Steve Weymouth – Good Robot, Bad Robot: Design for Social Robots Based on the Animation Principle of ‘Appeal’

Social robotics is a relatively new field of research that aims to integrate traditional robotics into human and public spaces. Social robots are designed to interact with humans as complementary agents to enhance human activity. Declining production costs means that sales for personal robots will increase over time, with robots having a greater reach into our daily lives. A 2016 report from the financial firm KPMG illustrated the increasing interest and decreasing costs (KPMG 2016). For example, the average cost of sensors more than halved between 2004 and 2016 (and is forecast to halve again in three year’s time), while average investments in robotics have increased dramatically. Investments started to rise in 2014, with 2015 showing almost a threefold increase on the previous year. Further, KPMG (2016) predicted that social robots will be an increasing part of our daily lives, moving from perceptive navigators with the basic interaction of a vacuum cleaner to affective and companionable learning with enough physical dexterity to be a sports coach. Social robots are predicted to play a greater role in the service industries by tutoring, helping with autistic children and in old age health care. The so-called caring industries are important testing grounds for the social acceptance of robots into people’s lives. Artist and robot researcher Petra Gemeinboeck stated that “he stakes for developing a better understanding of how to design socially competent machines are (therefore) high” (Gemeinboeck 2017). Techniques for movement generation in human–robot interaction (HRI) research include motion capture, puppeteering, animation and procedural software techniques that blend movements together. These approaches are exemplified in the work of Japanese roboticist Hiroshi Ishiguro, who created a self-likeness robot (Geminoid H1-1) and a female counterpart (Geminoid F) (Advanced Telecommunications Research Institute International 2017), as well as the work of David Hanson and Hanson Robotics (2017). Ishiguro uses motion capture to create robot’s facial expressions, while Hanson uses his background in sculpture and animation to inform the look and movement of robot’s faces. Both approaches focus on complex movement for the face and head, while movement in the body is more machine-like and is limited to the hands and arms. Both approaches push the boundaries in life-like robot design; however, attempts to replicate humans in look and movement tend to be cumbersome and disappointing, or at worst disturbing, as many discussions around Masahiro Mori’s Uncanny Valley (1970) and robotics attest. The above examples also fall into cliché role identification, presenting masculine or feminine social and sexual stereotypes.

Achieving human-like robots is technically challenging and extremely expensive. Socio-stereotypical roles of gender and labour are often reinforced for the elite few who can afford them. Historically, design for robots that have any form of interaction with humans has come from the fields of science, technology, engineering and mathematics, or STEM. Tracking from the 1970s we can see that the STEM fields were not only dominated by men but men whose attitudes to social roles were also influenced, and supported, by the predominate patriarchal systems of the past forty years. Over this same time, robot design has also tended toward the anthropomorphic or zoomorphic showing our human need for the familiar. It is only the industrial robot where function overcomes such design concerns opting instead for operability and price as the guiding principle. Aesthetics, behaviour and function are important considerations in designing social robots; however, if complex and meaningful interactions between machines and humans are to be achieved, a more imaginative approach is needed. New approaches are being explored in design and function based on social signals in movement and appearance informed by animation thinking.

Recent production of social robots has engaged with simplification in design by trading high-level functionality for engagement and interactivity. The new approach contrasts with the more traditional pursuits of the military or high-end computing research for artificial intelligence (AI). In 2016, new social robotics companies such as Jibo and Anki raised $157.5 million and $70.4 million respectively from investors to expand production. More established companies such as Hanson Robotics and Blue Frog are introducing companion robots for the home. Further, LG and Bosch have entered the market with offerings such as Hub and Mykie. Clearly, social robotics has a future where creative individuals can create traction in this space. With the aid of three-dimensional (3D) printing and rapid prototyping, off-the-shelf devices such as the Arduino operating system, artists and creative designers can develop thought-provoking and political responses to the market.

In this vein, manufacturers of social robots have recently turned to the animation industry for assistance with designs that enhance engagement. Animation approaches such as storyboarding, character design and the Disney twelve principles of (1995), provided new approaches to appearance and communication based on engagement. In 2016, game and robot company Anki employed Pixar animator Carlos Baena (2017) to help create palm-sized Cozmo (Anki 2017). Cozmo is a small game-playing interactive toy that moves around on tracktor-wheels, with large, expressive eyes based on the Pixar character Wall-E. Cosmo engages in play via a smartphone app and exhibits social signals in a range of expressions via screen-based and animated eye shapes. Como’s physical movements are based on a simple set of action and reaction responses.

