S.S. Ge et al. (Eds.): ICSR 2012, LNAI 7621, pp. 199–208, 2012.
© Springer-Verlag Berlin Heidelberg 2012
Anthropomorphism and Human Likeness in the Design
of Robots and Human-Robot Interaction
Julia Fink
CRAFT, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland
julia.fink@epfl.ch
Abstract. In this literature review we explain anthropomorphism and its role in
the design of socially interactive robots and human-robot interaction. We illu-
strate the social phenomenon of anthropomorphism which describes people’s
tendency to attribute lifelike qualities to objects and other non lifelike artifacts.
We present theoretical backgrounds from social sciences, and integrate related
work from robotics research, including results from experiments with social ro-
bots. We present different approaches for anthropomorphic and humanlike form
in a robot’s design related to its physical shape, its behavior, and its interaction
with humans. This review provides a comprehensive understanding of anthro-
pomorphism in robotics, collects and reports relevant references, and gives an
outlook on anthropomorphic human-robot interaction.
Keywords: anthropomorphism, design, human-robot interaction, literature re-
view, social robots, social factors in robotics.
1 Anthropomorphism and the Role of Anthropomorphic Design
Soon more and more robots will be used in everyday environments, and an important
aspect of developing “socially interactive robots” [1] is the design for effective hu-
man-robot interaction (HRI) as well as acceptance. One approach to enhance people’s
acceptance of robots is the attempt to increase a robot’s familiarity by using anthro-
pomorphic (humanlike) design and “human social” characteristics. This implies hu-
manlike parts of a robot’s physical shape, the usage of facial expressions and other
social cues, as well as natural humanlike interaction and communication (e.g. speech,
gaze, gestures). However, the role of anthropomorphism in robotics is not to build an
artificial human but rather to take advantage of it as a mechanism through which so-
cial interaction can be facilitated [2]. An underlying assumption is that humans prefer
to interact with machines in the same way that they interact with other people [1]. The
idea combines “anthropomorphic design” and the phenomenon of “anthropomor-
phism” – when people attribute human characteristics to objects. Researchers have
found that whenever artifacts show intentional behavior (e.g. when animated), people
tend to perceive them as characters or even as creatures [3] [4].
1.1 Anthropomorphism
“Anthropomorphism” originates from the Greek “anthropos” for “human” and
“morphe” for “shape” or “form” [2]. It describes people’s tendency to attribute
200 J. Fink
human characteristics to non-lifelike artifacts. The phenomenon of ascribing inten-
tions [5] and animacy to simple shapes based on motion has been intensively stu-
died in (developmental) psychology. But why do humans ascribe intentions and
emotions to objects? One interpretation is that attributing familiar humanlike
qualities to a less familiar non-humanlike entity can serve to make the entity be-
come more familiar, explainable, or predictable [6]. In the design of socially inter-
active robots [1], anthropomorphism plays an important role and is reflected in the
robot’s form (appearance), behavior (e.g. motion), and interaction (e.g. modality).
Robotics uses the mechanism to increase acceptance of robots and facilitate inte-
raction.
1.2 Anthropomorphic Forms in Robot-Design: Shape, Behavior, Interaction
Anthropomorphic design means an imitation of human (or natural) form [7]. Fong et
al. classify four categories of a robot’s aesthetic form: anthropomorphic, zoomorphic,
caricatured, and functional [1]. In robotics, “anthropomorphic design” refers to three
parts: a robot’s shape, behavior, and interaction/communication with the human [8]
[9]. Social robots make further use of “human social” characteristics, such as ex-
press/perceive emotions, communicate with high-level dialogue, learn/recognize
models of other agents, establish/maintain social relationships, use natural cues (gaze,
gestures, etc.), exhibit distinctive personality and character, learn/develop social com-
petencies [1]. One may ask how much human-likeness we want to have in non-human
objects. How will people react to a robot that resembles a human? In 1970, Mashiro
Mori formulated a theory called the “uncanny valley” [10]. It describes people’s reac-
tions to technologies that resemble a human too close while still not being one. Mori
hypothesized that a person’s response to a humanlike robot would abruptly shift
from empathy to revulsion as it approached, but failed to attain, a lifelike appearance
[10].
