DIS2002 | 00

All Robots Are Not Created Equal:The Design and Perception ofHumanoid Robot HeadsCarl F. DiSalvo, Francine Gemperle, Jodi Forlizzi, Sara KieslerHuman Computer Interaction Institute and School of Design, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213 USAcdisalvo@andrew.cmu.edu, gemperle@cmu.edu, forlizzi@cs.cmu.edu, kiesler@cs.cmu.edu

ABSTRACTThis paper presents design researchconducted as part of a larger project onhuman-robot interaction. The primary goalof this study was to come to an initialunderstanding of what features anddimensions of a humanoid robot’s facemost dramatically contribute to people’sperception of its humanness. To answer thisquestion we analyzed 48 robots andconducted surveys to measure people’sperception of its humanness. Through ourresearch we found that the presence ofcertain features, the dimensions of thehead, and the total number of facialfeatures heavily influence the perception ofhumanness in robot heads. This paperpresents our findings and initial guidelinesfor the design of humanoid robot heads.

Keywords human-robot interaction, socialrobots, interaction design, design research

INTRODUCTION AND MOTIVATIONAdvances in computer engineering andartificial intelligence have led tobreakthroughs in robotic technology.Today, autonomous mobile robots can tracka person’s location, provide contextuallyappropriate information, and act inresponse to spoken commands. In thefuture robots will assist people with avariety of tasks that are physicallydemanding, unsafe, unpleasant, or boring.

Because they are designed for a socialworld, robotic assistants must carry outfunctional and social tasks. Much of theresearch in robotics has focused onimproving the state of the currenttechnology. Our goal is to match thetechnology to the needs of users. Althoughthe technology exists to build a robustrobotic assistant [24], we lack a principledunderstanding of how to design robots thatwill accomplish social goals.

The goal of our project is to conductapplied research into the cognitive and

social design of robots. If robots are goingto be intelligent social products that assistus in our day-to-day needs, then ourinteraction with them should be enjoyableas well as efficient. We are interested inissues of product form, behavior, andinteraction in social robots as they relate toaccessibility, desirability, andexpressiveness. From our research, we willdevelop models of human-robot interactionthat support appropriate and pleasantexperiences and use these models to createguidelines for the design of assistiverobots.

This research is important for the fields ofinteraction and product design, human-computer interaction, and robotics. Human-robot interaction is a new area of researchand the impact of design on this field hasyet to be understood.

Most of the research efforts in human-robot interaction have not been focused ondesign [11-13, 25]. Relevant work has beendone in such related areas as

Permission to make digital or hard copies of all or

part of this work for personal or classroom use is

granted without fee provided that copies are not

made or distributed for profit or commercial

advantage and that copies bear this notice and the

full citation on the first page. To copy otherwise, or

republish, to post on servers or to redistribute to lists,

requires specific permission and/or a fee.

DIS2002, London © Copyright 2002 ACM 1-58113-2-

9-0/00/0008 $5.00 Figure 1. Elders interacting with a social robot (Pearl)

00 | DIS2002

anthropomorphism [9, 17, 20], computersas social actors [20, 21], facial interfaces[19, 26 - 28], and believable agents [7, 8,23]. Although basic and tacit knowledgefrom other areas of research and design canbe brought to inform human-robotinteraction, core design research is stillneeded to understand and articulatechallenges of interacting with anddesigning social robots.

Many robotics researchers are pursing ahumanoid robot form as the mostappropriate form for a social robot [11- 13,25]. These researchers have assumedimplicitly that the head will be the primaryplace of human-robot interaction. Whilethis assumption has yet to be scientificallyproven, we have chosen to pursue researchin the area of humanoid robot heads for apragmatic design goal: the design of a newhead for our robot, Pearl.

Pearl was developed as part of theNursebot project(http://www.cs.cmu.edu/~nursebot). Pearlis used in both laboratory experiments andfield settings as part of our research intohuman-robot interaction. We are currentlyre-designing Pearl s head to have modularfeatures. Modular features will allow us toeasily reconfigure Pearl’s head and conductfurther experiments on the impact of facialfeatures and dimensions.

This initial study was used to inform thatre-design process. The findings are beingused to identify what facial features anddimensions will be most important for us tohave control over and direct the industrialdesign team in the creation of a new headfor Pearl.

