arXiv:1606.02603v1  [cs.RO]  8 Jun 2016
Robot-stated limitations but not intentions promote user assistance
David Cameron1 and Ee Jing Loh2 and Adriel Chua2 and Emily Collins1 and
Jonathan M. Aitken1
and James Law1
Abstract.
Human-Robot-Interaction (HRI) research is typically
built around the premise that the robot serves to assist a human in
achieving a human-led goal or shared task. However, there are many
circumstances during HRI in which a robot may need the assistance
of a human in shared tasks or to achieve goals. We use the ROBO-
GUIDE model as a case study, and insights from social psychology,
to examine how a robots personality can impact on user cooperation.
A study of 364 participants indicates that individuals may prefer
to use likable social robots ahead of those designed to appear more
capable; this outcome reflects known social decisions in human inter-
personal relationships [7]. This work further demonstrates the value
of social psychology in developing social robots and exploring HRI.
1
INTRODUCTION
The use of autonomous, mobile service robots in the workplace is
expected to grow substantially in the coming years [16]. Robots
are anticipated to work side-by-side with people, assisting or col-
laborating with employees on a variety of tasks. As a consequence,
Human-Robot Interaction (HRI) research typically explores interac-
tions based around a robot in a supportive or assistive role for the
human user [14].
Current research in assistive robotics highlights the importance
of ensuring users’ trust that the robot can provide effective assis-
tance. In scenarios of robot-assisted navigation, this may require:
near-immediate user trust placed in a robot guide [28]; maintenance
of appropriate levels of user trust, so neither over-reliance nor under-
use occurs [19] [11]; and recovery of user trust after mistakes in au-
tomation [20, 15]. Across the literature, user trust is typically asso-
ciated with a robot establishing its capacity in meeting goals as an
autonomous agent.
However, there are instances in HRI where an assistive robot guide
may require user support to achieve its aims or complete shared tasks.
For example, a robot may hold incomplete information about its sit-
uation and require user input (e.g. asking for its location [29] or
for directions [2]) or face a task requiring manual intervention by
a user (e.g. autonomous mobile robots encountering physical barri-
ers to progress [5, 29]). At this juncture, a robot cannot effectively
demonstrate its capacity to operate as an autonomous agent and so
typical channels for engendering user trust may be at risk. Robots
requiring, rather than solely providing, user assistance in complex
and social environments is an emerging topic [4] and so it becomes
essential to study its impact on users’ experience of HRI.
Research on the topic of robots requiring assistance identifies ef-
fective means for robots to determine their limits and when human
1 Sheffield Robotics, University of Sheffield, UK, email: {d.s.cameron,
e.c.collins, jonathan.aitken, j.law}@sheffield.ac.uk
2 Department of Psychology, University of Sheffield, UK email: {ejloh2, dx-
achua1}@sheffield.ac.uk
intervention is required [29], or how to locate users that can offer as-
sistance [30]. However, determining effective means of how socially
adaptive robots request help to encourage user cooperation and as-
sistance is still a challenge in HRI.
To approach this challenge, we draw from social psychological
models exploring cooperation between agents. Social psychologi-
cal insights can be beneficial in exploring HRI, given it is a novel
and still-developing research topic [10]. Robots need to engage the
user, particularly those that require user input (either directly through
interacting with the robot or shaping the environment to meet the
robot’s needs). In particular, the emerging field of social robotics
considers the optimization of a robot’s morphology [16, 3] and even
its ‘personality’ [12, 6] as important to user perceptions of experi-
enced HRI.
In this paper, we compare the impact of two different robot per-
sonalities on individuals’ willingness to use a robot requiring assis-
tance. We construct a friendly robot personality that empahsises its
limitations and a capable personality that emphasises its intentions
in interactions. We further explore the impact that these personal-
ity features have on three factors identified in social psychology as
having impact on human-human cooperation and assistance: liking,
trust, and ambiguity. We test a model of robot personality; factors
of liking, trust, and ambiguity; and individuals’ willingness to help a
robot requiring assistance in an applied HRI scenario.
1.1
Pathways to cooperation
The following paragraphs introduce the three factors antecedent to
cooperation investigated in this experiment. We target specifically the
under-explored point of interaction: in which a robot needs human
assistance, within the broader context of it operating as an assistive
robot in a workplace environment.