Jibo was launched in 2014 and marketed as “the world’s first social robot” (Breazeal 2014) in a start-up that raised more than US$3 million. Jibo resembles a kitchen appliance more than an anthropomorphic robot. It interacts through verbal communication via smartphone app, and it has bodily movement but limited locomotion because it is designed to sit on a benchtop or table. Articulation is achieved on a simple three axes of rotation (or degrees of freedom [DoF]), which allows for an extensive range of physical actions such as turning its face towards sounds and people. Clear movement signals articulate surprise, interest and curiosity by employing some of the Disney principles of animation such as anticipation, overlap and follow through.

The Disney animators understood the power of appeal, which is the twelfth and final principle of animation according to Thomas and Johnston (Thomas 1981, p. 68). As appeal is last on the list, many believe that it is less important. Indeed, the concept of appeal is difficult to pin down because it is the most subjective of the 12 principles. Nonetheless, appeal is an important concept in the relationship between animation and social robotics. In animation, appeal can relate to the personality of a character or the way it moves, as well as design aspects and appearance. Etymologically, appeal has Latin roots, including the adjectives to drive and to address. Contemporary English usage includes legal and sporting notions of appeal, while the adjective form, appealing, relates to attractiveness and interest. The Merriam–Webster dictionary defines appealing as “having qualities that people like: pleasing or attractive” (Merriam–Webster 2017). In relation to the twelfth animation principle of appeal, attractive means to attract by engaging attention. In animation, appeal can relate to interesting characters, regardless of whether they are cute and lovable or hideous and evil; that is, interest attracts and holds attention. Disney animator Ethan Hurd (IMDb 2017) singled out appeal as carrying the most weight of the 12 Disney principles. He stated that appeal should be considered the culmination of the first 11 principles, “plus something extra” (Hurd 2017), which he identified as order and interest. Order is the pattern of logic in an animator’s considerations of movement, design and the sequence of events, whereas interest relates to the choices made “to keep the order from being boring” (Hurd 2017). If animation can encompass recognisable characters and anthropomorphised objects, as well as geometric shapes, simple scratches and lines, the viewer’s attention will be attracted by the order and interest in the movement, the look or design and the behaviour represented by the object, objects or thing. Although human relationships to things touches on the notion of animism or Jane Bennett’s “thing power” (Bennett 2009), appeal can then be identified as the order and interest created in the movement, appearance and behaviour of the thing with which we interact.
To consider the level of engagement afforded by appeal in this context, we need to consider the spectrum of development in robots to date. Currently, physical robots, such as those used on the production line or in warfare, often have an anthropomorphic appearance and machine-like movement that tend to alienate or disturb (Patel 2015). Human-like designs for robots with movement based on motion capture can create higher user expectations that often result in feelings of disappointment or otherness (Saenz 2010). Shifting the focus from machine and mechanical limitations to considerations based on animation thinking can help create more robust human-to-robot and robot-to-human interactions. Current research in social robotics relies heavily on the creation of anthropomorphic and zoomorphic design to create engagement. Robots that look and behave similar to humans and pets appear to be more readily accepted. However, this approach creates a high degree of expectation that robots will behave and move like those we know. Current research into interactive robots also relies on historical and science fiction notions of the robot as servant (or weapon), while notions of gender are similarly retrogressive.

From Industrial Robots to Artist’s Responses

HRI has rapidly moved from science fiction and fantasy to reality. The focus on current HRI has given more weight to the development of AI over thoughtful design for engagement based on behaviour, appearance and movement. Many approaches must contend with robotic movement generated through efficient (or inefficient) machine design processes that start with raw functionality and end with movement limitations.

One such outcome of machine efficacy is the industrial robot. Although it was never intended for robot-to-human interaction, the industrial robot has increasingly replaced human workers on the production line and the factory floor. Robots assemble cars and sort merchandise held in vast warehouses. They retrieve, reorder replacement stock, package, ship and deliver goods ordered online with minimal involvement from humans. Industrial robots do not need to be friendly or receive input from human voices because most of their conversations are in line code that efficiently operates mechanised movement based on shortest-route linear trajectories that have little to do with kinetic-based human communication. Industrial robots are employed to perform repetitive tasks that would otherwise require a high degree of precision and attention. For safety reasons, they are often separated from humans by exclusion zones and physical barriers. However, artists see other possibilities for technology that extend beyond robot’s intended use. Dancers and choreographers have relished the opportunity to put soft-fleshed human bodies within reach of robots that are typically tasked with welding, cutting and assembling at speeds much greater than their human counterparts.

