framed as a desirable assistant. The following is hypothesized:

H3: When a robot is described as competitor, a high level of interaction skills leads to a more

negative evaluation of a) the robot’s sociability and b) its competence, which consequently

leads to c) a more negative general evaluation of the interaction with the robot.

Users’ individual background

In addition to the robot’s behavior and what future role people expect this robot to have, the

individual background of the person interacting with the robot should have a great influence

on how the robot and the interaction with it are evaluated. First of all, people’s technological

background should be taken into account. On the one hand, there is technical affinity, which

describes a person’s positive attitude, excitement, and trust toward technology [19]. A high

technical affinity resembles a high enthusiasm about new technologies [19], which should also

lead to a very positive attitude towards robots in general. On the other hand, there is locus of

control when using technology, which describes whether a person feels capable and confident

or rather helpless and overwhelmed when handling technological devices [20, 36]. How com-

petent a person feels handling technological devices usually relates to his or her actual techno-

logical competence and knowledge [20]. Thus, a high technological locus of control should

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 4 / 18

lead to a more realistic view on what social robots are capable of, as it appeared to be the case

in previous research [11].

Moreover, a person’s experiences with real robots but also with robots portrayed in mass

media (real and fictional ones) should be considered. Most people have never interacted with a

real robot before and thus a heavily used source of information for this still rather unfamiliar

technology are media reports about real robots and, even more prevalent, science fiction repre-

sentations (cf. [11]). Particularly since these portrayals often lack accuracy and elicit a biased

picture of robots, their influences need to be considered.

Since negative expectations and fears regarding robots are often discussed in human-robot

interaction research (cf. [9, 28, 37, 38]), as well as in literature (cf. [27, 33–35]), their influence

should be taken into account as well. Especially the fear of robots gaining consciousness and

becoming more intelligent and thus superior to humans represents a common science fiction

scenario [30]. Since science fiction is widely spread and easily accessible for everyone, they are

believed to have a great impact on people’s attitude towards social robots (e.g., [10, 11, 30]).

Thus, the influence of people’s negative expectancies regarding social robots to become supe-

rior to and dominate humans is also considered in this study.

In conclusion the following hypotheses are postulated:

H4: In addition to the main and interaction effects of the robot’s level of interaction skills and

the robot’s expected future role, people’s technological background, their previous experi-

ences with robots, and their negative expectancies regarding robots influence people’s eval-

uation of a) the robot’s sociability and b) its competence, which consequently affects c) the

general evaluation of the interaction with the robot.

Material and methods

The laboratory study employed an experimental 2 (assistant vs. competitor expectation) x 2 (high

vs. low level of interaction skills) between-subjects design. Participants were randomly assigned to

the conditions. The ethics committee of the division of Computer Science and Applied Cognitive

Sciences at the Faculty of Engineering of the University of Duisburg-Essen approved the study

and written informed consent was obtained. The individual pictured in this manuscript has given

written informed consent (as outlined in PLOS consent form) to publish these case details.

Sample

Results of an a priori power analysis using G�power 3.1 software (based on 95% power and a

medium effect size of f 2

= 0.15; [39, 40]) recommended a sample size of 111 participants. Fur-

thermore, our goal was to have at least 40 participants in each condition, resulting in a mini-

mum size of 160 participants. In the end, 189 people participated in the lab study, of which 25

had to be excluded because they did not pass the manipulation check making sure people read

the vignettes with sufficient attention (see next section). Additionally, one person was removed

because of insufficient language skills and one because the vignette was accidently read after

instead of before the interaction. Of the remaining 162 participants, 97 reported to be female

and 65 to be male. The age ranged from 18 to 42 years with an average of 22.85 years

(SD = 3.88). With regard to education, most participants were university students (94.4%) and

either possessed university entrance-level qualifications (59.3%) or a university degree (38.9%).

Experimental setting and procedure

First, a cover story was presented explaining that the participants are invited to the lab for a

final evaluation of the robot’s improved interaction skills. After written consent was obtained

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 5 / 18

from the participants, they were asked to read a vignette, where in the first part the robot Nao’s

(SoftBanks Robotics) interaction skills were described as very elaborated. In the second part of

the vignette, the robot was either described as competitor (Nao’s interaction and communica-

tion skills will be superior to human skills and it will take over many tasks which are currently

executed by humans, because it will be able to do them in a more efficient, reliable, and safe

way) or as assistant (Nao will be very helpful and assist humans with many exhausting and

onerous tasks, to make them easier and more pleasant to do). This was followed by a manipu-

lation check asking about the robot Nao’s interaction skills (possible answers: high, low, or

medium) and its future role (possible answers: assistant, replacement, or no information on

future application) as it was described in the vignette. Since only participants were included

who passed this check, we are confident that we successfully manipulated people’s expectations

with regard to the robot’s future role.

For the subsequent interaction, participants sat in front of the robot Nao, which was placed

in the middle of a table (see Fig 1). After the main functions of the robot were explained, the

experimenter gave the robot a start signal and left the room. The robot then explained that it

needed a moment for preparation and will notify the participant when it was ready to start.

This gave the experimenter enough time to enter the adjacent room from which the robot was

then controlled during the interaction. By using a webcam installed in the lab, the experi-

menter was able to see and hear the participant and to let the robot react in accordance with

the participant’s answers (wizard of oz design; see [41]). To explain why the participants

would be alone with the robot during the interaction, they were told that the robot produces a

lot of data for the final evaluation which can only be processed by a high-performance com-

puter located a few rooms down the hall. The experimenter explained that she wanted to make

sure that the data will be transferred and saved correctly, which is why she would like to

observe these processes during the interaction.

The webcam was justified by explaining that in case of error messages in the data, the devel-

opers of Nao’s interaction skills could use the videos to see what went wrong. Participants

were also told that the robot would guide them through the interaction and that they should

ring a bell when both interaction tasks were completed, which would notify the experimenter.