1.1.1
Liking
Future robotic systems are considered likely to operate more as team-
mates rather than tools [25]. Understanding the impact this change
could have on HRI is therefore vital to adapt work-forces to accom-
modate future, robotic teammates. Occupational psychology models
of teamwork and cooperation in the workplace may provide a good
foundation for understanding HRI with these robotic teammates. Re-
search from occupational psychology indicates that individuals pre-
fer to populate their cooperative working networks with people they
like, ahead of those more capable in the cooperative role (but not
liked) [7, 33]. Liking of individuals is considered to play a substan-
tial role in motivations for cooperation and, particularly as the rela-
tionship develops, displace antecedent cognitive motivations for co-
operation [24].
1.1.2
Trust
A prominent model from occupational psychology of interpersonal
cooperation in teammates identifies trust as its foundation [22].
McAllister argues that to successfully achieve in a task requiring two
agents working together, both need to trust each other. Specifically,
cooperation is thought to benefit from both affective trust (built by
personable interactions from the partner) and cognitive trust (built
from evidence that one’s partner carries out responsibilities compe-
tently) [22]. Analogues for both forms are seen in HRI [14]. A user’s
perception of a robot’s performance (analogous to cognitive trust)
and a user’s perception of a robot’s attributes, such as personality
(analogous to affective trust), are found to positively contribute to-
wards user trust in robots [14].
1.1.3
Ambiguity
Classic social psychology research indicates that ambiguity in assis-
tive scenarios results in substantial detriment to individuals’ proac-
tive assistive behavior [9]. For many, HRI in cooperative environ-
ments may currently be entirely novel. As a result, individuals could
face ambiguity in HRI situations and, without an indication of a
robot’s limitations, uncertainty regarding whether the robot requires
user cooperation. Alternatively, lack of a clear plan or intention com-
municated by a robot may create further ambiguity in how the indi-
vidual may best cooperate with the robot, limiting the action taken.
2
Scenario
To explore how a robot’s personality may influence the above fac-
tors and user cooperation and assistance towards a robot, it is essen-
tial to consider an interactive social robotics scenario in which these
circumstances arise. The ROBOtic GUidance and Interaction DE-
velopment (ROBO-GUIDE) project [18] offers an ideal scenario to
explore these factors. ROBO-GUIDE is implemented on the Pioneer
LX mobile platform to autonomously navigate a multistory building,
leading users (visitors to the building) to their chosen destination.
Critically, there are elements in the tour which require user in-
tervention to remove barriers to progress, such as using an elevator
to navigate between floors. While the ROBO-GUIDE platform can
identify which floor it is on [23], it is currently unable to call for
an elevator itself: it can neither manually operate elevator buttons
nor remotely command elevator operation. As a result, the robot re-
lies on user cooperation to press buttons to call the elevator and se-
lect the required floor to progress. To direct the user, statements of
the robot’s limitations or intentions are communicated using the on-
board speech synthesizer.
Our focus for this study is the point at which the platform changes
between floors as it navigates the multistory building. Use of ROBO-
GUIDE requires users to place trust in an autonomous way-finding
robot, whilst also assisting the robot in overcoming obstacles or bar-
riers to progress. For both the user and the robot, this scenario is
identified as a simple and low-risk circumstance in which a human
user can act in order to meet a robot’s needs [5].
2.1
Promoting user cooperation in HRI
We identify how scenario-specific robot-stated limitations or inten-
tions can influence factors for user cooperation and assistance. The
following sub-sections show proposed control-condition statements
or requests in quotes and normal font, whereas supplemental, exper-
imental statements are in the same quotes and italics.
2.1.1
Liking
ROBO-GUIDE’s primary purpose is to lead new visitors to the
robotics laboratory. It is anticipated that user liking of the robot
would be greater with supplemental phrases regarding its offering
‘face-to-face’ direct assistance to the human user. To promote liking,
we include friendly and relatable references by the robot that is as-
sisting the user as a tour guide: “Please follow me; I am here as your
tour guide”.
2.1.2
Trust
It is anticipated that developing affective trust overlaps substantially
with developing user liking [22] and can be built by demonstrating
trust in others [21, 32]. To promote affective trust, we supplement re-
quests from the robot for user cooperation at the elevator with identi-
fication of its limitations: “Please press the down button; I cant quite
reach the buttons”.
In contrast, cognitive trust is developed through demonstration of
an agent’s competency in meeting its intended and/or required re-
sponsibilities [22]. It is anticipated that users’ cognitive trust of the
robot’s way-finding capabilities would be greater with the inclusion
of additional phrases that directly address its intended aims: “Please
follow me to get to the robot labs”.