Dr Margo Apostolos experimented with a PUMA industrial robot arm to explore ways to make machine movements more aesthetic and graceful within the limits of PUMA’s three axes of rotation in the arm (DoF). By identifying machine movement as linear and more human-like movement as curvilinear, Apostolos identified ways to make machine movement more expressive. In 1978, Ballet for one arm was programmed to use graceful movements, as demonstrated at the Stanford University Rehabilitation Research & Development Center, VA, Palo Alto. In 1983, Apostolos created three more choreographed performances using a PUMA 260 industrial robot. Apostolos is now a professor of dance at the University of Southern California Glorya Kaufman School of Dance and is investigating further potential between the efficient, artistic and effortless movement of industrial robots and new designs for movement in robots to best serve people with severe disabilities.

Margie Medlin is an Australian choreographer and media artist who works at the Laboratory at Südpol in Lucerne, Switzerland. In 2007, Medlin collaborated with fellow Australian Gerald Thompson to create a purpose-built robot for a human and robot dance piece called Quartet (2007), which was described as “a project to develop a real-time interactive robot to perform live on stage with a dancer and a musician” (Thompson 2017). Medlin wanted a real-time link between the dancer and robot to record the dancer’s movements and enable recording and playback on the robot at a later stage. The aim was to create movement that evoked a sense of life in the robot based on the graceful flow of the dancer. Although the robot in this case was not an autonomous or even semi-autonomous device, the project nonetheless formed a bridge between the aesthetic values in the movements of a human dancer and the movement desired and represented in the robot dancer to enhance the appeal of a mechanical device. Medlin’s earlier approach in sourcing aesthetic and meaningful movement from dancers and applying it to robots was reflected and extended in the more recent research of collaborators Petra Gemeinboeck and Rob Saunders.
Gemeinboeck and Saunders founded Machine Movement Labs (MML; De Quincey Co 2017), which is a collaborative project that seeks to develop new design methods for unique and abstract robot forms. Working with engineers and choreographers, MML uses professional dancers to interpret constrained movement parameters in geometric prototypes for robots to explore new ways of helping robots to learn meaningful and expressive modes of movement (De Quincey Co 2017). Rather than pursue anthropomorphic designs and be constrained to anthropomorphic behaviour, MML works with abstract shapes to create new movement languages that engender affect and empathy (De Quincey Co 2017). Using “Performative Body-Mapping” (Gemeinboeck & Saunders 2015), Gemeinboeck and Saunders aim to harness the dancer’s creativity and sensitivity to develop movement characteristics for strange, abstract shapes. They argued that “movement can provide a key to socializing non-anthropomorphic robots” (Gemeinboeck & Saunders 2015). In relation to animation, Gemeinboeck and Saunders pointed out that animators have a history of bringing life to non-human and abstract shapes. Further, they acknowledged John Lasseter (Lasseter 2001) and noted that it is not the story or narrative that brings out character and personality, but the timing and construction of movement.

The three approaches outlined above follow the same trajectory of seeking to inform robotic movement with the movement of dancers and thinking, as revealed through the choreography of the human body. A number of other performers have experimented with dancing with industrial robots, including Actuator by Finnish choreographer Thomas Freundlich in 2008 and KUKA by Huang Yi in 2012. The investigation of aesthetic junctures between dancers and robots contributes to the potential of appeal in the movement of social robots.

Dance, Neuroscience and Cognitive Thinking

Animators and dancers both consider the potential of movement to communicate complex emotional information through the poetry of movement itself. In particular, dance and animation thinking helps to inform more effective, and affective, methods of communication between humans and robots. Throughout history, many artists have been influenced and inspired by dance, including Matisse, Manet, Degas, Sargent and Titian. More recently, Céline Dion, Lady Gaga and Beyoncé stated that Michael Jackson’s dancing style has had a significant influence on them. Thus, the designer of social robots should also be inspired by dance. Dance is about performance and the embodiment of movement, as well as engagement with and enjoyment by an audience that mostly remains seated and physically inactive throughout the performance. Sitting (or standing) in a stationary position can be thought of as paying attention and being attracted and engaged. However, when people watch dance and animation or engage with a social robot, they engage their cognitive embodiment. Physically, they may appear to be relatively relaxed, but on the inside, there is quite a lot going on. Neuroscience has helped to uncover the internal workings of perception while paying attention to the external world. Although few studies have been conducted in regard to animation, neuroscience for motion perception and the watching of dance can provide an understanding of the powerful communication potential for movement in social robots.

Between 2008 and 2011, the University of Manchester, University of Glasgow, York St John University and Imperial College London undertook a collaborative research project entitled Watching Dance. The core research examined spectator’s responses to dance using transcranial magnetic stimulation (which measures the extent to which neural pathways from the brain to the muscles are primed for action) and functional magnetic resonance imaging (fMRI, which shows active brain regions). Their findings made broad use of the term “kinaesthetic empathy.”