The first interaction task was a figure guessing game. Participants received a list with the

names and small pictures of 32 different figures (celebrities as well as comic and movie

Fig 1. The experimental set-up.

https://doi.org/10.1371/journal.pone.0238133.g001

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 6 / 18

characters) and were instructed to choose one of them without telling the robot which one

they chose. The robot then asked different questions (e.g., “Is it a real person or a fictional fig-

ure from a comic or movie?”) to eventually guess the figure. A decision tree was constructed so

that the robot was able to name the chosen figure after five questions. After three rounds of

this game, the robot introduced the second interaction task: creating a profile about each

other. Here, the robot and the participant took turns asking each other personal questions to

fill out a given profile template (e.g., age, height, hobbies, and favorite movie).

In the high interaction skills condition, participants were able to communicate with the

robot in a rather natural and sophisticated way. In the low interaction skills condition, the

communication was more restricted (e.g., robot can only understand “Yes” or “No”), the robot

deliberately applied some false grammar (e.g., false conjugation of verbs) and misunderstood

the participants’ answers several times. After the two interaction tasks were completed, the

experimenter came back to the lab and asked the participant to continue with questionnaires

on the computer. At the end of the experiment, participants were debriefed and their time and

effort were compensated with money or course credits.

Questionnaires

Evaluation of the robot’s sociability and competence. To assess the robot’s perceived

sociability, the Social Attractiveness subscale (Interpersonal Attraction Scale [42]; Likert scale ranging from 1 = “strongly disagree” to 5 = “strongly agree”; 5 items; e.g., “I think the robot

Nao could be a friend of mine.”; α = 0.80) as well as the Sociability subscale (Source Credibility Scale [43]; semantic differential scale ranging from 1 to 5; 6 items; e.g., “unpleasant–pleasant”;

α = 0.84) were adapted to ask about the robot Nao instead of a person. Additionally, the two subscales Perceived Enjoyment (5 items; e.g., “I enjoy the robot talking to me.”; α = 0.87) and Perceived Sociability (4 items; e.g., “I find the robot pleasant to interact with.”; α = 0.84) of the Acceptance of a Social Robot Scale ([44]; 1 = “strongly disagree” to 5 = “strongly agree”) were

employed as well.

People’s evaluation of the competence of the robot Nao was assessed with the Task Attrac- tiveness subscale (Interpersonal Attraction Scale [42]; Likert scale ranging from 1 = “strongly

disagree” to 5 = “strongly agree”; 5 items; e.g., “The robot Nao would be a poor problem

solver.”; α = 0.81) and the Competence subscale (Source Credibility Scale [43]; semantic differ- ential scale ranging from 1 to 5; 6 items; e.g., “unintelligent–intelligent”; α = 0.89), after both scales’ items were adjusted to an interaction with a robot instead of with another person.

Evaluation of the interaction with the robot. To assess how people generally evaluate the interaction with the robot, adapted versions of the Evaluation subscale ([45]; 4 items; e.g., “I was enjoying the interaction with the robot Nao.”; 1 = „strongly disagree”to 5 = „strongly

agree“; α = 0.81) and the Overall Rewardingness scale ([46]; 4 items; e.g., “The opportunity to interact with the robot Nao again is very desirable.”; 1 = „strongly disagree”to 5 = „strongly

agree“; α = 0.85) were used. People’s individual background. The participants’ technological background was mea-

sured with the Locus of Control when Using Technology Scale (KUT [20]; 8 items; e.g., “I feel

so helpless regarding technical devices that I rather keep my hands off of them.”; 1 = “strongly

disagree” to 5 = “strongly agree”; α = 0.85) and the Technical Affinity as Handling of and Atti- tude toward Electronic Devices Scale (TA-EG [19]; 19 items; e.g., “I enjoy trying an electronic

device.”; 1 = “does not apply at all” to 5 = “applies completely”; α = 0.83). Their experiences with robots were measured by asking participants how often (0 = “never”;

1 = “very rarely” to 5 = “very often”) they have had contact with real robots before. In a similar

way, they were asked how often they have watched reports or something similar about real

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 7 / 18

robots and how often they have watched fictional movies or series where robots played an

important role before.

To assess participants’ negative expectancies regarding robots, 12 items asking about com- mon science fiction scenarios were employed ([11]; e.g., “Robots will try to free themselves

from humans.”; 1 = “strongly disagree” to 5 = “strongly agree”; α = 0.88). Demographical background. Participants’ age, sex, educational level, and current

employment or training status were assessed.

Further questionnaires assessed, but not analyzed for this paper. In addition, the Char-

acter subscale of the Source Credibility Scale [43], the Anxiety subscale of the Acceptance of a

Social Robot Scale [44], Themes of Relational Communication: Trust, Dominance, and Equal-

ity [47], Contact Intentions [6], the General Anxiety subscale of the Frankenstein Syndrome

Questionnaire [48] and Knowledge of Fictional Robot Characters [11] were assessed, but not

used for the analyses of this paper.

Results

For an overview of the descriptive values of the main influencing and dependent variables see

Table 1.

To test all hypotheses (H1 to H4), a structural equation model (SEM) was employed based

on the theoretical deliberations outlined before. The process of structural equation modelling

comprises several statistical techniques which enable analyses of the relationships between one

or more, either continuous or discrete, independent variables and one or more, either continu-

ous or discrete, dependent variables [49]. Structural equation modelling basically combines

two multivariate techniques: multiple regression analysis and factor analysis [50]. On the one

hand, it is possible to simultaneously estimate multiple dependence relationships akin to mul-

tiple regression equations and on the other hand, various measures can be incorporated for

each concept similar to factor analysis [50]. A SEM has three major advantages over traditional

multivariate techniques: measurement errors are explicitly assessed, latent (unobserved) vari-

ables can be estimated via manifest (observed) variables and a model can be tested to examine

whether a conceptual or theoretical structure fits the data [51]. The aim of the current study is

to examine the relative influences of a social robot’s interaction skills (H1), the robot’s ex-

pected future role (H2), a combination of the robot’s interaction skills and its expected future

role (H3) and people’s individual background (H4) on a) the robot’s perceived competence

Table 1. Descriptive statistics of the main influencing and dependent variables.