METHODWe began by collecting images of 48robots from websites, books, andmagazines. We sorted theses images into 3categories: Research, Consumer Products,and Fiction. The Research categoryconsisted of robots that have been createdin educational and industrial researchlaboratories (n=18). Pearl is an example ofa research robot (Figure 2).The ConsumerProducts category consisted of robots thathave been manufactured to be for sale asactual functioning products (n=14).ASIMO is an example of a ConsumerProduct robot (Figure 3). The Fictionalcategory consisted of robots fromtelevision, film, and toys (n=16). TheTransformer is an example of a fictionalrobot (Figure 4).

SurveysWe used the images of the 48 robots toconstruct two paper and pencil surveys.One survey contained an image of the headand body of each robot. The other surveycontained an image of each robot headonly. In both surveys, participants were

asked to rate each image on a 1 to 5 scale,from Not Very Human Like to Very HumanLike. We solicited 20 participants for eachsurvey. Participants either did the robothead or the whole robot survey, but notboth.

The results for each survey were correlatedto assess the validity of robot head scores.The two surveys were highly correlated,suggesting that our scores of the perceptionof humanness of robot heads are accurateand valid.

Robot Head AnalysisUsing images of the 48 robots wecollected, the heads were coded for thepresence of eyes, ears, nose, mouth,eyelids, and eyebrows, and the totalnumber of features present on the head.The heads were scaled to a height of 10inches so that all of the measurementswould be relative. The images of the facewere measured to record the height/widthratio of each face; the percentage of theforehead region, feature region, and chinregion, the size of the eyes, the distance

Figure 2. Pearl, our Research robot Figure 4.The Transformer, an example of a

Fictional robot

Figure 3. ASIMO, an example of a

Consumer Product robot

Figure 5. Diagram of the comparative physical measures taken

All heads were scaled to 10” height

between the eyes, and the width of themouth (Figure 5).

Using the data from the head analysis andthe robot head survey (we did not includethe ratings from the whole robot surveys),we constructed two statistical modelsrelating to the perception of humanness inrobot heads: The Presence of Features andThe Dimensions of The Head and TheTotal Number of Features and performed aregression analysis on these models tocome to our findings.

FINDINGSThe Presence of FeaturesOne would assume that all humanoidrobots would have facial features but this isnot the case. Of the 48 humanoid robotsthat we surveyed, six did not have anyfacial features. SIG is an example of ahumanoid robot without any facial features (Figure 6). However, the presence of facialfeatures is very important. The presence ofspecific facial features account for 62% ofthe variance in the perception ofhumanness in humanoid robot heads. Thethree features that increase the perceptionof humanness the most are the nose (p < .01), the eyelids (p = .01) and themouth (p < .05) (Figure 7).

The Dimensions of The Head andFeatures and The Total Number ofFeaturesThe shape of a human head, the dimensionsof facial features, and the distribution ofthose features on the head are fairlyuniform in humans, but this is not the casein robots (Figure 8). We saw a similarvariance in the width of the head relative tothe height. None of the dimensions of thefacial features are individually significant

in the perception of humanness in robotheads. However, the total number offeatures on the robot’s head is significant inthe perception of humanness (p < .01). Themore features that a robot head has, themore human like it will be perceived. Thewidth of the head is also significant, in theperception of humanness (p < .03); thewider the head compared with the height,the less human-like it is perceived.

Robot Heads In Comparison ToHuman HeadsWe were curious to know how much thedimensions of a robot’s head differed fromthe dimension of human heads. Wecombined the dimensions of the facialfeatures of the Mona Lisa, Michelangelo sstatue of David, George Bush, and BritneySpears to develop prototypical humanhead. We compared this prototypicalhuman head to a somewhat human-likerobot head and a very human-like robothead. For the very human-like robot headwe chose the robot from Metropolis whowas ranked the second most human-likeof our 48 robots. For the somewhat

human-like head we chose Lazlo, aresearch robot from MIT, who fell withinour median range of humanness and wasranked 19th most human-like robot(Figure 9).

How Human-Like is Humanoid?Although all of the robots included in thissurvey were classified as humanoid, themajority of them were not rated as beingvery human-like. The mean score on thescale of humanness for the robot headswas 2.74 (sd 0.68). This does not conflictwith their classification as humanoidrobots, for that simply means that theirform resembles a human more than itresembles any other form. This does,however, raise the issue of how human-likea robot can be perceived by form alone.Humanness will be defined not only byform but interactions through expression,communication, and behavior.

