2.1 Visual Processing of Aesthetics

DeLong [4] describes the clothed or adorned body as the “apparel-body-construct,” the combined influence of the visual properties of the body and the visual properties of worn artifacts. The body alone and a garment alone may have discrete properties, but both are modified as they are brought together. The body and the garment each have visual properties which encode meaning – sometimes literally, in the sense of a graphic t-shirt with a text-based message, and sometimes in a more abstract manner, in the sense of the social judgments that accompany wearing a fashion trend, or having a certain body shape. The intersection of the properties of the garment and the body also contributes to the viewer’s perception of meaning in “decoding” the identity and context of the individual. For example, a too-tight waistband may constrict the wearer’s waistline, producing folds above and below the waist, and distorting the original shape of the garment. This visual cue may be read in any number of ways, depending on other contributing visual factors, and attributed to elements of identity (too lazy to buy new clothes, too vain to admit a size change) or of context (temporarily wearing ill-fitting clothes, following a fashion for cinched waistlines), or both.

A wearable device worn under clothing may produce a bump or distortion to the body shape. This distortion may or may not be obviously attributable to something being worn under the clothes. Figure 1 shows three body-worn volumes (a rectangular shape with square corners, a curved shape, and a bracelet) worn under and on top of clothing. While curved shapes may be perceived as more “ergonomic” or comfortable, when worn under clothing they may be more easily read as a protrusion of the body’s surface rather than a concealed technology.

DeLong identifies two types of characteristics of the apparel-body-construct: expressive characteristics and referential characteristics (see the left side of Figure 2). Expressive characteristics are direct characteristics of the form itself (visual elements of shape, color, texture, etc.). Referential characteristics are interpreted by the viewer; they are characteristics of the form the viewer understands as related to something outside of the form (such as a brand logo, a visual reference to another time period or a symbol of an occupational role like a badge or white coat). In some ways expressive characteristics are less open to interpretation by the viewer. They tend to play on innate responses and associations (such as bolder colors being perceived as more aggressive, flowing shapes being perceived as softer and more gentle), whereas referential characteristics depend more on the experiences and prior knowledge of the viewer. For example, it is common for teen fashion trends to reference various decades of the past (e.g., 1970s fashion references in the late 1990s and early 2000s; 1980s references in the 2010s). However, while these references may evoke direct memories in older viewers, many of the adopters of these trends have no memory of the time periods they are referencing (and may not even be aware that there is a reference), and may therefore find a different kind of meaning in the aesthetic. The same holds true for wearable systems: while the aesthetic effect of the system may hold one meaning for the designer of the technology, to an observer this meaning may be completely lost, or interpreted as something else entirely. Starner et al. [5] found that the expressive characteristics of a wearable computer were often (at the time) interpreted as being those of a medical device, the nearest mental model that most viewers could compare the wearable to. They found that altering the color of the head-mounted display quickly translated the device into a new referential association: white or light-colored devices were more often interpreted as medical devices, but grey or black devices were more often interpreted as consumer products. This division in device color specifically has blurred to some extent since 1996, but other expressive characteristics of devices still display trends that afford referential grouping by viewers.

Meanings are often defined and agreed upon by groups and sub-groups within a society (Figure 2, right side). Bell described one facet of this definition process as “sartorial morality” – the codes and mores established by a society that govern “appropriateness” of dress. These codes change with context, such that a form of dress that may be appropriate on the beach is rejected in an office setting, or clothing that is appropriate for a younger person may be inappropriate for an older person [6]. While some codes may be explicitly enforced, more often they simply carry undesired social weight: social repercussions in the form of unwanted attention or negative responses that an individual may receive when “inappropriately” dressed [7].

2.2 Visual Expression of Individual and Group Identity

One of the most important ways in which the aesthetics of dress are interpreted is in understanding and assigning group identity. In some ways, the individual can be perceived as the intersection of the many groups of which she is a member: explicit groups such as her occupation, age, and socio-economic status, as well as less explicit groups like the kind of trends she adopts, products she buys, and the way in which she wears her clothes. In wearable sensing, a device may communicate group affinities like “handicapped,” “sporty,” or “high-tech,” depending on its aesthetics and context. For example, the wearable volumes shown in Figure 1 may evoke very different group affinities depending on the shape, body location, and visibility of the “device.” By assembling the many referential characteristics of her appearance, the observer forms an aggregate impression of this individual’s identity. This impression is also heavily influenced by the observer’s experience: for example, a young man wearing very tight jeans may be displaying allegiance to a certain trend, communicating to others that he is style-conscious enough to know that this is a trend and confident enough to pull it off. An older observer, however, may not be aware of this trend or that there exists such a group of younger people who all agree that this is a stylish way to dress, and may interpret the tight jeans by drawing on other previous experiences in which men were observed wearing tight jeans. If no suitably similar previous experiences are found, the viewer may perhaps interpret the visual appearance in another context, such as evidence of rebellion, lack of social awareness, or any number of other reasons.