Robot’s behavior Robot’s expected future role

High skills Lows skills Assistant

expectation

Competitor

expectation

Total

M SD M SD M SD M SD M SD Perceived competence Competence (SCS) 4.03 0.66 3.38 0.80 3.78 0.82 3.63 0.78 3.71 0.80

Task Attractiveness (IAS) 4.11 0.61 3.53 0.72 3.86 0.68 3.78 0.78 3.82 0.73

Perceived sociability Sociability (SCS) 4.32 0.53 3.97 0.65 4.23 0.55 4.06 0.67 4.14 0.61

Social Attractiveness (IAS) 3.24 0.91 2.89 0.93 3.21 0.91 2.93 0.95 3.07 0.94

Perceived Sociability (ASRS) 3.67 0.82 3.17 0.89 3.58 0.79 3.27 0.96 3.03 0.96

Perceived Enjoyment (ASRS) 4.16 0.68 3.74 0.92 4.11 0.77 3.79 0.86 3.95 0.83

Interaction evaluation Evaluation 4.23 0.58 3.61 0.87 4.08 0.72 3.77 0.85 3.92 0.80

Overall Rewardingness 3.95 0.73 3.32 0.89 3.73 0.83 3.54 0.89 3.63 0.87

SCS = Source Credibility Scale [43]; IAS = Interpersonal Attraction Scale [42]; ASRS = Acceptance of a Social Robot Scale [44].

https://doi.org/10.1371/journal.pone.0238133.t001

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 8 / 18

and b) its perceived sociability and how these subsequently affect c) the general evaluation of

the interaction with the robot.

For this SEM, perceived sociability, perceived competence and interaction evaluation as well as technological background and experiences with robots served as latent variables. Interaction skill level, expected future role, skill level x expected role, and negative expectancies regarding

robots were included as manifest variables (see Fig 2).

The SEM analysis was computed using IBM SPSS AMOS 25.0 for Windows (IBM SPSS sta-

tistics, released 2017). The latent variable perceived sociability is made up of the four manifest variables sociability (Source Credibility Scale [43]), social attractiveness (Interpersonal Attrac-

tion Scale [42]), perceived enjoyment, and perceived sociability (Acceptance of a Social Robot

Scale [44]). The latent variable perceived competence consists of the two manifest variables competence (Source Credibility Scale [43]) and task attractiveness (Interpersonal Attraction

Scale [42]). The latent variable interaction evaluation consists of the two manifest variables evaluation and overall rewardingness. The latent variable technological background consists of the two manifest variables locus of control when using technology and technical affinity, while

the latent variable experiences with robots consists of the three manifest variables contact with real robots, reception of reports about real robots, and reception of robot science fiction mov-

ies/series. The standardized model results for the latent dimensions can be found in Table 2.

There were no missing data and variables were tested for multivariate normality and multi-

collinearity. For the evaluation of the model fit, standard criteria were applied following the

suggestions from Hu and Bentler [52, 53]: The root mean square error of approximation

(RMSEA; values below 0.08 indicate an acceptable fit), comparative fit indices (CFI/TLI; values

above 0.90 indicate a good fit), and the standardized root mean square residual (SRMR; values

Fig 2. Structural equation model for the evaluation of the robot’s perceived competence and sociability as well as the general evaluation of the interaction.

https://doi.org/10.1371/journal.pone.0238133.g002

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 9 / 18

below 0.08 indicate a good fit with the data). For this model, the RMSEA was 0.05, CFI was

0.96, TLI was 0.95, and the SRMR was 0.06, indicating an overall good model fit.

Influence of the robot’s level of interaction skills

There was a significant direct effect of the robot’s interaction skill level on participants’ evalua-

tion of the robot’s sociability (β = .54, SE = 0.08, p = .001) as well as competence (β = .28, SE = 0.08, p = .005), which confirms the assumptions of H1a and H1b. There was a significant indirect effect of the robot’s level of interaction skills on people’s general evaluation of the

interaction with the robot via perceived sociability and competence (β = .41, SE = 0.07, p = .001), in support of H1c.

In summary, the robot’s behavior in terms of a high or low interaction skill level had a sig-

nificant effect on participants’ evaluation of the robot’s sociability and competence, which also

significantly affected people’s general evaluation of the interaction with the robot. The robot

was evaluated as less sociable and less competent when it displayed a low level of interaction

skills (see Table 1 for the descriptive values). Low interaction skills were also found to indi-

rectly lead to a less positive general evaluation of the interaction with the robot (see Table 1).

Influence of the robot’s expected future role

The expectation of the robot’s future role had a significant direct effect on the evaluation of the

robot’s sociability (β = .20, SE = 0.08, p = .011), but not on the evaluation of its competence (β = .12, SE = 0.08, p = .173), which is in accordance with H2a but not with H2b. There was also only a marginally significant indirect effect of the expectation of the robot’s future role on the

interaction evaluation via perceived sociability and competence (β = .15, SE = 0.07, p = .067). Therefore, H2c is not fully supported.

To sum up, how people evaluated the robot’s sociability was influenced by their expectation

regarding the robot’s future role. When the robot was described to aim to become a threaten-

ing competitor in the future, this robot was evaluated less sociable than when it was described

aiming to become a helpful assistant (see Table 1). The robot’s expected future role did not sig-

nificantly affect how competent this robot was perceived. On a marginal significant level, peo-

ple’s expectation regarding the robot’s future role indirectly affected their general evaluation of

the interaction with the robot. When the robot was expected to become a competitor, the

interaction with it was evaluated less positive (see Table 1).

Table 2. Coefficients of the manifest variables’ loadings on the latent dimensions.