2.1.3
Ambiguity
It is anticipated that many visitors would be unfamiliar with HRI
or social robots and could find the experience unusual. Interacting
with a novel robot, especially a robot that needs help, is anticipated
to create ambiguous situations for users. To reduce ambiguity, we
use friendly and relatable references by the robot about its limita-
tions, which a user could help with in the HRI scenario: “Please press
ground floor; good thing you’re here to do that for me”.
It is further anticipated that ROBO-GUIDE’s statements of in-
tentions when requesting help will reduce ambiguity and promote
cooperation. Without declaring why the tasks are to be completed,
requests for help could be ambiguous in their purpose. Intentions
should demonstrate the robot has a clear goal it is trying to achieve
but that it is now facing an obstacle and so asks for help: “Please
press the down button to call the lift” and “Please press ground floor
for the Robot Labs”.
2.1.4
A model of user cooperation
We predict that a robot stating its limitations and intentions when
requesting assistance will promote individuals’ willingness to coop-
erate with assisting the robot. Pathways by which this is anticipated
to occur, factors of liking, trust, and ambiguity, are over-viewed in
Figure 1.
The robot stating its limitations is predicted to: promote user lik-
ing, promote users’ affective trust towards the robot, and limit ambi-
guity in the situation. The robot stating its intentions is predicted to:
promote users’ cognitive trust towards the robot and limit ambiguity
in the situation. These outcomes are in turn anticipated to positively
impact on users’ perceptions of assisting the robot
3
METHOD
3.1
Design
A 2x2 independent measures design was implemented. The four con-
ditions comprised of the presence or absence of key statements re-
Figure 1.
Pathways by which robot-stated limitations and intentions impact
on user-cooperation
garding i) the robot’s limitations, and ii) intentions. Participants were
randomly allocated to a single condition and the Qualtrics survey en-
gine prohibited repeat participation.
3.2
Materials
3.2.1
Videos
Four HRI videos were prepared; one for each condition. The videos
represented a typical use of the ROBO-GUIDE: an individual is
greeted by the robot at the building entrance and instructed to fol-
low it. The robot and user travel along a corridor to an elevator, at
which point the robot instructs the user to press the call button (Ex-
ample of HRI in Figure 2). On entering the elevator, the robot again
instructs the user to press the relevant button for the floor. The robot
and user then leave the elevator at the target floor and travel along
another corridor to the robot-lab - the final destination.
All four films have minimal visual differences as non-critical
scenes were used across conditions. Critical scenes differ in audio
(i.e. the condition-specific words spoken by the robot) and subtitles
of the robot’s speech. Videos lasted 90 seconds.
3.2.2
Questionnaires
Perspectives on the observed HRI were assessed using the Godspeed
Questionnaire [1]. Measurements of liking were taken using the rel-
Figure 2.
Example of user assisting robot in HRI (Intentions and Limita-
tions condition)
evant Godspeed sub-scale. In addition, 3 ad-hoc measures were de-
veloped to assess participants’ perspectives of the trustworthiness of
ROBO-GUIDE, ambiguity of the interaction, and their likelihood of
using ROBO-GUIDE.
Trustworthiness was assessed using an ad-hoc 16-item scale de-
rived from words previously identified as strongly associated with
trust or mistrust in automated systems [17]. Each of the following
items was scored on a 7-point scale from not at all to extremely,
headed ‘To what extent would you describe ROBO-GUIDE as...’ ca-
pable, competent, confident, deceptive, false, honest, honorable, in-
capable, incompetent, loyal, misleading, reliable, trustworthy, unre-
liable, unsure, untrustworthy.
Ambiguity was assessed using a 4-item measure (alpha = 0.93)
concerning the clarity of the robot’s requests regarding, why,
whether, when and how a user may need to assist the robot. Each
item was scored on a 7-point scale from strongly agree to strongly
disagree.
Likelihood of use was assessed using a 4-item measure (alpha =
0.78), with each item scored on a 7-point scale from strongly agree
to strongly disagree. Items included: perceiving the robot as being
fun to use, and intentions to use the robot in an unfamiliar building.