The notion of kinaesthetic empathy is similar to Theodor Lipps’s aesthetic sympathy (Lipps 1903), which represents the sharing of feelings between a viewer and the object of their contemplation (in Lipps’s case, works of art). The focus on creative practice is invested in a belief derived from the Watching Dance project that “art is that which expresses feeling” (Reynolds & Reason 2012, p. 11). The term “kinaesthetic empathy” combines thinking and meaning from two uncertain terms. Kinaesthesia refers to the sensation of movement, which in turn is related to proprioception, or awareness of the location of one’s body parts without the aid of visual reference. Both have roots in Merleau-Ponty’s (1962) concepts of perception and phenomenology. Empathy is perhaps the more problematic of the two; it is often thought of as the act of projecting one’s self onto or into objects, objects of contemplation or others. However, empathy also affects inner sensations by mirroring or mapping externally perceived movement onto the viewer. Lipps suggested that animation resonates with the notion that artists communicate by leaving a trace through their creative objects. The effect of kinaesthetic empathy correlates with the communication potential of animated movement, and therefore the power of communication through movement.

According to Reynolds and Reason (2012), Paterson (2012, p. 492) and Parmenter (2013, p. 155), New York Times dance critic John Martin first attempted to define the passive act of watching dance as an internal and bodily sensation. He described the pleasure that audience members experience while watching dance as a feeling of vicarious physical enjoyment that “present muscular experience and awaken such associational connotations as might have been ours if the original movement had been of our own making” (Martin 1936, p. 117). Martin identified that the audience may look outwardly passive, but internally, their body and mind were moving in sympathy with the dance. However, he believed that the dancer also led the audience’s emotions—that it was ‘the dancer’s whole function to lead us … in order that we may experience (the dancer’s) feelings’ (Martin 1939, p. 52). Martin missed that humans bring emotional information to all of their experiences, and they are not universal in the way they respond to any given experience. Human emotions are a result of developmental growth, which differs from person to person. Humans internalise movement—particularly the movement of other people—to comprehend and predict what those people will do. Further, they interpret movement based on learned behaviour and cognitive abilities. Nonetheless, Martin was on the right track, and many researchers have identified the inner workings that humans experience while watching the movement of life around them.

Catherine Reed is a cognitive neuroscientist who works at Claremont McKenna College in California, where she explores a range of topics from a cognitive neuroscience perspective, including sensory and perceptual processing, selective attention and emotional perception. Reed (2012) argued that “the interface of visual-motor processes between the observer and observed others depends upon both innate and learned mechanisms” (p. 44). The human body in motion is one of the most frequently witnessed stimuli in people’s environments. It is also a unique experiential phenomenon because humans can perceive their body from the outside and the inside at the same time. Given that emotional and physical information can be assessed from a distance without the need to see facial expressions, the body in motion is an immediate and accessible source of sociological and emotional information based on movement. Reed referred to the comparisons that people make between themselves and other bodies in motion as “self-other correspondences” (Reed 2012, p. 44).

Thus, the perception of biological motion involves subconscious physical and emotional memory. Therefore, empathy is neither the act of putting ourselves into another’s shoes nor the act of projecting one’s self onto another. Rather, it involves internalising other’s actions, mapping them onto our neural motor and memory systems, and referencing individually formed emotional experiences to understand. Extending the notion of self-other correspondence to the perception of other animate objects, such as those represented in the appeal of movement (whether animation or robot), provides a deeper understanding of appeal in social robot design. Activating internal mimesis through movement based on social signals and gestures can help to engage people’s interest and hold their attention.

Contemporary Social Robots: Marketing the New Appeal

Social robots are being marketed at an increasing rate; however, in the race to dominate the market, many companies are attempting to identify the most marketable aspects of these robots (Ackerman 2017). For aesthetic purposes, many of the new releases have some similarities, such as a smooth surface, white and minimalistic design, and roughly the size (and appearance) of a large kitchen appliance with animated eyes. Many have a head with a degree or three of freedom that can swivel around when listening to voice commands or indicating the direction of attention.

In 2016 Bosch funded a new start-up company called Mayfield Robotics, which “make helpful home robots” (Mayfield Robotics 2017) and have recently launched Kuri. Kuri’s appearance is similar to the sleek lines and high-design of Eve in Wall-E, but it is slightly more overweight and grounded. Kuri can move around a house on Roomba-like wheels and perform activities such as shooing a dog off a couch. Kuri can understand voice commands, but it does not talk; instead, it responds with three different modulated sounds. Mayfield Robotics aimed to create a mobile robot that would be readily accepted into the home; thus, Kuri had a rounded design and blinking eyes that covered two internal high-definition cameras. Physical indicators for actions are critical in mobile robots. To help Kuri develop clear indications of intention and attention, Mayfield Robotics employed Pixar animator Doug Dooley. A robot navigating around a house needs to indicate intention of direction to human and animal inhabitants to avoid collisions. Using animation principles to help create smooth and unambiguous actions, Dooley incorporated leading actions into the eyes and head to show thought and consideration before movement began. Animation thinking also helped to make Kuri more life-like by adding reactions to tasks or events, because “showing a response to the task outcome can make the robot seem to be more intelligent/capable, even if it fails to achieve the functional task at hand” (Takayama, Dooley & Ju 2011, p. 76). Reactions such as surprise and disappointment helped to make Kuri’s behaviour more appealing and engaging. Dooley’s Pixar background enabled him to create prototypes of many movement behaviours on a computer. This prototyping method is frequently used in animation production and is useful for refining movement based on engagement and appeal before any economic cost is incurred.