Latent dimension Manifest variables β SE p Perceived competence Competence (SC) 0.65 0.07 .008

Task Attractiveness 0.67 0.06 .015

Perceived sociability Sociability (SC) 0.67 0.05 .002

Social Attractiveness 0.72 0.05 .001

Perceived Sociability 0.87 0.02 .001

Perceived Enjoyment 0.90 0.02 .004

Interaction evaluation Evaluation 0.92 0.02 .002

Overall Rewardingness 0.91 0.02 .005

Technological background Technological Locus of Control 0.56 0.11 .002

Technical Affinity 1.05 0.21 .002

Experiences with robots Contact with real robots 0.48 0.09 .003

Reception of robot reports 0.73 0.09 .004

Reception of science fiction 0.70 0.08 .002

https://doi.org/10.1371/journal.pone.0238133.t002

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 10 / 18

Combined influence of the robot’s level of interaction skills and its

expected future role

There was no significant direct effect of the interaction variable (robot’s skill level x robot’s

expected future role) on the robot’s perceived sociability (β = .03, SE = 0.08, p = .665) or com- petence (β = -.02, SE = 0.08, p = .807), which leads to a rejection of H3a as well as H3b. With regard to the combined influence of the robot’s expected future role and its displayed level of

interaction skills, there was also no significant indirect effect on the general evaluation of the

interaction (β = .00, SE = 0.08, p = .980), which leads to the rejection of H3c. Summing up, there was no significant direct interaction effect of the robot’s level of interac-

tion skills and the robot’s expected future role on the robot’s perceived sociability and its com-

petence, as well as no significant indirect interaction effect on the general evaluation of the

interaction with the robot.

Influence of people’s individual background

With regard to individual backgrounds, participants’ technological background (β = -.30, SE = 0.12, p = .001) as well as their experiences with robots (β = .37, SE = 0.13, p = .001) had a sig- nificant direct effect on negative expectancies regarding robots. However, neither negative expec-

tancies, nor technological background, nor experiences with robots had a significant direct effect

on the evaluation of the robot’s sociability (negative expectancies: β = -.06, SE = 0.08, p = .426; technological background: β = .14, SE = 0.10, p = .108; experiences with robots: β = .03, SE = 0.13, p = .808) or competence (negative expectancies: β = -.10, SE = 0.09, p = .235; technological back- ground: β = .11, SE = 0.11, p = .155; experiences with robots: β = -.07, SE = 0.14, p = .503). Like- wise, there were also no indirect effects (via negative expectancies) of technological background

and experiences with robots, neither on the robot’s perceived sociability (technological back-

ground: β = .02, SE = 0.03, p = .292; experiences with robots: β = -.02, SE = 0.04, p = .304) nor on its perceived competence (technological background: β = .03, SE = 0.04, p = .141; experiences with robots: β = -.04, SE = 0.04, p = .153). This leads to the rejection of the hypotheses H4a and H4b.

With regard to people’s evaluation of the interaction, experiences with robots (β = -.05, SE = 0.11, p = .615) as well as negative expectancies (β = -.08, SE = 0.08, p = .324) had no signif- icant indirect effect on the general evaluation of the interaction. There was a marginally signifi-

cant indirect effect of people’s technological background on the general evaluation of the

interaction with the robot (β = .15, SE = 0.09, p = .053). Therefore, H4c is not fully supported. In summary, none of the personality variables technological background, experiences with

robots, or negative expectancies regarding robots had a significant direct effect on the evalua-

tion of the robot’s sociability or its competence. There was also no significant indirect effect of

people’s experiences with robots and their negative expectancies regarding robots on the gen-

eral evaluation of the interaction with the robot. However, there was a marginally significant

indirect effect of people’s technological background revealing that people with a higher affinity

and competence with regard to technological devices appear to indirectly evaluate the interac-

tion with the robot more positively.

Post-hoc statistical power analysis

Post hoc power analyses using G�Power 3.1 were computed with the actual sample size of 162

participants. The statistical power for this study was 0.24 for detecting a small effect (f 2

= .02)

and exceeded 0.99 for a medium effect (f 2

= .15) as well as for a large effect size (f 2

= .35;

Cohen, 1988). Thus, more than adequate statistical power was reached for medium to large

effect sizes, but less than adequate power for a small effect size.

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 11 / 18

Discussion

Since social robots are increasingly entering various areas of people’s daily lives, it is crucial to

examine what affects people’s perceptions and evaluations of these robots and their interac-

tions with them. Ultimately this will affect people’s acceptance of social robots and their will-

ingness to use this technology [15]. For this study, our goal was to examine what influences

people’s evaluations of specific attributes of a robot with which they just interacted, but also

what subsequently influences people’s more general evaluation of the interaction with the

robot. For this purpose we decided to look at three aspects which we assumed to have substan-

tial influence: 1) the robot’s displayed level of interactions skills as a behavioral component, 2)

what kind of role this robot is expected to have in the future (helpful assistant or threatening

competitor), and 3) people’s individual background with regard to technology, experiences

with robots, and negative expectancies regarding robots. After setting people’s expectation

regarding the robot Nao to either become a threatening competitor or a helpful assistant in the

future, they interacted with Nao, which then either displayed a low or high interaction skill

level. That people had a personal encounter with a real robot in this study extends insights of

studies which are based on surveys only and allows a more realistic situation to examine peo-

ple’s evaluation of a robot’s characteristics and the interaction with this robot.

The robot’s level of interaction skills

With regard to the robot’s behavior during the interaction, Rickenberg and Reeves [13] were

among the first to emphasize the strong effects that the behavior of a non-human interaction

partner has on people. They further explain that these effects “are not unilaterally good or bad;

they can be either or both” and that “an animated character turns up the volume on social

presence, which means that it can accentuate the effects of everything presented” (p. 55). In

other words, the behavior of an interaction partner is of crucial importance when it comes to

the perception and evaluation of this interaction partner. In accordance with these findings,

the results of this study show that when a robot presents low interaction skills during people’s

interaction with it, the robot is evaluated as less competent and less sociable compared to

when the robot displays high interaction skills. Furthermore, these negative evaluations lead

people to also generally evaluate the interaction with the robot less positively. The robot with

low interaction skills was probably harder and less comfortable to interact with, which was

accentuated by its displayed behavior during the interaction. Another considerable explana-

tion is that this robot was also perceived as less useful, since a social robot displaying low inter-

action skills defeats its purpose of socially interacting with people [14]. And since perceived

usefulness is known to influence the attitude towards as well as the usage intention of a tech-

nology [15], this could explain why the competence and the sociability of the robot displaying

low interaction skills are evaluated more negatively. These negative evaluations of the robot’s

attributes then cause people to also negatively evaluate the interaction with the robot in

general.