Because wearable systems have not yet achieved widespread adoption, it may be difficult for observers to identify the group representation afforded by a visible device. As such, the aesthetics of the system may be grouped with a nearest-neighbor reference point, by interpreting expressive characteristics that fit a known category (such as the aforementioned influence of color).

In addition to communicating group identity to others, the way we dress also communicates some element of identity to ourselves. The “role” theory of social organization posits that an individual’s understanding of how he is perceived by others mediates the roles he believes he is able to perform, and can actually affect his abilities and skills. A 2012 study by Adam and Galinsky [8] found effects in cognitive performance between groups of participants wearing identical white lab coats. The group told that their coat was a doctor’s coat performed significantly better than the group told that their coat was a painter’s coat. In wearable systems, similar effects can be found. Wearing a medical device may result in the wearer adopting a “sick person” or “patient” role, causing them to reduce their physical activity and restrict their movements in some way. For example, Costa et al. [9] found that patients wearing an ambulatory blood-pressure monitor were significantly less active when wearing the device.

3.1 The Fashion Cycle: Aesthetic Change in Fashion

While the adoption of wearable devices has not been studied in depth, adoption of apparel fashion trends is well-researched. The term “fashion” as pertains to aesthetic changes in dress can be described as a gradual evolutionary process, inspired by changes in cultural identity and values, and established through a process of social consensus. A trend is not a trend until it is agreed upon by enough people, and innovation most often happens gradually, couched in the context of current norms [10]. (An aesthetic that is truly entirely new, with very few references to current norms, is more likely to look crazy or silly than to look cool.) Successful innovations often must resonate with the cultural “zeitgeist” [11] – the moods, values, and focus of a group or sub-group of a given society at a given point in time. This process is fuzzy and difficult to define (if it were straightforward, we’d all be rich!) and in practice relies heavily on the intuition and subconscious processing of members of the design community (i.e., designers, stylists, trend forecasters, journalists).

While the origin of trends and direction of trend propagation varies widely, the adoption and progress of a trend generally follows a similar structure. Innovators and early adopters are the first to adopt a new concept. (For fashion, early adopters are those invested in aesthetics and aesthetic change, who devote considerable time and attention to perceiving and developing trends, and who are usually more interested in differentiating themselves through aesthetics rather than conforming.) As the concept progresses and gains traction (is adopted by more and more people), certain factors are at play. Rogers [10] identifies the following factors that contribute to an individual’s decision to adopt or reject an innovation: relative advantage, compatibility, complexity, observability, and trialability. Many of these factors relate to questions like: Do I understand it? Do I like it? Have I seen anyone else doing it? (And if so, who? Were they people like me?) Can I try it out before committing? For wearable devices, the relative advantage of the innovation is usually the key focus of the designer (making the device do something that the designer perceives as useful). However, if the remaining factors are not also taken into account, relative advantage may not be enough for widespread adoption.

If the innovation is adopted by early adopters and opinion leaders (Figure 3), and also meets the needs of the mass-market consumer, it will continue to pick up speed. Fashion consumers in the middle of the consumer spectrum are generally less sensitive to noticing trends, and are more interested in conformity in aesthetics (“fitting in”) than in differentiating themselves (“standing out”) [12]. Therefore, a mass-market consumer or late adopter must see the device adopted by far more people than a fashion innovator before they perceive it as something they could adopt.

Finally, as innovations are adopted by the most-resistant adopters (fashion laggards), they can either become staple or non-fashion items (if they are still in use by fashion innovators and opinion leaders), or they can become symbols of insensitivity to fashion trends (e.g., become “uncool,” if they have passed out of use for fashion innovators and opinion leaders). Laggards have also been characterized as particularly insensitive to aesthetic trends: for technology, this user group may actually be more willing to adopt a technological innovation that fashion innovators might perceive as socially unwearable, precisely because they are less sensitive to the emotional and identity-based factors of social wearability.