Fig. 1. Examples for bio-/anthropomorphic robots; top row: AIBO, Pleo, Paro, iCat, Papero; bot-
tom row: Kaspa
r
, NAO, Nexi, Barthoc, iCub, NAO
Anthropomorphism and Human Likeness in the Design of Robots 201
1.3 Why Is Anthropomorphism Relevant for (Social) Robotics?
What brings together anthropomorphic design and social robotics is the fact, that the
appearance and function of a product impacts how people perceive it, interact with it,
and build long-term relationships with it [11]. On one hand, robots with humanlike
design cues can elicit social responses from humans which in turn can have a positive
impact on acceptance [12] [13] [14]. People responded more positively to an artifact
that displayed humanlike behavioral characteristics (emotions, facial expression) in
contrast to a purely functional design [3] [15] [16] [17]. However, user preferences
were task and context dependent [18]. Thus, the appearance of a robot should match
its capabilities as well as the users’ expectations [13] [19]. Anthropomorphizing a
technological agent appears to create some social connection to it, aids in learning
how to use it [6], and how pleasant and usable it is perceived [20] [16]. People pre-
ferred to collaborate with a robot that was able to respond socially [14] [18]. On the
other hand, robots that overuse anthropomorphic form, such as humanoids that almost
perfectly resemble a human but still remain unnatural copies, can have a contrary
effect and evoke fear or rejection [10]. Though the point of when this negative effect
can be observed is not yet identified, studies showed that especially humanoid robots
evoked more reluctant and negative responses than robots with a pet-like or more
functional shape [21]. Interestingly, the phenomenon seems to be culture sensitive
[22] and based on Epley et al.’s psychological determinants likely to be related to
other person-related factors, such as expertise/experience with a system [23] [24].
2 The Social Phenomenon and Socially Interactive Robots
2.1 Explaining the Social Phenomenon of Anthropomorphism
According to [25], there are two main perspectives when seeking to explain people’s
tendency to anthropomorphize artifacts. First one explains anthropomorphism from
the design of the artifact. It is assumed that humans directly respond to life-like or
social cues that an object or system emits, without thoughtful mental processing, by
simply applying stereotypes and heuristics to it. Schmitz [26] describes that within the
visual scope for design, the outer appearance can have an important impact on the
overall perception of an object. If this explanation of anthropomorphism is correct,
people may respond automatically to social cues emitted by a robot, and apply hu-
man-human social schemas and norms to these interactions [25].
A second explanation applies a human-centered, cognitive viewpoint where anth-
ropomorphism is described through people’s specific mental model [25] they have
about how an artifact works the way it does. If a system behaves much like a human
being (e.g. emits a human voice), people’s mental model of the system’s behavior
may approach their mental model of humans, but this model may differ in important
respects from their models of humans [25]. People’s estimation of a robot’s “know-
ledge” and its capabilities/abilities affects the way they relate to it. Research ex-
amined the validity of the mental model concept with various kinds of robots [25]
[27]. Findings suggest that people tend to hold richer mental models about anthropo-
morphic robots in contrast to mechanic ones [27] .
202 J. Fink
As an alternative to the two explanations given above, one can explain people’s
tendency to attribute human qualities to objects based on social psychology. As men-
tioned earlier, Epley et al. [6] established a three-factor theory of when people are
likely to anthropomorphize based on psychological determinants. Namely, the theory
describes that some people are more likely to anthropomorphize, so when (i) anthro-
pocentric knowledge is accessible and applicable to the artifact (elicited agent know-
ledge), (ii) they are motivated to explain and understand the behavior of other agents
(effectance motivation), and (iii) they have the desire for social contact and affiliation
(social motivation) [6]. Some work also discusses the inverse process to humanizing
artifacts, namely, dehumanization [6], or mechanomorphism [23].
2.2 Classification and Evaluation of Social Robots
Socially interactive robots can be classified in terms of (1) how well the robot can
support the social model that is ascribed to it and (2) the complexity of the interaction
scenario that can be supported [1]. Breazeal [28] and later extended by Fong et al. [1]
suggest seven classes of social robots: socially evocative, social interface, socially
receptive, sociable, socially situated, socially embedded, socially intelligent (for more
details, see Fong et al. [1]). This classification is based on Dautenhahn and Billard’s
[29] definition of social robots, as “embodied agents that are part of a heterogeneous
group: a society of robots or humans. They are able to recognize each other and en-
gage in social interactions, they possess histories (perceive and interpret the world in
terms of their own experience), and they explicitly communicate with and learn from
each other.” Since the time when the Turing Test was drafted, one of the benchmarks
for success in AI and HRI has been how well a system can imitate human behavior.