Not many people have interacted with a social robot before and thus often rely on available

information sources like mass media, which usually present a rather positively biased view on

the skills of social robots [11, 30]. This was shown to lead to heightened expectations [11],

which increases the possibility that these expectations are negatively violated when people

interact with a real social robot which does not measure up to those high expectations [24, 26].

These violations of expectancies are linked to detrimental communication outcomes in inter-

personal settings [32] and were also shown to lead to disappointment, mistrust, and rejection

in human-robot interactions [24, 26]. This might have also happened in this setting since peo-

ple were told that they will be interacting with a social robot in possession of sophisticated

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 12 / 18

interaction skills. In case of the robot which then displayed low interaction skills, this could

have caused a negative expectancy violation leading to a negative evaluation of this robot and

the interaction with it.

The robot’s expected future role

In addition to the display of different levels of interaction skills, it was examined what effects

different expectations of the robot’s future purpose have. The robot which was portrayed as

trying to compete with and eventually replace humans was evaluated as less sociable compared

to the robot which was expected to become a beneficial assistant. Moreover, when people

expected the robot to become a competitor, the negative evaluations of the robot’s attributes

also lead them to generally evaluate the interaction with the robot more negatively–at least on

a marginally significant level.

This confirms and extends previous research on the two different prominent views on

social robots [11, 28, 31] by showing that framing a robot as competitor or assistant signifi-

cantly influences how sociable this robot is perceived and also partly how an interaction with

this robot is evaluated subsequently. A previous study showed that people who know more

“bad” fictional robot characters have stronger negative expectancies regarding robots becom-

ing the superior life form which will outrace and dominate us humans one day [11]. The cur-

rent study extends these survey-based findings in a systematic-experimental way and shows

that the fear of a robot becoming a threatening competitor also causes people to evaluate the

sociability of the robot they just interacted with more negatively.

These findings hold important implications for the future shaping of the perception of

social robots. If people have no access to personal experiences or existing categories into which

something or someone can be encoded, they have to rely on statements and judgements from

third parties [54]. Since social robots are not very common yet, many people base their knowl-

edge about this technology on what they learn from others, a main source being mass media

[11]. Here, the negative and fearsome view of robots developing their own agenda, revolting

against and dominating humans is often promoted [28–30]. However, as this study shows,

people’s reliance on third party information can also be used to shape people’s expectations

and their subsequent evaluations regarding social robots in a positive way. This could be

achieved by explaining and emphasizing the positive features and anticipated usefulness of

social robots.

What kind of role people expected the robot to have in the future had no effect on how

competent the robot was perceived, probably because in both descriptions the robot was

described as having high interaction skills. After that, only the actual competence of the social

robot, represented by its displayed level of interaction skills, seemed to be able to influence

people’s evaluation of the robot’s competence. Also, the indirect influence of the robot’s

expected future role on the general evaluation of the robot was rather weak. This can be

explained by the circumstance that the robot’s perceived competence had a stronger influence

on the interaction evaluation than its perceived sociability. The robot’s expected future role

mainly influenced its perceived sociability and not so much its perceived competence.

The robot’s expected future role influencing the perception of its

interaction skill level

There was no direct interaction effect of the robot’s interaction skill level and the expectation

of its future role on the evaluation of the robot’s competence or sociability. Consequently,

there was also no indirect interaction effect on the general evaluation of the interaction with

the robot. The expectation of the robot to become an assistant or a competitor seems not to

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 13 / 18

have affected how the robot’s level of interaction skills were perceived. An explanation could

be that the effect of the robot’s behavior during the interaction was too strong by itself to be

influenced by what kind of role people expect this robot to have one day in the future. This is

also supported by the effect sizes, which were largest for the robot’s level of interaction skills.

The behavior is what people experienced live during their interaction with the robot, while the

expected role was probably perceived as being far in the future and thus not as present as the

behavior displayed live and in that moment. This explanation also goes in line with the insights

by Reeves and Rickenberg [13], who emphasize the extraordinary influence of an agent’s

behavior during an interaction on the subsequent evaluation of the agent.

Also, the low interaction skill level was probably brought to the fore during the interaction

since people usually expect precision, sophisticated abilities, and high efficiency from robots

[21, 23]. Since all these aspects were not given in the condition with the poorly skilled social

robot, this robot was likely perceived as useless given its purpose of socially interacting with its

environment [14]. This perception appears not to be affected by which role the robot is sup-

posed to take over in the future, probably because with these poor skills the robot seems

unsuitable for any kind of social setting.

People’s individual background

In this study, we additionally considered various personality variables. The influence of peo-

ple’s technological background, their experiences with real and fictional robots, and their neg-

ative expectancies regarding robots were taken into account while examining the main effects

of the robot’s level of interaction skills and the robot’s expected future role. However, there

was no significant effect found caused by these variables. There was a marginally significant

indirect effect of people’s technological background on the general evaluation of the interac-

tion with the robot. People with a more pronounced technological background, which

includes their technical affinity and their locus of control when using technology, evaluated

the interaction with the robot better. This can likely be traced back to these people being more

enthusiastic about technology in general, which was already found in a previous study to lead

to a more positive evaluation of the technology and the interaction with it [7].

An explanation for the overall not significant effects of people’s individual backgrounds

could be that the effects of the robot’s expected future role and, even more, the robot’s behavior

during the interaction were so strong that people’s technological background, their previous

experiences with robots as well as their negative expectancies regarding robots did not make

any further notable difference in the evaluation of the robot and the interaction with it. This

again is supported by the findings of human-agent interaction studies, such as the one by Rick-

enberg and Reeves [13], which suggest that the behavior of an agent plays a pivotal role for the

effect this agent has on a potential user.

Limitations and future research

The study has some limitations regarding the generalizability of the results since most partici-

pants were students and thus predominantly young and highly educated. In future studies, the

setting should also be moved to the participants´ personal environments and instead of a one-

time event, interactions with the robot should take place several times over a longer period of

time to examine long-term effects. Moreover, the dependent variable measures were self-

reported, an experiment with behavioral measurements could bring further insights, for exam-

ple, with regard to the willingness to interact with this robot and robots in general. However,

we would like to emphasize that the participants of this study had a real interaction with a

robot, which extend insights of studies solely based on surveys and offers a crucial basis for a

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 14 / 18

more realistic evaluation of the robot’s characteristics and the interaction with this robot. In

future studies, other pre-existing or interactional characteristics should be considered along

with other factors which influence the perception and evaluation of a robot and the interaction

with a robot. For instance, the appearance of a robot was found to have a great influence on

people’s perception of it [1–4] and might also affect how the robot’s present and future role are

perceived. Thus, appearance should also be considered in future studies dealing with the ques-

tion which variables may influence the evaluation of a social robot.