3.3
Participants
Participants were recruited through staff and student university vol-
unteer mailing lists; 442 participants signed up for the study. A sur-
vey timer indicated that 78 participants did not fully watch the HRI
video stimulus so were excluded from the study. Of the remaining
364 participants: 196 were female, 152 were male, and 16 declined
to identify gender; 240 participants were British, 114 were from over-
seas, 10 declined to identify nationality; aged M = 27.67, SD = 8.99.
Participants were given the opportunity to win one of two 50 gift
vouchers as recompense for their time.
3.4
Procedure
The survey was distributed through online mailing lists to reach
a broad audience and delivered through the online survey engine
Qualtrics (Version 06.2015; Copyright 2015; Qualtrics, Utah). Par-
ticipants were first presented with a study information and consent
page. On agreeing to participate, individuals were randomly allo-
cated to one of the four conditions described above and presented
with an HRI video. Following the video, participants completed the
Godspeed questionnaire, measures rating trustworthiness of ROBO-
GUIDE, interaction ambiguity, and willingness to use the ROBO-
GUIDE. At the close of the questionnaire, participants could share
demographic details and register for a chance to win offered gift
vouchers. The study took participants approximately 10 minutes to
complete.
4
RESULTS
4.1
Randomisation Check
Participant numbers were evenly distributed across conditions x2(1,
N = 364) = 0.01, p = 0.92. For each condition, there were even distri-
butions of participants in terms of gender and nationality (max x2(1,
N = 354) = 3.31, p = 0.07).
4.2
Primary results
4.2.1
Liking
There was a significant main effect for the robot stating its limitations
F(1,356) = 53.407, p < 0.01. Participants who saw the interactions
in which the robot stated its limitations reported liking the robot to a
substantially greater extent (M = 3.77, S.E. = 0.05) than those who
saw control statements (M = 3.24, S.E. = 0.05). This is a large effect
observed (d = 0.77). There was also a significant main effect for the
robot stating its intentions F(1,356) = 7.25, p < 0.01. Participants
who saw the interactions in which the robot stated its intentions re-
ported liking the robot to a substantially greater extent (M = 3.60,
S.E. = 0.05) than those who saw control statements (M = 3.40, S.E.
= 0.05). This is a small effect observed (d = 0.28). There was no
significant interaction effect between the robot stating both its lim-
itations and intentions on participants’ liking F(1,356) = 0.87, p =
0.79.
4.2.2
Trust
The 16 items for trust were subjected to a principal axis factor anal-
ysis for the full sample of the 364 participants; missing values were
treated pairwise. Three factors with Eigenvalues greater than 1.00
[8] were extracted from the matrix, explaining 59% of the variance.
Inspection of the pattern matrix showed that all items loaded above
0.40 on one of the three factors. Items on the first factor are mainly
related to falsity: deceptive, false, misleading, unreliable, unsure, and
untrustworthy. Items on the second factor are mainly concerned with
affective trust: honest, honorable, loyal, and trustworthy. Finally,
items on the third factor are mainly concerned with cognitive trust
capable, competent, confident, reliable, and two negatively scored
items of incapable and incompetent.
However, there were no significant main effects for the robot stat-
ing its limitations nor intentions for either factors of affective trust
(Limitations F(1,351) = .11, p = .75; Intentions F(1,351) = .21, p =
.65) or cognitive trust (Limitations F(1,351) = 2.14, p = 0.14; Inten-
tions F(1,351 = .01, p = .98). Furthermore, there were no significant
interaction effects.
4.2.3
Ambiguity
There was a significant main effect for the robot stating its limitations
F(1,352) = 134.95, p < 0.01. Participants who saw the interactions
in which the robot stated its limitations reported substantially less
ambiguity in the situation (M = 1.87, S.E. = 0.10) than those who
saw control statements (M = 3.52, S.E. = 0.10). This is a large effect
observed (d = 1.22). There was no main effect for the robot stating its
intentions on participants’ perceptions of ambiguity in the interaction
F(1,352) = 0.43, p = 0.52.
There was a significant interaction effect between the robot stat-
ing both its limitations and its intentions F(1,352) = 4.72, p = 0.03.
The robot solely stating its intentions promoted greater perceptions
of ambiguity in comparison to control statements (M = 3.72, SE =
0.14 versus M = 3.32, SE = 0.14). However, in combination with the
robot stating its limitations, participants’ ratings of ambiguity were
lower in comparison to the robot solely stating limitations only (M
= 1.76 SE = 0.14 versus M = 1.98, SE = 0.14). Simple main effects
analysis showed that the robot-stated intentions resulted in more am-
biguity than control statements, when presented without robot-stated
limitations (p = .04), but there were no differences when accompa-
nied with robot-stated limitations (p = .28). These results are pre-
sented in Figure 3.
