Donath, Psychologia
[ Pobierz całość w formacie PDF ]
//-->Mediated FacesJudith DonathMIT Media LabAbstract.Incorporating faces into mediated discussions is a complex designproblem. The face conveys social and personal identity; it reports fleetingchanges of emotion and the cumulative effects of often repeated expressions.The face both expresses and betrays: it shows what the person wishes to convey– and much more. We are highly attuned to recognizing and interpreting faces(though these interpretations are very subjective). Incorporating faces into medi-ated environments can be quite desirable: it helps the participants gain a strongersense of their community and can potentially provide finely nuanced expression.Yet there are significant disadvantages and difficulties. The immediate identify-ing markers revealed by the face, e.g. race, gender, age, are not necessarily theinitial information one wants to have of others in an ideal society. And much canbe lost in the path from user’s thought to input device to output rendering. Thisessay discusses key social, cognitive and technical issues involved in incorporat-ing faces in mediated communication.1IntroductionThe face is essential in real world social interactions: we read character and expressionin the face, we recognize people by their face, the face indicates where one’s attentionlies. Yet the face is mostly absent from online interactions – and this is in part whymany people find cyberspace to be only a pale substitute for real world contact.Today’s fast graphics cards and high bandwidth connections have eliminated manyof technical barriers to making the virtual world as fully visaged as the real world. Yetthe problem goes beyond perfecting geometric models of facial structure, for there arecomplex social and cognitive aspects to how the face is used in communication thatcannot be directly transplanted to a mediated environment. Furthermore, the desirabil-ity of faces cannot be assumed for all interfaces -- some online communities havethrived because of the absence of faces and their immediate revelation of race, gender,age and identity.Bringing the face to the interface requires radically reinventing the notion of per-sonal appearance, while remaining grounded in the cognitive and cultural meanings ofthe familiar face. It requires analyzing applications to understand what aspect of theface they need to convey - personal identity? level of attentiveness? emotional expres-sion? - and finding intuitive ways both to input and express this information. In somecases, the best interface is as realistic as possible, in others it has no face at all, whileothers may be best served by a synthetically rendered image that selectively conveyssocial information.Faces are used in many ways in computer interfaces, representing both people andmachines. This paper focuses on the role of the face in computer-mediated humaninteractions in which the face represents a particular individual, communicating withother people in a real-time, online discussion.Unlike much of the research in computer-mediated communication, we do notassume that the ultimate goal is to recreate reality as faithfully as possible. The com-puter makes it possible to go “beyond being there”[21] – to create environments thathave features and abilities beyond what is possible in the ordinary everyday world. Wecan create environments in which the face shows expression, but does not reveal theuser’s identity; we can create worlds in which traces of the user’s history of actions aresketched into the lines of the face. Yet introducing faces into a mediated communica-tion system must be done carefully, for the face is replete with social cues and subtlesignals; a poorly designed facial interface sends unintended, inaccurate messages,doing more harm than good.2Why use faces in mediated human to human communication?There are many reasons to uses faces in mediated communication. The face is verycognitively rich and holds great fascination for us. Even newborn babies, a few hoursold, will gaze longer at a face-like image than at a random array [24]. An environmentfilled with faces can be endlessly interesting to observe. People-watching is a perenni-ally favorite pastime, in which we scan the surrounding scene for familiar faces andimagine the identity of the individual behind a stranger’s visage [48]. An online envi-ronment populated with “people” with faces may seem more sociable, friendly,intriguing than a textual or purely abstract space.Faces convey important social information about who you are and what you arethinking. We are cognitively wired to recognize and remember faces and your individ-ual identity is uniquely represented by your face. The face also conveys social identity,with features that indicate basic categories such as age and gender as well as those thatare associated with particular personality types. The face conveys emotional state andintent, displaying a wide range of expressions, from puzzlement to terror, fury todelight.The faces helps to moderate and choreograph conversations. We use gaze to indi-cate attentiveness, to direct our remarks at an individual, to hold and yield the floor.Facial expressions soften our words, expressing humor, regret, etc. The face is veryimportant in conveying responses, to show understanding, agreement, etc.People behave more “socially”, that is, more