Measuring Product Semantics with a Computer

Transcription

1 San Jose State University SJSU SchoarWorks Facuty Pubications Art and Art History & Design Departments October 1988 Measuring Product Semantics with a Computer De Coates San Jose State University, dcoates@decoates.com Foow this and additiona works at: Part of the ndustria and Product Design Commons Recommended Citation De Coates. "Measuring Product Semantics with a Computer" nnovation (1988): This Artice is bt to you for free and open access by the Art and Art History & Design Departments at SJSU SchoarWorks. t has been accepted for incusion in Facuty Pubications by an authorized administrator of SJSU SchoarWorks. For more information, pease contact schoarworks@sjsu.edu.

2 Measuring Product Semantics with a Computer i With so much recent interest in product semantics, where itte existed a few years ago, some designers might concude that expressive products are a new phenomenon. They might aso concude that product semantics are optiona, depending on the designer's choice. n fact. products have aways conveyed meaning. Designers have ony the choice of whether and how to manipuate a product's semantics. They can try tu detenninc which meanings their designs convey and, at best, they' choose appropriate ones. My own interest in semantics took root in the mid-1950's when fe in with beatnik artists and writers who worshipped Afred Korzybski, the father of genera semantics. Our bibe was Korzybski 's Science and Sanity: An ntroduction to Non-Aristoteian Systems and Genera Semantics. Words, Korzybski says, are not the ony raw stuff of semantics. Every human act, every human artifact, is a semantic instrument which conveys meaning. nanimate products don't actuay speak to us, of course. nstead, we project meaning into them in the form of our feeings, emotions and attitudes-a process aready famiiar to artists and designers known as empathy (first named by Robert Vischer a hundred years ago but described much earier, even by Aristote). Yet, we aone are not responsibe for this emoriona bridge. Somehow, the object shares responsibiity, because some quaity in it evokes the empathic response. and causes one set of meanings to coor it rather than another. Feeings: The Meaning of Meaning Another guiding ight of my youth was Susanne Langer, [ that "phiosopher in a new key" who was equay at home deveoping the emerging science of symboic ogic or i expaining art. For Langer, too, a perceived things, from the most premeditated and carefuy crafted museum piece to a randomy funned stone. are vesses of meaning... j,; Feeings. Feeings are the key, the raw materia of semantics. Langer stresses in, Mind: An Essay on Hui?Un Feeing, that we cannot know anything without first feeing it. A concepts, a knowedge, our very abiity to think, arc growded in our abiity to fee-in both senses of the word. Knowedge of a thing begins with the senses when we touch it and fee it with our hand. But knowedge is incompete unti it touches us, in tum. th our feeings, emotions and attitudes. t is no mere coincidence that, when something fits ogicay, we say "it makes sense" or "it fees right." (f Langer is right, no computer wi ever dispay true human inteigence unti it trst earns to fee as humans do; artificia emotion wi precede artificia inteigence.) Feeings arc a by-product, in part, of a compex set of physioogica events (irreguar heart rate, constricting and diating bood vesses, etc.) known coectivey as arousa. Arousa can be triggered by any noteworthy occurrence: a oud noise, a person's name, hunger, sexor a product of unusua or distinctive appearance. Once provoked to an eevated state of arousa, the viewer instinctivey begins a search for meaning in the stimuus_ Arousa cranks up the sensitivity and infonnation processing capabiities of the viewer's nervous system th hormona secretions and other mechanisms. To abet the information gathering, viewers instinctivey turn a their sensory apparatus toward the stimuus. Whie eyes scan the object, the nose scans the air for tetae odors and taste buds anticipate something more tangibe to work on. Senses of touch. kinesthesis (body orientation) and baance aso are on the aert for reevant cues. f any components of the sensory nervous system fai to find rea, environmenta stimui (as they do in the case of something merey seen), the system obiges the appetites by conjuring imaginary stimui from the interna storehouse of memories. Every time we have touched sharp edges. our memory has been reinforced. So we fee the edge of a new product, even though we caress it ony from a distance with our eyes, and we know it is reativey sharp, not du. Sharpness thus becomes part of the product's meaning. Simiary, we earn to fee and know with our eyes aone when a product is or smooth, or, ight or, and when a surface is sweet or a coor is sour. Other sensations, associated with the arousa reaction itsef, are more eusive and diffic.ut to smooth sow ordered du pas.'1ve static controed feminine rationa 'impe cod ight gente peacefu dea 4-door Sedan (1987) fast chaotic strong sha' acve dynamic accidenta mas.cuhne emotiona compex hot vioent [erociuu:s De Coates, DSA De Coates, DSA, teaches industrijj design at San Jose State University. He is aso the academic director ofthe university's CADRE nstitute, which is devoted to the appication ofcomputers to art and design. De is a widey pubished expert on the rok ofcomputers in design and has spoken at numerous design and computer graphics conferences. Most recenty, he organized the session on CAD at WORLDESGN88/NEW YORK. Figure : 1987 survey resut~ representing the "idea 4-door sedan" stereotype. The white diamonds represent mean (average) scores. The shaded ban represent one standard deviation (twothirds ofthe subjects' responses) on either side of each mean and indicate the reative consensus among subjects. 7