The Importance of DesignAs designers we would like to believe thatthe design of facial features is important inthe perception of humanness. Not all robots

Figure 6. SIG, a robot without facial

features

Figure 8.The dimensions of features on robot heads

Note: All dimensions are relative to the head at 10” height

0

20

40

60

80

100

EarsMouth*Nose*EyebrowsEyelids*Eyes

81.25%

18.75 16.67%

29.16%

62.5%

43.75%

Figure 7.The presence and influence of features on robot heads

* The presence of these features had a statistically significant effect on the perception of humanness

Eye Diameter

Pupil to Pupil

Mouth Width

Head Width

Top

Middle

Bottom

Minimum Median Maximum

0.25" 1.75" 4.75"

1.375" 4" 11.75"

0.875" 3.75" 9.25"

5.25" 9.63" 20"

8.75% 35% 62.5%

28.75% 60% 100%

6.25% 11.88% 27.5%

00 | DIS2002

have designed facial features. Manyhumanoid robots express their humanqualities only through the suggestion offeatures. However, designed features dohave a significant effect on the perceptionof humanness (p < .01). Whether or not thefeatures had been designed accounts for23% of the variance in the perception ofhumanness in humanoid robot heads.Kismet (Figure 10) is an example of arobot with highly designed features, DB(Figure 11) is an example of a robot whosefeatures are merely suggested.

This finding suggests that in situationswhere it is not possible or feasible todesign the actual facial features providingsuggestions of those features, in effectaffordances for those features, may sufficein creating an overall perception ofhumanness in the robot head.

IMPLICATIONS FOR DESIGNHow Human Should a Head Be?There is a dual challenge in our designproblem. First, we must understand whataspects of robot form lend themselves tobeing sufficiently human-like to carry on

social interaction in an appropriate andpleasant way. Next we need to understandthe aspects of the robot that need to remainrobotic enough to clearly display therobot’s non-human capabilities andemotional limitations.

Mashiro Mori developed a theory of TheUncanny Valley (Figure 12), which statesthat as a robot increases in humannessthere is a point where the robot is not 100%similar to humans but the balance betweenhumanness and machine-like isuncomfortable. Mori provides andexample: If you shake an artificial hand[that you perceive to be real] you may notbe able to help jumping up with a scream,having received a horrible, cold, spongy,grasp . According to Mori there is areasonable degree of familiarity that shouldbe achieved and maintained in humanoidrobots [22].

Between the three categories of Research ,Consumer Product, and Fictional robots,Fictional robots are on average the mosthuman-like and Consumer Products are onaverage the least human-like. Although the

difference between the categories is notlarge, it is enough to suggest a trend inConsumer Product robots to appear morerobotic than human. Whether this trend isdue to the technical constraints of creatinga robot for everyday use or reflects theactual preferences of users has yet to bedetermined and is an important topic forfuture research.

The relationship of the body to the headand the importance of the body in theoverall perception of humanness is anotherimportant topic of inquiry. Although thisstudy focused on the form of the head thebody clearly plays a role in the perceptionof humanness.

We are working toward identifying thethreshold of humanness that is mostappropriate for social robots. We knowfrom existing literature that the face isextremely important in scenarios of human-to-human interaction and we know how thehuman face functions in those scenarios[10, 14, 29]. However, a robot is not ahuman and its form will always be differentthan that of a human. A need exists for a set

Figure 9. Contrast of measures between our comparison human head, a very human-like robot head, and a median human-like robot head

0

10

20

30

40

50

60

70

80Lazlo

Metropolis

Average Huma

bottomfeaturestop

(% o

f hea

d)

0

1

2

3

4

5

6

7

8

width at 10" height

inch

esvery human-like head median human-like head

sources of our comparison head

fam

iliar

ity

0% 100%

The uncanny valley

similarity

Figure 12. Mori’s The Uncanny Valley,

adapted from Reichard 1978

Figure 10. Kismet, a robot with designed

facial features

Figure 11. DB, a robot whose facial features

are merely suggested

DIS2002 | 00

of heuristics to define the appropriatedesign of a humanoid robot that interactswith humans.