politely and with greater restraint,when interacting with a face. Sproull et al. [39] found that people responded quite dif-ferently to questions posed by computers when they were presented as text or as facialdisplays. For instance, when asked questions about themselves via text they answeredwith little embellishment but when queried by a facial display they attempted topresent themselves in the best possible light.Some of these reasons for using faces in mediated communication are advantagesonly in certain circumstances. The “social” responses that Sproull et al. detected canmake a friendly discussion forum more sociable, but may be detrimental at other times.Isaacs and Tang [22] noted that non-facial interfaces could be more efficient, since theparticipants attended to the problems at hand, rather than to the time-consuming ritualsof greetings and small-talk that ordinary politeness requires; Sproull and Kiesler [38]found that hierarchical distinctions were flattened in text-only discussions – it is plau-sible (though untested) that such distinctions would regain prominence in a mediatedenvironment populated with visible faces (the desirability of maintaining or flatteningsuch distinctions is context dependent).The face allows us to characterize people at a glance. In the real world, the firstthings one learns about another are such social categories as age, gender and race, forthe cues for these categories are embodied in the face. In an ideal world, would thatnecessarily be one’s first impression? The online world has been touted as a placewhere one is identified first by one’s words and ideas, free from the stereotypesimposed by such categorization; online spaces in which one’s face is visible afford nosuch freedom. There is no simple metric for measuring the desirability of conveyingthis information, with numerous factors such as the purpose of the forum and the back-ground of the participants affecting the evaluation. What we can do is understand fullywhat social cues the face does convey and use that knowledge to help determine wherea facial display is appropriate.Including faces in the interface is very difficult to do well. This is to a large extentdue to the fact that the face is so expressive, so subtle, so filled with meaning. Weascribe character to and read emotion in any face, especially a realistically renderedone. There is no truly “neutral” face. A face in the interface is replete with social mes-sages, but a poorly designed one will send many unintended ones.In real world social situations we are constantly adjusting our face to do the appro-priate thing – to hide or show our feelings and to gaze (or not) in the proper direction.We expect the same from mediated faces and when they elide a particular social proto-col, we read an unintended message in the absence of a required expression or the acci-dental invoking of an inappropriate one. Making the “right” expression is extremelycomplex, for it is not a single motion, but a precisely timed choreography of multiplemovements: a smile that flashes briefly conveys a different message than a smile thatlingers.One of the goals of this paper is to better understand what the fundamental limitsare using mediated faces. Can the problems with mediated faces sending unintendedmessages be ameliorated with better input sensors and better renderings? Are theaspects of the face’s social role that cannot be transferred to the mediated world? Wewill address these questions by first looking more closely at what social informationthe face conveys and then examining the technologies through which we bring thesefeatures to the mediated world.3What does the face convey?Almost every aspect of the face provides some sort of social cue and we are very adeptat perceiving minute details of its configuration1. Knowing how to act toward someoneand what to expect from them is fundamental to social interaction, and this knowledgedepends upon being able to distinguish men from women, expressions of anger fromthose of joy, children from adults, friends from strangers – information that we read inthe face. Our societal structures and mores have developed with the assumption thatthis face-conveyed information is available as the context for interaction.The face conveys information through its structure, its dynamics, and its decora-tions[49]. The structural qualities include the overall head shape, the size and place-ment of the eyes and other features, the lines and texture of the skin, the color andquantity of scalp and facial hair. From these, viewers assess personality and make clas-sifications such as race, gender and age. The dynamic qualities include gaze direction,pupil dilation, blushing, smiling, squinting, and frowning. From these, viewers reademotional expression and attention. Decorations include eyeglasses, cosmetics andhairstyle from which viewers read cultural cues, ranging from large scale group mem-bership to subtleties of class distinctions and subcultural membership. There is alsoconsiderable interplay in how these qualities convey social cues. Haircuts affect theassessment of age, cultural mores modify the production and interpretation of emo-tional expressions, gender determination based on structural cues impacts the culturalinterpretation of fabricated elements such as very short hair or lipstick. Recognition isprimarily structural, though many