Figure 3.
Interaction effects for intentions and limitations for user per-
ceived ambiguity
4.2.4
Willingness to use
There was a significant main effect for the robot stating its limitations
f (1,352) = 8.74, p < 0.05. Participants who saw the interactions in
which the robot stated its limitations reported greater willingness to
use the robot (M = 4.57, SE = 0.11) than those who saw control
statements (M = 4.26, SE = 0.11). There was no significant main
effect for the robot stating its intentions on participants’ willingness
to use the robot F(1,352) = 1.07, p = 0.30. There was no significant
interaction effect between the robot stating both its limitations and
intentions on participants’ willingness to use the robot F(1,352) =
0.09, p = 0.86.
Mediation analysis was run to determine if the main effect of the
robot-stated limitations promoting individuals’ willingness to use
ROBO-GUIDE could be explained by the effects seen above for
robot-stated limitations on user liking and ambiguity3. Indirect ef-
fects were computed following 1,000 bootstrapped samples at 95%
confidence intervals. The positive relationship between robot-stated
limitations and individuals’ willingness to use ROBO-GUIDE was
mediated by user liking but not ambiguity (see Table 1). Figure 4
identifies the pathways examined in the mediation analysis and coef-
ficients between constructs.
[htbp]
Table 1.
Mediators for robot-stated performance limitations on individuals’
willingness to use ROBO-GUIDE
Dependent
variable
Mediators
Point estimate
effects
95% confidence
intervals
Lower
Upper
Willingness
Liking
0.56
0.39
0.74
to Use
Ambiguity
0.12
-0.04
0.031
Total
0.31
0.01
0.62
R2 = 0.27, F(3,352) = 42.54, p < 0.01
3 User trust is not included as there were no main effects observed.
Figure 4.
Regression coefficients for the relationship between robot-stated
limitations and individuals’ willingness to use ROBO-GUIDE as mediated
by user liking and ambiguity. The regression coefficient between robot stated
limitations and willingness to use ROBO-GUIDE, controlling for user liking,
is in parenthesis (* p < 0.5; ** p < 0.01)
4.3
Demographics
There were significant main effects of nationality for both partici-
pants’ reports of liking the robot F(1,343) = 6.03, p = 0.02 and their
perceptions of the robot as being trustworthy F(1,342) = 21.68, p <
0.01. Individuals from overseas reported liking the robot more (M =
3.66, SE = 0.07) than those from the UK did (M = 3.45, SE = 0.05)
and showed more affective trust towards the robot (M = 4.50, SE =
0.12), than those from the UK did (M = 3.81, SE = 0.09). These are
small and medium effects (d = 0.28; d = 0.53) respectively. There
were no significant main effects for gender (Max F(1, 354) = 2.58, p
= 0.11).
There were significant interaction effects between nationality and
gender for both perceptions of robot competency F(1, 343) = 5.83, p
= 0.02 and willingness to use such a robot F(1,343) = 7.80, p 0.01.
In both cases, female participants from the UK scored higher (M =
5.31, SE = 0.08; M = 3.96, SE = 0.13) than female participants from
overseas (M = 4.94, SE= 0.13; M = 3.60, SE = 0.21), respectively;
whereas, male participants from overseas scored higher (M = 5.26,
SE = 0.12; M = 4.12, SE = 0.20) than male participants from the UK
(M = 5.11, SE = 0.10; M = 3.48, SE = 0.16), respectively.
5
Discussion
The results demonstrate a substantial difference in individuals’ re-
sponses to observing a friendly versus capable social robot. The con-
trast between a friendly, limitations-focused, and capable, intentions-
focused, robot is seen in individuals’ willingness to use the robot in
the future; only the former significantly differs from the control con-
dition. Moreover, the observed effect can be explained by the medi-
ating influence of individuals liking a limitations-focused robot. As
observed in occupational psychology [7], individuals prefer to fur-
ther interact with those they like, rather than those presented as be-
ing more capable. This study provides new evidence that strategies
used by individuals in interpersonal relationship development can be
extended to apply to social robotics HRI.