In addition to Kuri, three other social robots for home use have been released in 2017: LG’s Hub, Bosch’s Mykie and Jibo. In contrast to Kuri, these three robots are immobile and are designed to interact with members of the household by being carried around the house. Even so, all three have mobile heads that can communicate a level of attention by swivelling on a limited number of DoF. To varying degrees, they all show evidence of animation thinking in the movement of their head and eyes (in the case of Jibo, a single cyclopoid eye). While Kuri has physical eyes, the other three have screen-based eyes that move with anticipation, and they can squash and stretch. Hub and Jibo are the most life-like because they have more DoF, which allows movement to be executed on curvilinear paths, whereas Mykie maintains its robotic linear movement, and its actions are limited to swivelling its head from left to right.

Guy Hoffman is the Assistant Professor and the Mills Family Faculty Fellow at the Sibley School of Mechanical and Aerospace Engineering at Cornell University. He has created compelling and engaging movement for robots based on animation thinking since seeing John Lasseter’s Luxo Jr. (1986). Hoffman’s experiences in learning animation helped him to realise the valuable contributions that animation thinking could make to social robotics. In 2012, Hoffman and Ju published a paper on thinking in movement informed by animation; however, the present article focuses on the combined thinking for appeal as the sustained connection between behaviour, appearance and movement to keep the appeal from becoming boring. With this in mind, it is worth analysing Hoffman’s approach.
Hoffman has researched HRI and interaction fluency based on non-anthropomorphic, non-verbal social robot design. He has developed several working prototypes of non-verbal communication potentials for social robots. Most notably, Shimon, an improvising marimba player, has a head that moves on:

four chained movement DoFs: a base pan moving the whole head left and right, a base tilt roughly 40 percent up from the base allowing for a bowing motion and the 40-degree coupled head pan-tilt mechanism. In addition, a servomotor controls both upper and lower shutter to support the opening and closing of the head. The head also contains a single high-definition digital video camera. (Hoffman & Ju 2012)

The DoF allows Shimon’s head to move up and down in time to the music and direct attention towards fellow human musicians. The single blinking eye (a pincher-action horizontal slit) can watch human musicians and react to physical cues. Shimon can play music with humans and improvise in a naturalistic way based on inbuilt algorithms. The non-anthropomorphic design and simple mobility enriched by animation thinking has formed the basis for all of Hoffman’s robot designs.

Another robot of note is Hoffman’s Travis, which is a speaker-bot roughly 30 cm high that plays music from a smartphone and reacts by bobbing its head in time with the music. Travis has a similar mechanism to Shimon that allows its head (which contains the speakers) to move in complex ways based on the three DoF based on animation thinking. Hoffman’s robots have visual roots in Luxo Jr., which had the same number of DoF. Hoffman intentionally shifted his focus to well-designed movement and non-anthropomorphic visual design to create more opportunities for affective and appealing interaction: “Movement can give interactive robots a larger dynamic range of expression and enables spatiotemporal affordances” (Hoffman & Ju 2012). Limiting both the range of movement and the non-anthropomorphic nature of the visual design creates a higher degree of focus appeal.

Cynthia Breazeal is Jibo’s company director and the Associate Professor of Media Arts and Sciences at the Massachusetts Institute of Technology (MIT). In 2014, Breazeal raised US$25 million in a crowdfunding campaign to fund Jibo, which is “an adorable-looking social robot, looks a little bit like Luxo Jr., Pixar’s signature lamp” (Kosoff 2015). Breazeal was also Hoffman’s supervisor for his 2007 PhD at MIT. Jibo shows evidence of Hoffman’s design thinking; similar to Shimon and Travis, Jibo has a limited number of DoF and a non-anthropomorphic visual design. Although it is not mobile, as with Kuri, Jibo is nonetheless engaging and appealing because of the focus on the limited range of movement based on animation thinking. Unlike Hoffman’s focus on the power of engaging movement, Breazeal wanted Jibo to have an engaging voice. Audio engineers spent considerable time and effort creating what they described as an earnest, energetic, helpful, attentive voice based on a young male character inspired by Marty McFly in Back to the Future (1985). The extensive selection process included a casting call that attracted almost 500 respondents (Robohub 2017). This process indicates the level of thinking in creating appeal through the power of voice design.