Our research has led us to create a set ofsuggestions for the physical design of anew head for Pearl. It is important to notethat this set of suggestions does not includethe design of movement. These designsuggestions for a robot head take intoaccount three considerations: the need toretain an amount of robot-ness so that theuser does not develop false expectations ofthe robots emotional abilities but realizesits machine capabilities, the need to projectan amount of humanness so that the userwill feel comfortable socially engaging therobot, and the need to convey an amount ofproduct-ness so that the user will feelcomfortable using the robot. Together,these suggestions create a balance betweenwhat we expect of a human, a robot, and aproduct for an effective design. In the nextsix months we plan to execute theseguidelines in the design of a new head forour robot Pearl (Figure 13).

Design Suggestions for a HumanoidRobotic Head1. Wide head, wide eyesTo retain a certain amount of robot-ness, bymaking the robot look less human, the headshould be slightly wider than it is tall andthe eye space should be slightly wider thanthe diameter of the eye.

2. Features that dominate the faceThe feature set, from browline to bottom ofmouth, should dominate the face.Proportionally, less space should be givento forehead, hair, jaw or chin. Thisdistribution is in contrast to a human s andcombined with the size of the head, willclearly state the form of the head as beingrobot-like.

3. Complexity and detail in the eyesHuman eyes are complex and intricateobjects. To project humanness a robotmust have eyes, and the eyes shouldinclude some complexity in surface detail,shape of the eye, eyeball, iris, and pupil.

4. Four or more featuresThe findings from our study show that thepresence of a nose, a mouth, andeyebrows, greatly contribute to the

perception of humanness. To project a highlevel of humanness in a robot thesefeatures should be included on the head.

5. SkinFor a robot to appear as a consumerproduct it must appear finished. As skin, orsome form of casing is necessary toachieve this sense of finish. The headshould include a skin or covering ofmechanical substructure and electricalcomponents. The skin may be made of softor hard materials.

6. Humanistic form languageThe stylized appearance of any productform is important in directing ourinteraction with it. To support the goal of ahumanoid robot the head shape should beorganic in form with complex curves inthe forehead, back head and cheek areas.

FUTURE RESEARCHThe effect of interaction in the perceptionof humanness should not beunderestimated. While this study wasconducted with static images of robotsisolated from any context and devoid ofanimation or interaction, our futureresearch will conduct similar measures ofhumanness with physically present,animated, and contextually situated robots.We believe that interaction through speechand movement will greatly effect theperception of humanness in robots.

Our future research will also address robotforms that are not humanoid. Weacknowledge that the importance of using a

humanoid form is still an assumption thathas yet to be proven. We plan to exploreother robotic forms and their effect onfacilitating social human-robot interaction.

CONCLUSIONThis study showed that the presence ofcertain features, the dimensions of thehead, and the number of facial featuresgreatly influence the perception ofhumanness in robot heads. Some robots aremuch more successful in the portrayal ofhumanness than others. This success is due,at least in part, to the design of the robot shead. From these findings we have createdand initial set of guidelines for the designof humanoid robot heads. Specifically, wehave identified features and dimensionsthat can be used to modulate how human-like a robot head will be perceived. Thesefindings should serve as a connectionbetween ongoing robot research and thesocial products of the future.

ACKNOWLEDGEMENTS:We would like to thank Jen Goetz andKyle Kirby for their help and support withthis study. This research was supported bygrants from the National ScienceFoundation, # 0085796 & #112018.

REFERENCES1. Bates, J. The Role of Emotion in

Believable Agents. Communications ofthe ACM, 37 (7). 122-125.

2. Bates, B. The Human Face. BBCWorldwide Limited, New York, 2001.

Figure 13. Initial sketches for our new robot head

00 | DIS2002

3. Breazeal, C. and Velasquez, J., TowardTeaching a Robot "Infant" usingEmotive Communication Acts. InProceedings of Simulation of AdaptiveBehavior, workshop on SociallySituated Intelligence 98, (Zurich,Switzerland, 1998), 25-40.

4. Breazeal, C. and Velasquez, J., Robotin Society: Friend or Appliance? InProceedings of Agents 99 Workshop onemotion-based agent architectures,(Seattle, WA, 1999), 18-26.

5. Breazeal, C. and Scassellati, B., Howto Build Robots That Make Friendsand Influence People. In Proceedingsof IROS-99, (Kyonju, Korea, 1999).