One key strength of the current work is the available sample size,
which would require an extensive effort to produce in a field robotics
study. The large sample size enabled a sufficiently powered 2x2 de-
sign study and a full factor analysis of the ad-hoc trust scale. Again,
the factor analysis indicates that it is extremely worthwhile explor-
ing HRI in terms of social psychological models of interpersonal re-
lationships. Factors extracted from the user responses closely corre-
sponded to the constructs of affective and cognitive trust [22]. Par-
ticipants may be applying their own understanding of human-human
social relationships to novel contexts (HRI) containing social agents.
There are several outcomes apparent due to individual differences,
such as an interaction effect of gender and nationality on willingness
to use ROBO-GUIDE. This emphasises the importance of human-
focused study design and robotic development [4]. The demographic
differences in individuals’ responses indicate that developing a so-
cially adaptive robot responsive to users, rather than attempting a
‘one-size-fits-all’ approach could greatly benefit user perceptions of
HRI. In terms of application to the ROBO-GUIDE scenario, this
could comprise of adjusting synthetic personalities following a user’s
feedback in preparation for their next use of ROBO-GUIDE.
5.1
Further Directions
The promising findings so far offer many directions to further exam-
ine user assistance in HRI.
The scenario chosen demonstrates a user assisting a robot at mini-
mal cost to themselves. While the explored factor of ambiguity is im-
portant in determining proactive assistive behaviour [9], relative cost
or risk to the individual is also a key factor [26]. The present work
indicates the two robot personalities both reduce ambiguity in inter-
action. Alternative scenarios (such as in manufacturing), in which
a user must expend greater effort to assist a robot, may further see
changes in the impact of ambiguity on a user’s willingness to assist.
A complex social interaction between user and robot, such as the
scenario presented, can be further explored across a variety of social
dimensions and factors not-yet considered. Empathy towards robots
in need or at risk [31] may contribute to user willingness to assist.
This may impact on the current pathways identified, particularly to-
wards a friendly robot or for individuals more likely to anthropomor-
phise the robot [27]. Alternatively, the interaction could be further
considered in a broader, social context, rather than the applied, occu-
pational psychology background for this workplace-inspired interac-
tion scenario and draw from related social cognition literature [13].
Further revisions to the scenario could include altering the robot’s
behavioural competency. It is possible that the robot performing ef-
fectively throughout the task (elevator operation not-withstanding)
places a ceiling on individuals’ trust towards the robot and limits the
impact of the experimental conditions. As identified in the literature
[14], errors from the robot impinge on user trust. Introductions of er-
rors (such as departing the elevator on the wrong floor) may lower
baseline user trust, enabling any potential influence from limitation
or intentional statements. This could further develop work examin-
ing the impact of robots addressing their mistakes on improving user
HRI experiences [34], again suggesting the importance of user liking
ahead of competency in social robots.
This study forms a solid foundation to conduct a field experiment
exploring user behaviour in the HRI scenario. Direct measures of
user behaviour, such as human-robot ‘interpersonal’ distance [35]
and user speed of response to robot requests, would build on the es-
tablished results of willingness to engage, liking of, and ambiguity
felt towards the social robot.
Our findings have implications for the development of social
robots and study of social HRI. The study demonstrates that individ-
uals may apply known human-human interpersonal social decisions
to social robots. The careful development of robot personalities to
account for this may be instrumental in fostering positive user expe-
riences of social robots and effective HRI.
ACKNOWLEDGEMENTS
The research team thank Man Tik Cheung (Hugo) for his contribu-
tion to the video materials used in the study. This work was sup-
ported by European Union Seventh Framework Programme (FP7-
ICT-2013-10) under grant agreement no. 611971 and the University
of Sheffield’s SURE scheme
REFERENCES
[1]
Christoph Bartneck, Dana Kuli´c, Elizabeth Croft, and Susana Zoghbi,
‘Measurement instruments for the anthropomorphism, animacy, like-
ability, perceived intelligence, and perceived safety of robots’, Interna-
tional journal of social robotics, 1(1), 71–81, (2009).
[2]
Andrea Bauer, Klaas Klasing, Georgios Lidoris, Quirin M¨uhlbauer,
Florian Rohrm¨uller, Stefan Sosnowski, Tingting Xu, Kolja K¨uhnlenz,
Dirk Wollherr, and Martin Buss, ‘The autonomous city explorer: To-
wards natural human-robot interaction in urban environments’, Inter-
national Journal of Social Robotics, 1(2), 127–140, (2009).
[3]
Cynthia Breazeal and Brian Scassellati, ‘How to build robots that make
friends and influence people’, in Intelligent Robots and Systems, 1999.