Conclusion

The current study provides novel insights regarding the effects of a social robot’s level of inter-

action skills as a behavioral component, the robot’s expected future role, and people’s individ-

ual backgrounds on the evaluation of this robot’s sociability and competence and subsequently

on the general evaluation of the interaction with this robot.

Since knowledge of “malevolent” fictional robots was found to be related to negative expec-

tancies regarding real robots [11], this study systematically examined whether a robot which is

expected to become a threatening competitor is evaluated differently after people had a real

interaction with it. Results showed indeed that the robot presented as a future competitor is

evaluated as less sociable.

However, the most central insight from this study confirms previous findings [13] in show-

ing that the behavior of the robot in the actual interaction has the most pivotal influence on

how the robot and the interaction with it are evaluated. With the largest effect sizes of this

study, a social robot displaying low interaction skills leads to a more negative evaluation of the

robot’s sociability and competence as well as subsequently of the interaction with the robot.

Summing up, whether people expect a robot to become a competitor to them in the future

plays an important role in how this robot’s sociability is evaluated. However, the robot’s behav-

ior in the actual interaction appears to be the key variable influencing people’s evaluations of

the robot and the interaction with it in general. Thus, when people get to the point where they

interact with a social robot themselves, the robot’s behavior will be more decisive for how this

robot is perceived and evaluated than people’s negative expectations regarding the robot’s

future role or prior attitudes and personality traits.

Supporting information

S1 File. Complete data set.

(SAV)

Author Contributions

Conceptualization: Aike C. Horstmann.

Data curation: Aike C. Horstmann.

Formal analysis: Aike C. Horstmann.

Investigation: Aike C. Horstmann.

Methodology: Aike C. Horstmann, Nicole C. Krämer.

Project administration: Aike C. Horstmann, Nicole C. Krämer.

Resources: Aike C. Horstmann, Nicole C. Krämer.

Software: Aike C. Horstmann.

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 15 / 18

Supervision: Nicole C. Krämer.

Validation: Aike C. Horstmann, Nicole C. Krämer.

Visualization: Aike C. Horstmann.

Writing – original draft: Aike C. Horstmann.

Writing – review & editing: Aike C. Horstmann, Nicole C. Krämer.

References

1. Goetz J, Kiesler S, Powers A. Matching robot appearance and behavior to tasks to improve human-

robot cooperation. In: Proceedings of the 12th IEEE International Workshop on Robot and Human Inter-

active Communication—RO-MAN ’03; 2003 Oct 31-Nov 02; Millbrae, CA. Piscataway, NJ: IEEE; 2003.

p. 55–60. https://doi.org/10.1109/ROMAN.2003.1251796

2. Li D, Rau PLP, Li Y. A cross-cultural study: Effect of robot appearance and task. International Journal of

Social Robotics. 2010; 2:175–86. https://doi.org/10.1007/s12369-010-0056-9

3. Rosenthal-von der Pütten AM, Krämer NC, Herrmann J. The effects of humanlike and robot-specific

affective nonverbal behavior on perception, emotion, and behavior. International Journal of Social

Robotics. 2018; 10:569–82. https://doi.org/10.1007/s12369-018-0466-7

4. Syrdal DS, Dautenhahn K, Woods SN, Walters ML, Koay KL. Looking good? Appearance preferences

and robot personality inferences at zero acquaintance. In: Tapus A, Michalowski M, Sabanovic S, edi-

tors. Papers from the AAAI Spring symposium: Multidisciplinary collaboration for socially assistive

robotics. Menlo Park, CA: AAAI Press; 2007. p. 86–92.

5. Mutlu B, Yamaoka F, Kanda T, Ishiguro H, Hagita N. Nonverbal leakage in robots: Communication of

intentions through seemingly unintentional behavior. In: Scheutz M, Michaud F, Hinds P, Scassellati B,

editors. Proceedings of the 4th International Conference on Human-Robot Interaction—HRI ’09; 2009

Mar 09–13; La Jolla, CA, USA. Piscataway, NJ: IEEE; 2009. p. 69. https://doi.org/10.1145/1514095.

1514110

6. Eyssel F, Kuchenbrandt D, Bobinger S. Effects of anticipated human-robot interaction and predictability

of robot behavior on perceptions of anthropomorphism. In: Billard A, Kahn P, Adams JA, Trafton G, edi-

tors; 2011 Mar 09–16; Lausanne, Switzerland. New York, NY: ACM Press; 2011. p. 61–68. https://doi.

org/10.1145/1957656.1957673

7. Horstmann AC, Bock N, Linhuber E, Szczuka JM, Straßmann C, Krämer NC. Do a robot’s social skills

and its objection discourage interactants from switching the robot off? PLoS ONE. 2018; 13:e0201581.

https://doi.org/10.1371/journal.pone.0201581 PMID: 30063750

8. Straßmann C, Rosenthal-von der Pütten AM, Krämer NC. With or against each other? The influence of

a virtual agent’s (non)cooperative behavior on user’s cooperation behavior in the Prisoners’ Dilemma.

Advances in Human-Computer Interaction. 2018:2589542. https://doi.org/10.1155/2018/2589542

9. Bartneck C. From fiction to science–A cultural reflection of social robots. In: Proceedings of the CHI

2004 Workshop on Shaping Human-Robot Interaction—CHI ’04; 2004 Apr 24–29; Vienna, Austria. New

York, NY: ACM Press; 2004. p. 1–4.