6. Bruce, A., Nourbakhsh, I. andSimmons, R. The Role ofExpressiveness and Attention inHuman-Robot Interaction. InProceedings, AAAI Fall Symposium.(2001).

7. Cassell, J. Embodied ConversationalAgents. MIT Press, Cambridge, Mass.,2000.

8. Cassell, J., Bickmore, T., Vilhjlmsson,H. and Yan, H., More Than Just aPretty Face: Affordance inEmbodiment. In Proceedings of 2000International Conference on IntelligentUser Interfaces, (New Orleans, LA,USA, 2000).

9. Don, A., Brennan, S., Laurel, B. andSchneiderman, B.,Anthropomorphism: From Eliza toTerminator 2. In Proceedings ofHuman Factors in Computing Systems92, (1992), 67-70.

10. Ekman, P. (ed.), Emotion in theHuman Face. Cambridge UniversityPress, Cambridge: UK, 1982.

11. Humanoid Interaction Lab, NationalInstitute of Advanced IndustrialScience and Technology, Japan.

http://www.etl.go.jp/etl/robotics/Projects/Humanoid/

12. Humanoid Robotics Institute, WasedaUniversity, Japan.http://www.humanoid.waseda.ac.jp/

13. Humanoid Robotics Group,Massachusetts Institute of Technology,USA.http://www.ai.mit.edu/projects/humanoid-robotics-group/

14. Hassin, R. and Trope, Y. Facing faces:Studies on the cognitive aspects ofphysiognomy. Journal of Personality& Social Psychology, 78 (5). 837-852.

15. Kiesler, S. and Sproull, L. "Social"Human Computer Interaction. InFriedman, B. ed. Human Values andthe Design of Computer Technology,CSLI Publications, Stanford, 1997,191-199.

16. Klingspor, V., Demiris, J. and Kaiser,M. Human-Robot-Communication andMachine Learning. Applied ArtificialIntelligence Journal, 12 (7-8). 573-617.

17. Laurel, B. Interface Agents: Metaphorswith Character. in Friedman, B. ed.Human Values and the Design ofComputer Technology, CSLIPublications, Stanford, 1997, 207-219.

18. Menzel, P. and D’Aluisio, F.RoboSapiens. MIT Press, Cambridge,Mass, 2000.

19. Moon, Y. When the Interface is theFace: Social Responses to InteractiveTechnologies. Report (113). 23-24.

20. Nass, C. and Steuer, J.Anthropomorphism, Agency, andEthopoeia: Computers as SocialActors. Human CommunicationResearch, 19 (4). 504-527.

21. Nass, C., Steuer, J. and Tauber, E.,Computers are Social Actors. InProceedings of Human Factors inComputing Systems 94, (1994), ACMPress: New York, 72-78.

22. Reichard,J. Robots: Fact, Fiction, andPrediction. Penguin Books, 1978.

23. Reilly, W.S.N., A Methodology for Building Believable Social Agents. In Proceedings of Autonomous Agents 97, (Marina Del Rey, California USA,

1997), 114-121.

24. Roy, N., Baltus, G., Fox, D.,Gemperle, F., Goetz, J., Hirsch, T.,Margaritis, D., Montemerlo, M.,Pineau, J., Schulte, J. and Thrun, S.Towards Personal Service Robots forThe Elderly.http://www.cs.cmu.ed/~nursebot/web/papers.html

25. Social Robots Project, CarnegieMellon University, USA.http://www2/cs.cmu.edu/afs/cs/project

/robocomp/social/www/SRPmain.html

26. Sproull, L., Subramani, M., Kiesler, S.,Walker, J.H. and et al. When theinterface is a face. Human-ComputerInteraction, 11 (2). 97-124.

27. Takeuchi, A. and Nagao, K.,Communicative Facial Displays as ANew Conversational Modality. InProceedings of INTERCHI 93,(Amsterdam, the Netherlands, 1993),ACM Press: New York, 187-193.

28. Takeuchi, A. and Naito, T., SituatedFacial displays: Towards Socialinteraction. In Proceedings of HumanFactors in Computing Systems 95,(1995), ACM Press: New York, 450-455.

29. Zebrowitz, L.A. Reading faces:Window to the soul? Westview Press,Boulder, CO, USA, 1997.