3 Measuring Product Sematics with a Compurer Figure 2: 1987 survey re suits representing the "idea station wagon" stereotype. The computer arranges scaesfrom top to bottom in order oftheir usefuness to the designer. The most usefu scaes have more extreme means judgements of 1 or 7 are more meaningfu thaf 4s) and smaer standard de''iations (greater con.~en.m.~). 8 abe because we ca!u1ot see their causes directy--constricting and diating bood vesses, fo r instance. These sensations can be especiay strong: That first sharp edge we fet as an infant and attempted to taste may have been a knife bade (Ouch!). Usuay, though, the minged sensations have been bunted by time, burred hy mutipe exposures to simiar events that are not quite identica from one time to the next. So it is difficut to distinguish them or to name them. They are just "feeings.'' "emotions" or "moods." Nevertheess, they coor immediate perceptions and the meanings we come to associate with products. The Semantic Differentia The "semantic differentia," deveoped more than 30 years ago by Chares Osgood and his students at the University of inois, remains probaby the most reiabe and easiy impemented too for tapping meanings associated with products--or any other object, concept or stereotype. t has been used to measure the meanings ofthings as diverse as poitica candidates, motherhood, appe pie and Chevroets. have used it in my aesthetics research and a~ a teaching too for neary 20 years. Professionay, have appied it most often to the anaysis ofautomotive designs. But 1aso have used it for product design and deveopment ofproduc t and corporate names. The premise of the semantic d ifferentia is simpe; Anything can be described with pair.; of antonyms (hot-cod, -ight, sharp-du, etc.) paced at opposite e nds of seven-vaued scaes. n practice, subjec ts are presented with a piece of paper with scaes printed on it. They indicate which word ofeach scae best describes ordered controed s ow smooth rationa sratic simpe b"'ntk peacefu ight du passive cod femin:ine dea Station Wagon (1987) chaotic accidenta fa.st strong emotiona d ynamic compex vioent ferocious sharp acti ve hm ma~uine the thing being considered by marcing the scae coser to one end or the other (or in the midde if no distinction seems appropriate). For instance, subjects usuay judge the concept automobie to be reativey "fast" rather than "sow," "smooth" rather than " " and "active" rather than "passive." To some extent, the same word pairs can be used to measure the meaning of any object. Simiary, subjects have judged the idea persona computer to be reativey fast, smooth and active. The same pairs coud be used, as we, to measure the meaning of motherhood or mother-in-aw. Note that, in each case cited, the subject is not referring to an actua automobie, motheror motherin-aw, but to their genera concepts instead. refer to th e resuting profie as a "stereotype." Surveys are easy to conduct. nstructions are simpe, usuay printed on the back of the survey form. They are efficient because subjects are urged not to contempate their judgments. First impressions are the rue. Typicay, a subject spends no more than a minute or two on a survey. Anaysis is another story. Manuay recording and anayzing a survey invoving just 20 scaes and 30 subjects is a time-consuming, error-prone chore. Survey design eaves something to be desired, too, because each subject sees the same form with scaes in the same order. deay, scaes woud be presented in random order, with the eft-right order of word pairs varied randomy, too. Using a Computer To overcome these probem s and extend the usefuness of the semantic differentia, J am deveoping ware for Appe Macintosh computers that assists in the deveopment, impementation and anaysis ofsemantic differentia surveys. The ware has three modues. The "Survey Deveoper" modue eads the designer th the process ofdeveoping a survey step-by-step. The computer presents a ist of over 100 word pairs to choose from for making up the scaes (the designer can add to this ist). The computer then asks the designer to indicate which subject traits the survey wi coect data on (age, sex, occupation, etc.). Once the survey has been designed, it can be saved and conducted at any time. On aunch ofthe survey, the "Survey Conductor" modue eads each subject th simpe, graphic instructions (optiona, if the subject needs them). When the subject. is ready to proceed with the survey, the computer dis pays the first scae on its screen. Using the computer's mouse, the subject moves a pointer aong the scae and registers a judgment with a simpe cick of the mouse button. This causes the next scae to automaticay appear. Presentation of scaes is superior to the usua paper survey because the computer seects the r