3.2 Social Leadership in Fashion

Many researchers and theorists have discussed possible influences of social roles on diffusion of fashion ideas. One of the first of these is the “trickle-down” theory, reinforced by Veblen’s theory of “conspicuous consumption” [13]. In this model, the wealthy upper classes consume new fashions as a means of displaying wealth and influence. These status-symbol fashions are then disseminated through lower classes. However, with the advent of mass manufacturing and mass communications, innovations became more accessible at all social strata. King [14] discussed the “horizontal flow” theory, wherein innovations “trickle across” populations rather than down from higher social classes. Finally, “trickle-up” effects, where innovations are generated by youth and social minorities and subsequently adopted by a broader section of the population, have become more prominent in the last few decades (an example of this would be the rise of “street fashion” trends).

While these theories differ in their perspective on the origin and direction in which an idea spreads, the mechanisms in each theory are similar. The influences outlined by Rogers hold true regardless of the social status of individuals in each adopter group.

In the “fashion” of behavior (including adoption of technology), there are similarly several theories that have been used to explain the manner in which new technologies are explored and adopted by user groups. Ajzen’s [15] theory of planned behavior describes the influences on an individual’s consciously intended behavior as being influenced by both the person’s attitude toward the behavior, the person’s perception of his ability to perform the behavior, and the person’s normative beliefs about the behavior (perception of what others believe he should do).

Davis’s Technology Acceptance Model [16] reduces influences on a user’s decision to adopt a technology to two major facets: the perceived usefulness and perceived ease-of-use of the technology. Malhotra and Galletta [17] expand this model to incorporate the social aspect of “psychological attachment,” the user’s perception of how well the technology fits her value system.

Wearable systems can be seen as the intersection between the processes that influence technology adoption, the processes that influence behavior change and the processes that influence fashion adoption: the usability and utility influence of technological systems, the social influence on behavior, and the emotional identity influence of fashion and worn artifacts.

4.1 Conspicuity and Social Weight

While active gestures are inherently more conspicuous than passive gestures, even passive gestures can be conspicuous when performed in an inappropriate context. For instance, a device that uses eye movements as inputs to perform specific commands (such as seen in [18]) may rely on movements that are often performed in daily life, such as blinking. However, if such a movement is performed when there is no readily observable reason for it to occur, it can become conspicuous as an active command (i.e., excessive blinking or long, drawn-out blinking).

Similarly, not all active gestures have the same conspicuity. Toney et al. [7] discuss the impact of device interactions on social interactions as “social weight”: the degree to which interacting with a device has a negative impact on a simultaneous social interaction. This is an effect familiar to most users of technology: for example, interacting with a smartphone while mid-conversation can have a distinct social effect. If an interaction can be disguised as a “natural” gesture (in the case of Toney et al., gestures like touching buttons in a pocket or hem of a suit jacket), it may decrease the perceived social weight of the interaction.

4.2 Impact of Body Location and Handedness

Conspicuity of gesture is tightly coupled to two other design variables: body location and handedness. In previous work [19], we have explored the effect of body location on perceived acceptability of on-body interactions. When viewers observe users interacting with devices placed in various body locations, the perceived acceptability shows some interesting trends. For example, interactions that occurred in body areas where existing technology is already heavily worn and interacted with (the wrist and the forearm) were much more preferred to socially sensitive areas on the body core (e.g., the torso, which is proximal to a woman’s breasts, or the pocket area, which is proximal to a male’s genitalia).

Gestures that access readily reachable body locations (such as the forearm, as seen in Figure 4, or torso) may be more “naturally” performed than gestures that require contortion of the body to reach (such as the ankle or back). It is also clear that the more intensive the interaction with respect to body resources it occupies, the more conspicuous it may become. For example, a gesture that needs only one hand is likely less to be conspicuous than a gesture that needs both hands (even more so when one or more hands may be occupied prior to initiating the gesture). Furthermore, a gesture that also demands visual attention becomes even more conspicuous.

Humans are sensitive to the direction of gaze in others – for example, McAtamney and Parker [20] found that a head-mounted visual display re-directed gaze sufficiently to interrupt the flow of a conversation. To mitigate the effect of re-directed gaze, in circumstances where interface visibility is somewhat or entirely obscured, device surface topography can be employed to help tactilely guide the user toward the appropriate interaction (see an example in Figure 5). In previous work we employed design techniques to improve the “gropability” of a wearable display to minimize the demands for visual attention in interaction [21,22]. Tangible protrusions such as buttons, beveled or chamfered edges, or discernible apertures can help to orient the user for interface operation and reduce fumbling or accidental triggering.

4.3 Impact of Cultural Norms

From our previous research (Figure 6) it has also become clear that visual perceptions are strongly dependent on cultural norms and background. As these perceptions are largely judged relative to previous experience, the source of this influence is clear. The physical appearance and fashionable properties of an item may largely be driven by culture and context [23]. Additionally, perceptions of technology use may vary as normative behavior is not identical in all cultures [24].