10. Bruckenberger U, Weiss A, Mirnig N, Strasser E, Stadler S, Tscheligi M. The good, the bad, the weird:

Audience evaluation of a “real” robot in relation to science fiction and mass media. In: Jamshidi M, edi-

tor. Advance Trends in Soft Computing: Proceedings of the World Conference on Soft Computing—

WCSC ’13. Cham, Switzerland: Springer; 2013. p. 301–310. https://doi.org/10.1007/978-3-319-02675-

6_30.

11. Horstmann AC, Krämer NC. Great expectations? Relation of previous experiences with social robots in

real life or in the media and expectancies based on qualitative and quantitative assessment. Front Psy-

chol. 2019; 10:939. https://doi.org/10.3389/fpsyg.2019.00939 PMID: 31114525

12. Kriz S, Ferro TD, Damera P, Porter JR. Fictional robots as a data source in HRI research: Exploring the

link between science fiction and interactional expectations. In: Proceedings of the 19th IEEE Interna-

tional Symposium on Robot and Human Interactive Communication—RO-MAN; 2010 Sep 09–15; Viar-

eggio, Italy. Piscataway, NJ: IEEE; 2010. p. 458–463. https://doi.org/10.1109/ROMAN.2010.5598620

13. Rickenberg R, Reeves B. The effects of animated characters on anxiety, task performance, and evalua-

tions of user interfaces. In: Turner T, editor. Proceedings of the SIGCHI conference on Human Factors

in Computing Systems; 2000 Apr 1–6; The Hague, The Netherlands. New York, NY: ACM; 2000. p. 49–

56. https://doi.org/10.1145/332040.332406

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 16 / 18

14. Dautenhahn K. Socially intelligent robots: Dimensions of human-robot interaction. Philosophical Trans-

actions of the Royal Society of London. 2007; 362:679–704. https://doi.org/10.1098/rstb.2006.2004

PMID: 17301026

15. Davis FD. User acceptance of information technology: System characteristics, user perceptions and

behavioral impacts. International Journal of Man-Machine Studies. 1993; 38:475–87. https://doi.org/10.

1006/imms.1993.1022

16. Burgoon JK. Interpersonal Expectations, Expectancy Violations, and Emotional Communication. Jour-

nal of Language and Social Psychology. 1993; 12:30–48. https://doi.org/10.1177/0261927X93121003

17. Burgoon JK, Le Poire BA. Effects of communication expectancies, actual communication, and expec-

tancy disconfirmation on evaluations of communicators and their communication behavior. Human

Communication Research. 1993; 20:67–96. https://doi.org/10.1111/j.1468-2958.1993.tb00316.x

18. Horstmann AC, Krämer NC. When a Robot Violates Expectations. In: Belpaeme T, Young J, Gunes H,

Riek L, editors. Companion of the 2020 ACM/IEEE International Conference on Human-Robot Interac-

tion—HRI ’20; 2020 Mar 23–26; Cambridge, UK. New York, NY: ACM; 2020. p. 254–256. https://doi.

org/10.1145/3371382.3378292

19. Karrer K, Glaser C, Clemens C, Bruder C. Technikaffinität erfassen: Der Fragebogen TA-EG [Measur-

ing technical affinity—the questionnaire TA-EG]. Der Mensch im Mittelpunkt technischer Systeme.

2009; 8:196–201.

20. Beier G. Kontrollüberzeugungen im Umgang mit Technik [Locus of control when using technology].

Report Psychologie. 1999; 9:684–93.

21. Arras KO, Cerqui D. Do we want to share our lives and bodies with robots? A 2000 people survey. Tech-

nical Report Nr. 0605–001, Autonomous Systems Lab, Swiss Federal Institute of Technology, EPFL

2005. https://doi.org/10.3929/ETHZ-A-010113633

22. Dautenhahn K, Woods S, Kaouri C, Walters ML, Koay KL, Werry I. What is a robot companion—Friend,

assistant or butler? In: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and

Systems—IROS ’05; 2005 Aug 02–06; Edmonton, Canada. Piscataway, NJ: IEEE; 2005. p. 1192–

1197. https://doi.org/10.1109/IROS.2005.1545189

23. Ezer N, Fisk AD, Rogers WA. Attitudinal and intentional acceptance of domestic robots by younger and

older adults. In: Stephanidis C, editor. Proceedings of the 5th International Conference on Universal

Access in Human-Computer Interaction—UAHCI ’09; 2009 Jul 19–24; San Diego, CA. Berlin/Heidel-

berg, Germany: Springer Berlin Heidelberg; 2009. p. 39–48. https://doi.org/10.1007/978-3-642-02710-

9_5

24. Kwon M, Jung MF, Knepper RA. Human expectations of social robots. In: Bartneck C, Nagai Y, Paiva

A, Sabanovic S, editors. Proceedings of the 11th ACM/IEEE International Conference on Human Robot

Interation—HRI ’16; 2016 Mar 07–10; Christchurch, New Zealand. Piscataway, NJ: IEEE; 2016.

p. 463–464. https://doi.org/10.1109/HRI.2016.7451807

25. Fong T, Nourbakhsh I, Dautenhahn K. A survey of socially interactive robots. Robotics and Autonomous

Systems. 2003; 42:143–66. https://doi.org/10.1016/S0921-8890(02)00372-X

26. Komatsu T, Yamada S. Adaptation gap hypothesis: How differences between users’ expected and per-

ceived agent functions affect their subjective impression. Journal of Systemics, Cybernetics and Infor-

matics. 2011; 9:67–74.

27. Asimov I. Little Lost Robot. New York, NY: Street & Smith; 1947.

28. Ray C, Mondada F, Siegwart R. What do people expect from robots? In: Proceedings of the IEEE/RSJ

International Conference on Intelligent Robots and Systems—IROS ’08; 2008 Sep 22–26; Nice,

France. Piscataway, NJ: IEEE; 2008. p. 3816–3821. https://doi.org/10.1109/IROS.2008.4650714

29. Weiss A, Igelsböck J, Wurhofer D, Tscheligi M. Looking forward to a “robotic society”? International

Journal of Social Robotics. 2011; 3:111–23. https://doi.org/10.1007/s12369-010-0076-5

30. Khan Z. Attitudes Towards Intelligent Service Robots. Stockholm, Sweden: Royal Institute of Technol-

ogy; 1998.