The above examples outline approaches to the marketing of social robots for the home based on similar characteristics: limited or minimalistic non-anthropomorphic design, limited range of movement based on animation thinking and appeal in voice or sound design. While these examples inhibit the range of appeal in a positive or charming way, appeal can also engage people’s attention by raising concerns or challenging the consequences of the interaction. While disturbing or challenging the interaction may not be considered a marketable attribute for social robots, it is nonetheless an important consideration for affective responses from humans who work with robots. For example, it may be prudent for social robots to demonstrate a degree of distress or discomfort in a caring situation, such as monitoring a patient in hospital or assisting an ambulance crew at a roadside accident. Distress and discomfort may also help to protect social robots from physical abuse. Again, creating a fuller range of movement capabilities for social robots will contribute to maintaining users’ attention and creating richer interaction possibilities.

Creative Robots: Artists Working with Cybernetic Machines

As indicated in the introduction, a more imaginative approach is needed to help inform higher levels of acceptance and interactivity in social robots. This section will examine how artists have worked with robots to explore diverse approaches based on the focus on appeal. A major consideration relating to innovation by artists is that many artists have non-engineering backgrounds and therefore make discoveries through alternative experimentation and innovation. While aesthetic and behavioural design is an indispensable aspect of social robotics, technical barriers must be reduced to generate innovative approaches. Some artists team up with computer scientists and engineers, while others pursue their own path. Many artists have a ready acceptance of new technologies, which they explore in creative and innovative ways. This section examines how some artists approach technology by focusing on appeal through autonomous and semi-autonomous robot movement, appearance and behaviour to reveal new, surprising and inspiring ways to think about creative robots.

In 1968, roboticist and sculptor Edward Ihnatowicz exhibited a Sound Activated Mobile (SAM) listening device that turned its head towards sound at the Institute of Contemporary Art in London. SAM looked like a futurist’s impression of a flower and was part of a larger show that addressed “systems of communication and control in complex electronic devices like computers, which have very definite similarities with the processes of communication and control in the human nervous system” (Robots in Art 1968). In 1971, Ihnatowicz created the first computer-controlled interactive robotic work of art. Senster, which looked like a collapsed electricity pylon, could turn its giant head towards sound in the manner of an animatronic dinosaur. It was commissioned by electronics company Philips as a showpiece for its head office in Eindhoven from 1970 until 1974, when it was dismantled by the artist. Senster was part of an art movement called cybernetic sculptures, which included works by Bill Vorn, Louis-Philippe Demers, Simon Penny and Robert Breer.

In 1993, Australian artist Simon Penny built an autonomous robot called Petit Mal, which was a self-levelling body suspended between two bicycle wheels and resembling a wheelchair. Penny stated that the simple behaviour of following a human around while maintaining a respec” (Penny 2017). Penny’s interests extended from the neuroscience research area of enactive cognition, which is premised on the “non-separation of perception and action, it is a constant loop” (Kim & Galvin 2012). Penny’s approach foreshadowed the importance of successful human-to-robot and robot-to-human interactions based on clarity and non-ambiguity of movement in non-anthropomorphic robots. Petit Mal successfully interacted with humans through appeal in behaviour that manifested in personality and action.

Since the early 1990s, Spanish artist Carlos Corpa has been creating “humanizing machines” (Iglesias 2013). In 2004, Corpa created PaCo in conjunction with AI and natural language specialist Ana María García-Serrano (García, 2013). PaCo is a wheelchair-bound begging robot made from found objects that approaches people in public spaces and rattles a coin box. If a person places money into the beggar’s box, PaCo uses words from its database to recite a synthetic poem in a metallic voice before printing out the poem as an “economic–literary exchange” (Iglesias 2013). In 2006, Corpa created perhaps his most disturbing robot in the form of Sufrobot, which is another wheelchair-bound robot made from a combination of natural and found items. It has exposed inner clockwork workings and an exoskeleton made from an array of wood, plastic pneumatic pipes and hardware driven by hidden software. It is reminiscent of Nam June Paik’s 1964 man-machine K-456, a bi-pedal array of metal tubing and exposed wires with a whiring mechanical and palsy movement. However, Sufrabot is unique among social robots because it visibly suffers in the company of humans. In fact, the more people that surround it, the more it suffers. This suffering is initially exhibited by Sufrabot rocking back and forth—a movement that is reminiscent of autistic behaviour. As human presence increases, Sufrobot’s anxiety increases, and it rocks faster and beats its lower limbs with its upper limbs. The limbs are made of dried tree branches and exhaust pipes from cars, and the knocking sound is sharp and urgent. Once Sufrobot is left alone, it cries an oily tear from a plastic tube at the top of its head. Audience reactions reflect those in broader human society; people are either empathetic and show a desire to comfort Sufrobot by keeping a respectful distance or they demonstrate teh cruel desire to torment. The essence of materiality and exposed design, combined with a basic human reaction to anxiety through movement, makes Sufrobot one of the most humanised machines.