Several measurements and methods have been suggested for the evaluation of anthro-
pomorphic robots: psychological benchmarks [30], as well as properties of a social
robot rated by humans [19] [31]. From a methodological point of view, questionnaires
and content analyses [32] [33] have been used to analyze anthropomorphism in robot-
ics but also more implicit measures (e.g. psychophysical onses), such as gaze cues
[34], motor/perceptual resonance [35], and neurologic metrics [4].
3 How Anthropomorphism Impacts Human-Robot Interaction
3.1 Impacts of Anthropomorphic Shape of a Robot
A robot’s physical embodiment is one of the most obvious and unique attributes and
thus of high importance for interaction. The role of the physically visible design of
robotic products has been discussed and investigated by designers [8] [19] [36]. HRI
studies have so far verified that there are differences in how people interact with
anthropomorphic and non-anthropomorphic robots [37] especially in terms of social
interaction. However, while no real evidence exists, theory suggests a negative corre-
lation between the robot’s physical realism and its effectiveness in HRI [37]. A hu-
man shaped robot can raise specific expectations from the user side [25], which can
lead to a negative effect when the robot’s behavior does not meet these expectations.
Anthropomorphism and Human Likeness in the Design of Robots 203
In general, many studies so far, contribute (partly) to the “uncanny valley” effect,
however, one has to take a more detailed look at which dimensions of the interaction
are affected: Kanda et al. conducted a study with two different humanoid robots
(ASIMO and Robovie) and showed that different appearance did not affect the partic-
ipants’ verbal behavior toward the robot but did affect their non-verbal behavior such
as distance and delay of response [38]. Similarly, comparing a pet-robot (AIBO) to a
humanoid robot (ASIMO), people seem to prefer the pet-shaped robot [21]. While
there was no significant difference in how people gave verbal commands to both ro-
bots, the way participants gave positive and negative feedback to AIBO and ASIMO
differed significantly [21]. While AIBO was treated similarly to a real dog and petted
to give positive feedback, the humanoid ASIMO was touched far less [21].
In evaluating how humanlike a robot appears, especially a robot’s head and face
receives considerable attention, since this body part is crucial in human-human com-
munication (most non-verbal cues are mediated through the face). DiSalvo et al. [8]
found that particularly the nose, the eyelids and the mouth increase the perception of
humanness in robotic heads. Further, the width of the head had a significant effect.
Also, a robot’s physical embodiment and presence has been investigated in terms of
anthropomorphic interactions compared to robot-like agents or a remote robot [39]
[40]. Kiesler et al. [39] conducted a study where a robot-like agent interviewed partic-
ipants about their health. People were either present with the robot/agent, or interacted
remotely with it, projected life-size on a screen. Results indicated that participants
were more engaged, disclosed less undesirable behavior, and forgot more with the
robot versus the agent [39]. People viewed the robot as more dominant, trustworthy,
sociable, responsive, competent, and respectful than the agent and rated it more life-
like. The collocated robot was anthropomorphized the most [39].
In conclusion, studies suggest a positive effect of embodied robots that use anthro-
pomorphic shape. However, there is the tendency that participants prefer a pet-shaped
robot to a human-shaped robot. Overall, research confirmed that the physical shape of
a robot strongly influences how people perceive it and interact with it, thus visible
design is crucial. However, demographic, cultural factors [22] [41], individual prefe-
rences, and the context of use need to be considered as well. This makes it hard to
identify concrete universal guidelines for how to design an acceptable social robot.
3.2 Impacts of Robots Using Human Social Cues / Social Interaction
Besides the shape, a robot’s effectiveness in HRI is also related to its behavioral so-
cial success which is a fundamental component of the interaction. Studies showed that
the social identity of the robot (both the personality and the role of the robot) [37] has
an effect on the user’s task performance. The use of social interaction in HRI is ex-
pected to make the interaction more natural and thus more effective. Efforts have
been made in making a robot’s behavior social by giving it a personality, letting it
display facial expressions, making it communicate in a polite way, or even making it
cheat [42], for example. Also the ability of recognizing and being aware of the human
counterpart’s emotional state was used as one possibility for socially intelligent ma-
chines. In the following we present results of studies with robots that used human
social cues to interact with people and outline how this affected the interaction.