times one will not recognize an acquaintance whohas grown a beard or is shown in a photograph with an uncharacteristic expression.The face conveys four major types of social information: individual identity, socialidentity, expression, and gaze. (This is not an all-inclusive list, for there are importantfunctions that fall outside the scope of this paper, such as, as any lip-reader knows, dis-playing the words one is saying). These types may seem unbalanced: social identity isa broad conglomeration of all sorts information about one’s gender, genetics, andgeniality, whereas gaze is really a means by which the faces conveys information (suchas conversational turn openings and attention). Yet this division is useful for thinkingabout mediated interactions, for addressing these communicative functions indepen-dently brings a great deal of flexibility and creative possibilities to the design of theinterface.3.1Individual identityWe are very adept at recognizing people. We recognize them at a distance, from vari-ous viewpoints, with different expressions and as they change with age [49]. We canfind a familiar face in a crowd with remarkable speed, especially considering howcomplex this task is: one’s mental construct of the sought face is compared to each ofthe visible faces, all of which are quite similar in overall structure and are seen fromdifferent angles, in a range of lighting conditions, and feature different expression.There is strong evidence for specific neurological bases for recognizing faces. Forexample, injury to a particular area of the brain (the occipitotemporal section of the1Our ability to distinguish minute differences among faces is so acute that Chernoffproposed taking advantage of this ability to do multivariate statistical visualizationwith faces as the graphical representation: “Chernoff faces” map data to facial fea-tures as nose length, eye tilt, head shape, etc. [6]. The resulting faces may lookhappy, sad, surprised or pained - but the underlying data is independent of the inter-preted social meaning of the face.central visual system) leaves people with their vision intact, but nearly unable to rec-ognize faces, a condition known as prosopagnosia[7]. Indeed, our notion of personalidentity is based on our recognition of people by their face. To be faceless is to be,according to the Oxford English Dictionary, “anonymous, characterless, without iden-tity.”In today’s online, text-based worlds, facelessness is the norm and the extent towhich participants are identified or left anonymous is a design feature of the variousenvironments. Both anonymous and named forums exist and flourish, though eachproduces a different tone and is suited for a different purpose [10]. Anonymous orpseudonymous spaces provide an arena for exploring alternate personas and a safehaven for discussing highly sensitive subjects; they are also more likely to devolve intoan endless exchange of flames or spam. Named forums bring the weight of one’s realworld reputation to the online world; in general, people behave in them more as theywould in real life.Online forums in which the participants’ real faces are featured – as in, for example,a videoconference – are essentially named environments. Much of the discussionabout the desirability of video as a medium focuses on issues such as bandwidthrequirements and the common gaze problem (discussed below). The fact that it makesthe forum into a public sphere in which everyone is seen and known needs to also bekept in mind, for it has a deep effect on the mores of the space.3.2Social identity and characterWe recognize people not only as individuals, but also as types. Based on the cues wesee in the face we quickly categorize people according to gender, ethnicity and age andmake judgements about their character and personality.These classifications tell us how to act toward the other, what behaviors to expectfrom them, how to interpret their words and actions. In many languages, it is difficultto construct a grammatically (or at least culturally) correct sentence without knowingthe other’s age, gender or relative social status. Such distinctions are also the basis ofprejudice, with significant biases are found even among people who consciously decryrace or gender based stereotypes[2]. More subtle but perhaps even more pervasivebiases can be found in character judgements made on the basis of facial structure, e.g.a person with a baby-ish facial structure (large eyes, small nose, large forehead, smallchin) will be judged to be more child-like in nature - trusting, naive, kind, weak [49].This, and many other character judgements based on the face derive from “over gener-alization effects”. According to Zebrowitz [49], we have very strong responses to cuesfor important attributes such as health, age, anger etc., so strong that they get over gen-eralized to people whose faces merely resemble those with that attribute or emotion.Cyberspace (the text version) has been touted as an ideal realm where the visualabsence of these cues means that people are known and judged by their words, ratherthan by their gender, race, or attractiveness. Yet it is not simply a matter of text=good,face-based classification=bad. The cues we gather from the face are basic to much ofour established social interactions, and many people find that they need to “put a face
[ Pobierz całość w formacie PDF ]