IROS’99. Proceedings. 1999 IEEE/RSJ International Conference on,
volume 2, pp. 858–863. IEEE, (1999).
[4]
David Cameron, Jonathan M Aitken, Emily C Collins, Luke Boor-
man, Adriel Chua, Samuel Fernando, Owen McAree, Uriel Martinez-
Hernandez, and James Law, ‘Framing factors: The importance of con-
text and the individual in understanding trust in human-robot interac-
tion’, in IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS), Workshop on Designing and Evaluating Social Robots
for Public Settings, (2015).
[5]
David Cameron, Emily C. Collins, Adriel Chua, Samuel Fernando,
Owen McAree, Uriel Martinez-Hernandez, Jonathan M. Aitken, Luke
Boorman, and James Law, ‘Help! I cant reach the buttons: Facilitating
helping behaviors towards robots’, in Biomimetic and Biohybrid Sys-
tems, Living Machines 2015, volume 9222 of Lecture Notes in Com-
puter Science, 354–358, (2015).
[6]
David Cameron, Samuel Fernando, Emily Collins, Abigail Millings,
Roger Moore, Amanda Sharkey, Vanessa Evers, and Tony Prescott,
‘Presence of life-like robot expressions influences childrens enjoyment
of human-robot interactions in the field’, in Fourth International Sym-
posium on New Frontiers in Human-Robot Interaction, 36–42, (2015).
[7]
Tiziana Casciaro and Miguel Sousa Lobo, ‘Competent jerks, lovable
fools, and the formation of social networks’, Harvard Business Review,
83(6), 92–99, (2005).
[8]
Raymond B Cattell, ‘The scree test for the number of factors’, Multi-
variate behavioral research, 1(2), 245–276, (1966).
[9]
Russell D Clark and Larry E Word, ‘Why don’t bystanders help? be-
cause of ambiguity?’, Journal of Personality and Social Psychology,
24(3), 392, (1972).
[10]
Emily C Collins, Abigail Millings, and Tony J Prescott, ‘Attachment to
assistive technology: A new conceptualisation’, in Proceedings of the
12th European AAATE Conference (Association for the Advancement
of Assistive Technology in Europe), (2013).
[11]
Munjal Desai, Kristen Stubbs, Aaron Steinfeld, and Holly Yanco, ‘Cre-
ating trustworthy robots: Lessons and inspirations from automated sys-
tems’, (2009).
[12]
Samuel Fernando, Emily C Collins, Armin Duff, Roger K Moore,
Paul FMJ Verschure, and Tony J Prescott, ‘Optimising robot personal-
ities for symbiotic interaction’, in Biomimetic and Biohybrid Systems,
392–395, Springer, (2014).
[13]
Susan T Fiske, Amy JC Cuddy, and Peter Glick, ‘Universal dimensions
of social cognition: Warmth and competence’, Trends in cognitive sci-
ences, 11(2), 77–83, (2007).
[14]
Peter A Hancock, Deborah R Billings, Kristin E Schaefer, Jessie YC
Chen, Ewart J De Visser, and Raja Parasuraman, ‘A meta-analysis of
factors affecting trust in human-robot interaction’, Human Factors: The
Journal of the Human Factors and Ergonomics Society, 53(5), 517–
527, (2011).
[15]
Guido Herrmann, Martin Pearson, Alexander Lenz, Paul Bremner,
Adam Spiers, and Ute Leonards, Social Robotics: 5th International
Conference, ICSR 2013, Bristol, UK, October 27-29, 2013, Proceed-
ings, volume 8239, Springer, 2013.
[16]
Pamela J Hinds, Teresa L Roberts, and Hank Jones, ‘Whose job is it
anyway? a study of human-robot interaction in a collaborative task’,
Human-Computer Interaction, 19(1), 151–181, (2004).
[17]
Jiun-Yin Jian, Ann M Bisantz, and Colin G Drury, ‘Foundations for an
empirically determined scale of trust in automated systems’, Interna-
tional Journal of Cognitive Ergonomics, 4(1), 53–71, (2000).
[18]
James Law, Jonathan M. Aitken, Luke Boorman, David Cameron,
Adriel Chua, Emily C. Collins, Samuel Fernando, Uriel Martinez-
Hernandez, and Owen McAree, ‘Robo-guide: Towards safe, reliable,
trustworthy, and natural behaviours in robotic assistants’, in Towards
Autonomous Robotic Systems (TAROS) 2015, volume 9287 of Lecture
Notes in Computer Science, 149–154, (2015).
