Large public movements carry a performance cost

What it is

Raising an arm at a screen, striking a pose, speaking loudly in public turns the act itself into a show for unspecified others. Performance cost is the social risk paid in advance: looking odd, blocking the path, or appearing to make a fool of oneself with a machine. The cost rises with amplitude, duration, and how readable the act is to strangers, even after the person already knows the screen works. This entry does not cover how to lower that performance cost.

Why it happens

Public bodies are under a visibility constraint. In a Goffman-like presentation of self, an act has to fit a script bystanders can parse—phoning, waiting, reading an ad. Waving at air or aiming at glass has no script; watchers can only fill in “what is that person doing.” The emptier the script, the more explanation the actor owes.

Amplitude widens the audience. A finger tap at the waist is readable only to someone adjacent; an arm above the shoulder, a whole body following, turns half a plaza into seating. People simulate being seen before they fully reach, so many trials die in a preparatory pose: the shoulder moves, the hand never finishes. Lab throughput and recognition rates strip out that social calculation, which is why “the recognizer is excellent” and “nobody will do it in public” can both be true.

Studying it

Gesture-acceptability work in the line of Rico and Brewster asks people, in imagined or real public settings, how willing they are to perform an act and how they think others would read it. In the field, code amplitude (finger, forearm, over-shoulder, whole body) and the abort point. Reeves and colleagues treat the user as also a performer for spectators, which is the right grain for “what does this look like from outside.”

Independent variables: amplitude and duration, whether input makes sound, whether the body faces the flow. Dependent variables: stated willingness, share of field trials that start but do not finish, whether completed acts are shrunk relative to the instructed form.

A private lab booth deletes performance cost. Questionnaire “I would” runs higher than the field, because an imagined audience is easier than a real one. Do not let recognizer accuracy stand in for acceptability—the sensor can see an act people refuse to give it.

Where it stops holding

At home, on an empty street at night, or in a gallery already scripted as “you come here to play,” large movements cheapen. Professional performers, athletes, and guides already have a licence to enlarge the body. Talking into a phone with earphones has an established script, cheaper than gesturing at a public screen. When a large movement is the only accessible input, the cost remains; “nobody did it” must not be read as “this channel should not exist.”

Applying it

• Keep two ledgers: how well the system recognizes, and whether people in a trafficked place finish the act. Logs of recognition success do not answer the second.
• Code abort points: shoulder without a finished hand, a gesture started and pulled back, trials squeezed into gaps with no one passing. Treat those as performance cost before treating them as sensor failure.
• When comparing input candidates, add a column for “what this looks like in a passer-by’s sight-line,” not only speed and error. Schemes that look like flailing, blocking, or talking to oneself can be refused even when they win on throughput.
• Verify by running the same input in a booth and in real flow, and comparing attempt-completion rates. A booth far above the street is the cost; while the gap is there, do not forecast public conversion from lab throughput.

Related

• Within the group: K7.03.2 Onlookers suppress first attempts · K7.03.3 The honeypot effect: occupancy attracts more users
• Adjacent: K7.08 Social Embarrassment and Performance Cost · K7.02 Discoverability of Interactivity · K7.06 Touchless Interaction and Hygiene
• Search terms: performance cost · gesture acceptability · spectator experience · social embarrassment