4 1 word pairs randomy and randomy switches their eftright order to mitigate bias effects in accordance with principes of good survey design ("fast-sow" is just as ikey to appear as "sow-fast"). At any point in the survey, the designer can invoke the "Survey Anayzer" modue to statisticay anayze the data and iustrate them gr<~phicay, as shown in Figures 1-3. Figures and 2 show resuts of surveys of idea 4 door sedan and idea station wagon stereotypes conducted in conjunction with my Fa, 1987 automotive design cass, which was preparing to design sedan wagon variants of the same basic car. Thirty university students between the ages of 18 and 27 participated in the sedan survey (16 maes and 14 femaes). Thirty-tive students between 18 and 27 participated in the station wagon survey (25 maes and ten femaes). Subjects were not considering actua sedans or wagons, ony imaginary ones. Athough most of the subjects were automotive design students, and may have conjured quite specific images of their own designs, the imaginary nature of the concepts nevertheess woud assure that they woud be reativey abstract-more so than if they were ooking at actua cars, iustrations or modes. The surveys were aimed at defining idea stereotypes, quaities which carne to mind when contempating 4-door sedans and station wagons as they shoud be. This differs from an existing stereotype which purports to represent meanings associated with things as they are. The computer sorts the resuts and ists the scaes from top to bottom in the order of their importance and usefuness in determining design strategies. The computer ooks for two things when ranking scaes: how cose the means (represented by the white diamonds) are to either end of the scae Uudgments of 1 or 7 are more important than 4 's); and the consensus of the subjects. (Other weighting factors wi be incorporated imo the formua ater.) Consensus is indicated by the engths of the shaded bars. Each bar represents one standard deviation on either side of the mean. Statisticay, it can be assumed that approximatey two-thirds of the subjects' responses wi ie within one standard deviation of the mean. The shorter the bar (or standard deviation), the greater the consensus; the onger the bar, the ess the consensus. there is no particuar aesthetic potentia in making either body type seem definitey mascuine or feminine. By the same token, however, the designer is free to make them either mascuine or feminine without dire consequences. We coud draw the same concusions if judgments varied randomy from to 7, resuting in arge standard deviations that spanned the entire scae. Semantic Distance or mage Differentiation Other usefu information can be geaned by comparing the resuts of two surveys. Genera! y, the resuts shown here suggest that "sedan" is a more potent or energetic concept than "wagon." We know this by noting that adjectives on the right-hand side of the graph are a more highy correated with notions of power and activity than their opposites on the eft, and that sedan judgments ie further to the right on most scaes than wagon judgments. n those notabe cases where the difference exceeds 0.5, the sedan is perceived as faster, sharper and more active than the wagon. t is aso thought of as more emotiona and ferocious. The wagon is more potent than the sedan on ony four scaes: t is er (not as smooth), stronger, heavier, and more accidenta (ess controed). This tes the designer who is deveoping a station wagon variant of a sedan that it shoud seem reativey stronger and heavier than its sedan counterpart. t aso say~ that, if the resut isn't quite as smooth or controed as the basic sedan, it's okay-witness the reative cutter of the Taurus wagon's roof rack, compared to the sedan's cean roof, and the abrupt discontinuity of the wagon's betine at the rearrnost door cut. Aesthetic Potentia Comparisons of idea and existing stereotypes yied measures of aesthetic potentia-indications of pent-up "demand," if you wi-for specific meaning which the designer can use in deveoping design strategies. A survey of "the typica four-door sedan," for instance, might yied a mean of 3.0 on the ordered-chaotic scae. Compared with the more extreme 2.4 of the idea stereotype, this woud mean that sedans currenty on the streets are ess than idea in this regard and that consumers (reprcsemed by the particuar group participating in the two surveys) woud be receptive to a sedan with a consideraby more ordered ook. Trends Comparisons of stereotypes deveoped at different times yied measures of potentia trends. Consider, for exampe, a comparison of the idea 4-door sedan stereotype with an automobie stereotype (Figure 3) deveoped from a survey of 23 mae industria design students at the Center for Creative Studies in (Athough didn't coect it, References Coate:=-., De, ''Aesthetics,'' Proceedings of the Human Factors Engineering Summer Conference, spommred by The University of Michigan. Ann Arbor, MT (each year since 1970). Coates. F.D. "n Scan;h of the Sense of Beauty," Design Journa (DSA), 1972, Vo.5 pp Coates. De. "79 Autos: The European Look Spreads-Or s t Just Good Design?" ndu.rtria D~.rif{n Magazine, September/October, 1978, pp Coates, De, "Under:standing Aesthetics: From Od Shoes to a Teacup." ndustria De>ign Magazine. September/October pp Langer. Susanne K.. Phiosophy irt a New Key. Cambridge, MA.; Harvard University Press Langer. Susanne K., Mind: An essay on Human Feeing. Batimore, \.10; The