In our work, which looked at societal attitudes toward textile-based on-body technology placement and interactions [19], we found that Americans and South Koreans reflected a similar affinity for on-body technology placement and interactions that occurred on the wrist and forearm, and a similar distaste for their least preferred on-body locations and location-based interactions: the torso and collarbone. However, in America, there were some distinctions between attitudes toward on-body interactions. While, overall, the torso was one of the least preferred areas for device placement, interactions occurring at the torso were still significantly less acceptable on a female actor versus a male actor. Opposingly, interactions occurring at the pocket area were significantly less acceptable on a male actor versus a female actor. In South Korea, by contrast, this gender distinction was minimally present. Despite overall positive attitudes toward a particular placement or wearable, it was almost always less acceptable for the interactions to be conducted by a female actor versus a male actor, especially at areas on the upper body core.

4.4 The “Vocabulary” of Gesture

Social acceptability involves the manner in which one presents oneself so as to act comfortably within society [25], thus, societally prescribed conventions will ultimately play a role in how the physicality of interaction methods are perceived – i.e., whether they are deemed acceptable when performed within a particular context. Furthermore, the previously discussed facets of wearable technology (interaction technique, on-body placement, appearance, and aesthetics) and their perceived level of social acceptability will have a strong influence on their overall adoption.

Given these circumstances, how should gestural interactions be “designed”? Several precedents may lend insight. Many gestural interactions employ techniques common to hand-held and stationary devices, such as button-presses, taps, and touches. Newer gestural norms have emerged with hand-held touch screens, often based on real-world interactions, such as “swipe” (to move) and “pinch” (to zoom in/out) gestures. However, not all gestural inputs based on real-world interactions have taken hold: the “pinch” and “swipe” may be nearly ubiquitous now, but the “shake” gesture (to erase or undo) is not as well-entrenched.

Ashbrook and Starner [26] used a gestural toolkit to gather gesture inputs for a music player as designed by naïve consumers. While participants were relatively successful at creating distinct gestures for the audio player operations, a few participants expressed having quite a bit of difficulty devising gesture-based interactions that did not overlap with “everyday” gestures. Based on this study, a number of design strategies, such as “gestures should be easy to remember” and “gestures should be socially acceptable,” were developed to help guide future gesture design.

Many investigations of gestural input have focused on gestures made by the hands, perhaps because affixing sensor devices to phones, watches, and other devices mounted on or held in the hands is a relatively accessible application. Other natural gestures like foot tapping, shrugging, head tilting, etc., have been less fully explored, perhaps because they require sensing more “distant” body parts. An electronic-textile interface that is seamlessly integrated into clothing might benefit from input methods that exploit current clothing interactions, such as pushing long sleeves up one’s arm, or adjusting one’s tie. Capitalizing on current clothing-based interactions may help minimize any cognitive dissonance experienced by a third-party viewer.

4.5 Differentiating Passive and Active Gestures

In the space between conspicuous and inconspicuous active gestures is the ambiguity of differentiating between active and passive gestures. Gestures that appear more “normal” often begin to approximate passive gestures, and can therefore begin to be confused with actual passive gestures (risking false positive triggering of interactions with the device). Large sweeping gestures may be easily and reliably detected as inputs to a device, but may also bring unwanted attention to a user, not to mention physical fatigue or obstruction of the use of other devices. Therefore, gestural interactions should be designed so that operations delineate intent from passive actions [7]. Ashbrook and Starner [26] describe the “push-to-gesture” interaction, where an “intent” button is activated before commencing the input gesture. Our “gropable” interface used a thumb anchor-pad to both tangibly orient the hand relative to the interface, and to enable touch-based activation of interface inputs [21]. Similar techniques include micro-gestures that frame an interaction to trigger the “input” phase: an audio analog is the “OK Glass” verbal activation command interaction with Google Glass.

There is also a crucial interaction between the static aesthetics of the device and the dynamic aesthetics of the interaction. Interaction with a device that is visible to third parties may self-evidently communicate intent; however, a conspicuous interaction with an inconspicuous device may obscure intent and induce confusion for observers. A solution is to pair an inconspicuous device with an inconspicuous interaction, as seen in Karrer et al. [27] in their Pinstripe e-textile interface, developed for eyes-free continuous input. Pinstripe can be embedded on the underside of one’s shirt sleeve or pant pocket, and is manipulated by pinching the fabric between one’s fingers. User preference was found in matching the inconspicuity of the device with a subtle interaction (pinching the inner and outer side of one’s front pant pocket), thus not calling attention to the placement of the device.