[19]
John D Lee and Katrina A See, ‘Trust in automation: Designing for ap-
propriate reliance’, Human Factors: The Journal of the Human Factors
and Ergonomics Society, 46(1), 50–80, (2004).
[20]
Erika Mason, Anusha Nagabandi, Aaron Steinfeld, and Christian
Bruggeman, ‘Trust during robot-assisted navigation’, in 2013 AAAI
Spring Symposium Series, (2013).
[21]
Roger C Mayer, James H Davis, and F David Schoorman, ‘An integra-
tive model of organizational trust’, Academy of Management Review,
20(3), 709–734, (1995).
[22]
Daniel J McAllister, ‘Affect-and cognition-based trust as foundations
for interpersonal cooperation in organizations’, Academy of Manage-
ment Journal, 38(1), 24–59, (1995).
[23]
Owen McAree, Jonathan M Aitken, Luke Boorman, David Cameron,
Adriel Chua, Emily C Collins, Samuel Fernando, James Law, and Uriel
Martinez-Hernandez, ‘Floor determination in the operation of a lift by
a mobile guide robot’, in Proceedings of the European Conference on
Mobile Robotics, pp. 1–6, (2015).
[24]
Carolyn Y Nicholson, Larry D Compeau, and Rajesh Sethi, ‘The role
of interpersonal liking in building trust in long-term channel relation-
ships’, Journal of the Academy of Marketing Science, 29(1), 3–15,
(2001).
[25]
Scott Ososky, David Schuster, Elizabeth Phillips, and Florian G
Jentsch, ‘Building appropriate trust in human-robot teams’, in 2013
AAAI Spring Symposium Series, (2013).
[26]
Irving M Piliavin, Jane A Piliavin, and Judith Rodin, ‘Costs, diffusion,
and the stigmatized victim.’, (1975).
[27]
Laurel D Riek, Tal-Chen Rabinowitch, Bhismadev Chakrabarti, and Pe-
ter Robinson, ‘How anthropomorphism affects empathy toward robots’,
in Proceedings of the 4th ACM/IEEE international conference on Hu-
man robot interaction, pp. 245–246. ACM, (2009).
[28]
Paul Robinette, Alan R Wagner, and Ayanna M Howard, ‘Building and
maintaining trust between humans and guidance robots in an emer-
gency’, (2013).
[29]
Stephanie Rosenthal, Joydeep Biswas, and Manuela Veloso, ‘An ef-
fective personal mobile robot agent through symbiotic human-robot
interaction’, in Proceedings of the 9th International Conference on
Autonomous Agents and Multiagent Systems: volume 1-Volume 1, pp.
915–922. International Foundation for Autonomous Agents and Multi-
agent Systems, (2010).
[30]
Stephanie Rosenthal and Manuela M Veloso, ‘Mobile robot planning
to seek help with spatially-situated tasks.’, in AAAI, volume 4, p. 1,
(2012).
[31]
Astrid Marieke Rosenthal-von der P¨utten, Frank P Schulte, Sab-
rina C Eimler, Laura Hoffmann, Sabrina Sobieraj, Stefan Maderwald,
Nicole C Kr¨amer, and Matthias Brand, ‘Neural correlates of empathy
towards robots’, in Proceedings of the 8th ACM/IEEE international
conference on Human-robot interaction, pp. 215–216. IEEE Press,
(2013).
[32]
Mark A Serva, Mark A Fuller, and Roger C Mayer, ‘The reciprocal
nature of trust: A longitudinal study of interacting teams’, Journal of
Organizational Behavior, 26(6), 625–648, (2005).
[33]
Ramadhar Singh and Xue Ling Tor, ‘The relative effects of competence
and likability on interpersonal attraction’, The Journal of social psy-
chology, 148(2), 253–256, (2008).
[34]
DHJ Snijders, ‘Robot’s recovery from invading personal space’, in 23rd
Twente Student Conference on IT, volume 23. University of Twente,
(2015).
[35]
Michael
L
Walters,
Kerstin
Dautenhahn,
Ren´e
Te
Boekhorst,
Kheng Lee Koay, Dag Sverre Syrdal, and Chrystopher L Nehaniv, ‘An
empirical framework for human-robot proxemics’, Procs of New Fron-
tiers in Human-Robot Interaction, (2009).