Corpa is also a member of Amorphic Robot Works (AWR), which was set up by robotic sculptor Chico MacMurtrie in 1991 in Brooklyn, New York (MacMurtrie 2017a). MacMurtrie set up AWR to bring together a loose collective of artists, engineers and computer scientists to explore the possibilities for machine intelligence, interactivity and movement. The collective has more than 60 members across 10 countries and continues to create monumental and challenging artworks such as Totemobile. Totemobile is a converted Citroën DS that slowly unfolds to form an 18 m high flowering totem that blooms with light to reveal an organic nature that is not usually associated with robots or vehicles. Since 2006, MacMurtrie has explored inflatable robots by creating and extending a soft-machine approach to the social behaviour of machines. In 2006, the Australian Experimental Arts Foundation commissioned one of MacMurtrie’s earliest inflatable interactive robots, Birds, which was first exhibited in the Adelaide Bank Festival of Arts. Birds consisted of 16 soft-fabric inflatable wings (around 6m in length) that inflated to assume the shape of a child’s drawing of a bird’s wing. Suspended from the ceiling at eye level, the wings slowly inflated when the audience first entered the space. Interactivity for the work included an aspect of consequence for audience presence. If an audience member approached too close to a wing, the wing would deflate, thereby exhibiting a death cycle. The deflation of one wing would infect all wings, thus ending their life (MacMurtrie 2017b) until the next wave of visitors entered the space. MacMurtrie has created many inflatable robots—some as large as several storeys high. He views their gradual inflation and deflation cycles as a sign of life; the pneumatic pumps and switches that operate them are part of a breathing machine. MacMurtrie’s sculptural robots manifest a sensual and rhythmic approach that includes materiality and scale. Here, appeal resides in the idea that a robot can be soft, lightweight and compact while deflated, and it can then inflate and extend its presence as a soft machine to create interest in tactility by adding new parameters for the design of social robots.

The above examples show that HRI can have social weight and consequences, and they explore social conditioning and human-to-human interaction represented in the machine. The concept of distress or comfort manifested through movement and appearance based on materiality and sense of presence can help to inform creative approaches to the response of appeal in social robot design.


The availability of, and possibilities for, social robots entering people’s everyday lives will only increase with decreasing manufacturing costs and increasing investment. This article has discussed ways that help to lay the groundwork for new approaches to the design of social robots. The foundations developed by military and industrial or physical robots expose not only the high expense of robot development in the past, but also human reactions to a high degree of expectations, discomfort or failed levels of affective interaction. Artists have made a valuable contribution to design thinking for social behaviour in robots by adapting functionality and intepretations for interactivity. Approaches to movement for industrial robots, such as the curvilinear results developed by Apostolos, have helped to create a new way of thinking for HRI. Apostolos and others have demonstrated that robotic movement can be enhanced and made more appealing through the grace of dance. Dance studies have also revealed the contributions that embodied thinking and kinaesthetic empathy can make to the understanding of physical movement. The findings in this article have been supported by the cognitive aspects of human motion perception that show how understanding movement in others, and movement in animated social robots, can be understood. Finally, the review of artists working with robots has revealed a darker or alternative approach to appeal in movement; interaction, behaviour and appearance, that shows a certain cost of, or consequence for, human interaction. Designers of social robots can benefit by engaging with artists, animators and dancers to best create appeal in both look and movement for their devices. This new juncture can also benefit animators by creating new interpretations for interaction between animate devices and humans. Animation has long engaged with appeal through movement by observing and interpreting the actions of movement in the world around us; social robots are now part of that world.

The approaches outlined in this article are by no means prescriptive; rather, they seek to contribute to thinking and development for the design of social robots that extends beyond the approaches the market has demonstrated to date. If efficient and affective interactions with humans are to be achieved, consideration for clear and transparent signals afforded by appeal in behaviour, appearance and movement (whether charming or concerning) will be an important part of the future development and design of social robots.

Steve Weymouth is a full time academic in Media Arts at the University of New South Wales Australia, faculty of Art and Design.