31. Oestreicher L, Eklundh K. User expectations on human-robot co-operation. In: Proceedings of the 15th

IEEE International Symposium on Robot and Human Interactive Communication—RO-MAN ’06; 2006

Sep 06–08; Hatfield, UK. Piscataway, NJ: IEEE; 2006. p. 91–96. https://doi.org/10.1109/ROMAN.2006.

314400

32. Burgoon JK, Hale JL. Nonverbal expectancy violations: Model elaboration and application to immediacy

behaviors. Communication Monographs. 1988; 55:58–79. https://doi.org/10.1080/

03637758809376158

33. Asimov I. I, Robot. New York, NY: Gnome Press; 1950.

34. Clarke AC. 2001: A Space Odyssey. New York, NY: New American Library; 1968.

35. Dick PK. Do Androids Dream of Electric Sheep? New York, NY: Doubleday; 1968.

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 17 / 18

36. Gaul S, Ziefle M, Arning K, Wilkowska W, Kasugai K, Röcker C, et al. Technology acceptance as an

integrative component of product developments in the medical technology sector. In: Proceedings of

the 3rd Ambient Assisted Living Conference—AAL ’10; 2010 Jan 26–27; Berlin, Germany. Berlin, Ger-

many: VDE-Verlag; 2010. p. 26–27.

37. Bar-Cohen Y, Hanson D, Marom A. The coming robot revolution: Expectations and fears about emerg-

ing intelligent, humanlike machines. New York, NY: Springer; 2009.

38. Nomura T, Kanda T, Suzuki T, Kato K. Prediction of human behavior in human-robot interaction using

psychological scales for anxiety and negative attitudes toward robots. IEEE Transactions on Robotics.

2008; 24:442–51. https://doi.org/10.1109/TRO.2007.914004

39. Cohen J. Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum; 1988.

40. Wullenkord R, Fraune MR, Eyssel F, Sabanovic S. Getting in Touch: How imagined, actual, and physi-

cal contact affect evaluations of robots. In: Proceedings of the 25th IEEE International Symposium on

Robot and Human Interactive Communication—RO-MAN ’16; 2016 Aug 26–31; New York, NY. Piscat-

away, NJ: IEEE; 2016. p. 980–985. https://doi.org/10.1109/ROMAN.2016.7745228

41. Dahlbäck N, Jönsson A, Ahrenberg L. Wizard of Oz studies: Why and how. Knowledge-Based Systems.

1993; 6:258–66. https://doi.org/10.1016/0950-7051(93)90017-N

42. McCroskey JC, McCain TA. The measurement of interpersonal attraction. Speech Monographs. 1974;

41:261–6. https://doi.org/10.1080/03637757409375845

43. McCroskey JC, Young TJ. Ethos and credibility: The construct and its measurement after three

decades. Central States Speech Journal. 1981; 32:24–34. https://doi.org/10.1080/

10510978109368075

44. Heerink M, Krose B, Evers V, Wielinga B. Measuring acceptance of an assistive social robot: A sug-

gested toolkit. In: Proceedings of the 18th IEEE International Workshop on Robot and Human Interac-

tive Communication—RO-MAN ’09; 2009 Sep 27-Oct 02; Toyama, Japan. Piscataway, NJ: IEEE;

2009. p. 528–533. https://doi.org/10.1109/ROMAN.2009.5326320

45. Burgoon JK, Walther JB. Nonverbal expectancies and the evaluative consequences of violations.

Human Communication Research. 1990; 17:232–65. https://doi.org/10.1111/j.1468-2958.1990.

tb00232.x

46. Burgoon JK, Walther JB, Baesler EJ. Interpretations, evaluations, and consequences of interpersonal

touch. Human Communication Research. 1992; 19:237–63. https://doi.org/10.1111/j.1468-2958.1992.

tb00301.x

47. Burgoon JK, Hale JL. Validation and measurement of the fundamental themes of relational communica-

tion. Communication Monographs. 1987; 54:19–41. https://doi.org/10.1080/03637758709390214

48. Nomura T, Sugimoto K, Syrdal DS, Dautenhahn K. Social acceptance of humanoid robots in Japan: A

survey for development of the Frankenstein Syndrome Questionnaire. In: Proceedings of the 12th

IEEE-RAS International Conference on Humanoid Robots; 2012 Nov 29-Dec 01; Osaka, Japan. Piscat-

away, NJ: IEEE; 2012. p. 242–247. https://doi.org/10.1109/HUMANOIDS.2012.6651527

49. Ullman JB, Bentler PM. Structural Equation Modeling. In: Weiner IB, Schinka JA, Velicer WF, editors.

Handbook of Psychology: Research Methods in Psychology. 2nd ed. Hoboken, NJ: John Wiley &

Sons, Inc.; 2003. p. 661–690. https://doi.org/10.1002/0471264385.wei0224

50. Hair JF, Black WC, Babin BJ, Anderson RE. Multivariate data analysis. Andover, Hampshire, UK: Cen-

gage Learning EMEA; 2019.

51. Novikova SI, Richman DM, Supekar K, Barnard-Brak L, Hall D. NDAR. In: Urbano RC, editor. Using

secondary datasets to understand persons with developmental disabilites and their families. Amster-

dam: Elsevier/Academic Press; 2013. p. 123–153. https://doi.org/10.1016/B978-0-12-407760-7.

00003–7

52. Hu L-t, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria

versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal. 1999; 6:1–55.

https://doi.org/10.1080/10705519909540118

53. Hu L-t, Bentler PM. Evaluating model fit. In: Hoyle RH, editor. Structural Equation Modeling Concepts

Issues and Applications. London, UK: Sage Publications, Inc.; 1995. p. 76–99.

54. Darley JM, Fazio RH. Expectancy confirmation processes arising in the social interaction sequence.

American Psychologist. 1980; 35:867–81. https://doi.org/10.1037/0003-066X.35.10.867

PLOS ONE Expectations vs. actual behavior of a social robot

PLOS ONE | https://doi.org/10.1371/journal.pone.0238133 August 21, 2020 18 / 18

Copyright of PLoS ONE is the property of Public Library of Science and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.