Johns Hopkins University Press, 1967 (vo. ), 1972 (vo.t), 1982 (vo..t). Korzybski. Afred, Science and Sanity: An ntroductirm to Non-Aristnreian Systems ar~d Cenera Senuznric$/ Lakevie, CT: The nternationa Non Aristoteian Library, Kreiter. Hans & Krciter, Shuamith, Psychoogy ofthe Arrs. Durham. NC: Duke University Press, Osgood. C.E., Suci. G.J., & Tannenhaum, P.H., The Mea.uremenr of Meaning. Urbana, L: University of 111inois Pres., According to the resuts, expectations are so high that a 4 door sedan shoud be "smooth" and "fast," that a designer woud be fooish to design one that seemed "" and "sow." The feminine-mascuine scae, on the other hand, is reativey unimportant in either the sedan or wagon case because means for both are on or near the neutra midine. We can draw two concusions from this. First, 9

5 Me asuring Product Sematics with a Computer Figure 3: 1977 survey resuts representing the "auwmobie" stereotype. Not ony do judgements on individua scaes differ from those ofa simiar survey done ten years ater (Figure 1), but the reative importance of scaes differ. pa <~ve s ow static du Automobie (1977) s 6 7 s ordered mooth ~:=~==~~ simpe feminin e controed gente peacefu ight cod rationa acti ve f ast <tmng d ynamic shaq> c haotic compex mascuine accidenta viuc::nt ferocious hot emotiona age data woud be simiar to the 1987 surveys.) The comparison is not idea because the concept names were not identica. Athough the subjects were preparing to design a 4-door sedan (and 1 presumed they h ad that in mind), notions ofsports cars and station wagons may have affected their judgments, as we. Nor were they instructed to think specificay of typica or idea cases. Nevertheess, for purposes of demonstration, et's con sider the sudden impetus for design change--and change in taste~provided by the e nergy crises of the 1970s and the effects they might have had on stereotypes then and now. We shoud see shifts consisten with downsizing of cars and increased aerodynamic efficiency. Tndeed, there was a shift from (in 1977) to ight (in 1987), perhaps as the pubic earned to appreciate smaer (ighter) cars for their fue efficienc y. Judgments aso went from compex to simpe, from to, from mascuine to feminine, coser to smooth and away from sharp-just as we woud expect if tastes for the straight-ined, crispedged boxes of the '70s have given way to affection for the vouptuous curves, fush surfaces and ed ges of the ate '80s' "jey-beans." Athough the 1987 surv eys yieded on y a sighty more extreme judgment toward smooth than the earier survey. the importance of smoothness, a prime mark of good stream1ining, became paramount in the case of the idea4-door sedan, jumping from the midd1e to the top of the graph. Subjective Concinnity The computer can cacuate the subjective concinniry of a particuar product design by comparing surve ys of it and its reevant stereotype. A survey of the idea teephone, for instance, woud be compared with a s urvey of an actua mode or prototype of a proposed teephone. The same subjects coud be used for both surveys; if not, subject.~ shoud at east be drawn from the same market segment being targeted by the product. (For more information on the concept of concinnity, see references 6-9.) The compute r cacuates a dimensioness coefficient from 0 to 1.0 which meas ures the semantic distance between the idea and rea concepts aong a ine, in effect, th a three-dimensiona semantic space. f the concepts arc cose together (hig h subjective concinnity), the proposed design wi be compatibe in the cassic sense defined in the human factors iterature. The proposed design wi tend to meet a expectations. t wi not seem nove or unusua and, thus, wi not offend the intended consumer. Neither wi it stir the emotions in ways that create much eye-grabbing, heart-pounding appea. The greater the semantic distance between a product and its stereotype. the greater its aesthetic potentia (its abiity to stir the viewers' feeing s). Unfottunatey, aesthetic potentia is a doube-edged sword. A product dispaced in one di rection away from its stereotype wi evoke peasant excitement, which attracts the viewer. But dispaced in the opposite direction on this ine th the semantic space, the produc t wi provoke uneasiness, which the viewer woud rather avoid. The computer sorts things out and suggests to the designer how to optimay baance aesthetic potentia and subjective concinnity for the best resuts. Computation of a product desig n 's objective concinnity requires an anaysis of the design's geometry wh ich is beyond the scope of this ware. However, the ware can provide an estimate of reative objective concinnity based on the fact that certain scaes of the semantic differentia (ordered-c haotic, for instance) are highy correated with objective concinnity. Future Deveopments Appe has announced pans to introduce a ap-top portabe version of the Macintosh within a year. That machine shoud extend the utiity of the ware dramaticay by making fied surveys much more practica. Future versions of the ware wi incude an "Advisor" modue, based on expert system technoogy, which wi1 more activey hep he designer to deveop strategies for improving a design. Evenruay, expect to tie this modue ro computer-aided design ware so that the computer can try fixes to the product's geometry automaticay in order to optimize its semantics. 10