Ackerman, E (2017), “CES 2017: why every social robot at CES looks alike,” viewed 18th September 2017,

Advanced Telecommunications Research Institute International (2017), Hiroshi Ishiguro Laboratories, viewed 18th September 2017,

Anki (2017), “Cozmo,” viewed 18th September 2017,

Apostolos, MK (1990), Robot choreography: moving in a new direction, Leonardo, 23 (1), pp. 25–9.

Baena, C (2017), Carlos Baena, viewed 18th September 2017,

Bennett, J (2009), Vibrant matter: a political ecology of things, Duke University Press, Durham, NC.

Breazeal, C (2014), “JIBO, the world’s first social robot for the home,” viewed 28th August 2017,

Corpa C (2017), “Biography,” viewed 28th August 2017,

De Quincey Co (2017), “BodyWeather & robotics: machine movement labs,” viewed 28th August 2017,

García, RI 2013, ‘Carlos Corpa: una estética crítica robótica’, Revista: Estúdio 4 (7), pp. 43–9.

Gemeinboeck, P (2017), FCJ-203 creative robotics: rethinking human machine configurations.

Gemeinboeck, P & Saunders, R (2015), ‘Towards socializing non-anthropomorphic robots by harnessing dancers’ kinesthetic awareness’, In International workshop in cultural robotics, Springer International Publishing, pp. 85–97.

Hanson Robotics (2017), “We bring robots to life,” viewed 16th July 2017,

Hoffman, G & Ju, W (2012), ‘Designing robots with movement in mind’, Journal of Human–Robot Interaction 1 (1), pp. 78–95.

Hurd, E (2017), “Appeal: The 12 Basic Principles of Animation,” viewed 19th July 2017,

IMDb (2017), “Ethan Hurd,” viewed 16th July 2017,

Johnston, O & Thomas, F., (1981), The illusion of life: Disney animation, Disney Editions.

Kim, JF & Galvin, K (2012), ‘An interview with Simon Penny: techno-utopianism, embodied interaction and the aesthetics of behavior’, Leonardo Electronic Almanac 17 (2), pp. 136–45.

Kosoff, M (2015), “An adorable family robot just raised $25 million,” viewed 16th August 2017,

KPMG (2016), “Social robots,” viewed 16th August 2017,

Lasseter, J 2001, ‘Tricks to animating characters with a computer’, ACM Siggraph Computer Graphics 35 (2), pp. 45–7.

Lipps, T (1903), Einfühlung, innere Nachahmung und Organenempfindungen.

MacMurtrie, C (2017a), “Amorphic robot works,” viewed 16th August 2017,

MacMurtrie, C (2017b), “Interactive birds,” viewed 20th September 2017,

Martin, J.J (1936), America dancing: the background and personalities of the modern dance (vol. 10), Dance Horizons, Brooklyn.

Media Art Net (1968), “Cybernetic serendipity,” viewed 16th September 2017,

Merleau-Ponty, M (1962), Phenomenology of perception, trans. C Smith, Routledge, New York, NY.

Merriam–Webster (2017), Appealing, viewed 3rd September 2017,

Mori, M, (1970), “The uncanny valley,” Energy, 7 (4), pp.33-35.

Parmenter, M (2013), ‘Body English kinaesthetic empathy, dance and the art of Len Lye’, Moving imagination: Explorations of gesture and inner movement, 89, p. 149.

Patel, NV (2015), “The creepy collective behavior of Boston Dynamics’ new robot dog,” viewed 15th July 2017,

Paterson, M (2012), ‘Movement for movement’s sake? On the relationship between kinaesthesia and aesthetics’, Essays in Philosophy 13 (2), p. 7.

Penny, S (2017), “Petit Mal,” viewed 15th July 2017,

Reed, CL (2013), ‘Seeing you through me: creating self-other correspondences for body perception’, in KL Johnson, M Shiffrar (eds), People watching: social, perceptual, and neurophysiological studies of body perception, Oxford University Press, New York, NY, pp. 44–62.

Reynolds, D & Reason, M (eds) (2012), Kinesthetic empathy in creative and cultural practices, Intellect Books, Bristol.

Robohub (2017), “Creating the voice behind Jibo,” viewed 29th July 2017,

Saenz, A (2010), Creepy but very life-like telepresence replicant from Kokoro, viewed 17th June 2017,

Takayama, L, Dooley, D & Ju, W (2011), Expressing thought: improving robot readability with animation principles’, in Proceedings of the 6th International Conference on Human–Robot Interaction, Lausanne, Switzerland, pp. 69–76.

Thomas, F., Johnston, O, (1995), The illusion of life: Disney animation New York: Hyperion.

Thompson, G (2017), “Margie Medlin’s live interactive robot performance,” viewed 6th September 2017,

© Steve Weymouth

Edited by Amy Ratelle