The honeypot effect: occupancy attracts more users

What it is

An empty public screen often stays empty. Once someone has actually stopped and is using it, later arrivals become more likely to approach and to try. Brignull and Rogers called this the honeypot effect: the current user is the strongest social evidence that the thing works, that using it is not humiliating, and that something joinable or watchable is happening now. It is the other face of spectator inhibition of first attempts. With no user, a crowd reads as a ban; with a user, the same crowd reads as an invitation. This entry does not cover how to lower that performance cost.

Why it happens

People learn two things from whoever is already there. Interactivity: a hand on glass that produces change overthrows the read-only class on the spot, faster than any copy. Social licence: someone is already doing it, so doing it oneself is no longer the job of breaking the calm; the performance cost is shared with an activity already underway. Passers-by are also pulled by the cluster itself—people are an attention magnet, and the screen is understood afterwards.

The effect is therefore state-dependent. At occupancy zero, a newcomer must privately finish “discover it works” and “dare to be first.” At occupancy one or more, both steps are replaced by a live demo. The same display then behaves as two machines: cold start is hard; once lit, it feeds itself. Averaging interactions over a day kneads those regimes into a medium number that explains neither the dead dawn nor a sudden crush.

Studying it

Brignull and Rogers’ observation at a large public screen is the source: record whether someone is already interacting, then whether later people stop and try. Later field work treats occupancy as a time-varying condition and plots current user count against stop rate and attempt rate of new arrivals. Do not take “confederates making a scene” as the only evidence—that measures a manipulated flow, not how natural occupancy rolls.

Independent variables: whether someone is already using the screen at approach, user count, whether the use looks fluent. Dependent variables: stop rate of new arrivals, conversion from spectating to hands-on, duration of an occupied episode from first user to dispersal.

A lab can rarely rebuild “happened to walk by and saw someone using it.” Deployment logs that store sessions but not how many people were already there at session start will book the honeypot as ordinary peak footfall. Peaks in flow correlate with occupancy; use contemporaneous passers-by as the denominator, or the finding is only “noon is busier.”

Where it stops holding

When the apron is full or the cluster blocks the path, the honeypot flips to crowding aversion and later people detour. Use that looks like failure, argument, or staff interruption becomes a live counter-example and attraction turns into warning-off. A kiosk people must queue for is brought by the task, not by a honeypot. Remote viewing (a webcast of a screen) has no body-joinable scene; the effect does not travel. A tiny one-person window machine has no room for a second body, so the honeypot has nowhere to grow.

Applying it

• When evaluating a public screen, report conversion in two windows—“occupancy zero” and “someone already there.” A day-wide average may not represent either.
• Watch cold start (zero to first user) and sustain (arrivals after occupancy) as separate durations and success rates. A large gap is a social-evidence problem, not different copy in the afternoon.
• Notice what the live demo is teaching: later people mimicking input, confirming that touch is allowed, or only spectating. Spectating without hands-on can still raise noticing without raising use.
• Verify by tagging, for each new arrival, occupancy at that instant, then stop and attempt. A clearly higher stop rate above occupancy zero means the honeypot is working; almost no first user from an empty apron means natural ignition is weak. How to ignite it is another layer. This one only requires that evaluation stop pretending the empty state is ordinary life.

Related

• Within the group: K7.03.1 Large public movements carry a performance cost · K7.03.2 Onlookers suppress first attempts
• Adjacent: K7.08 Social Embarrassment and Performance Cost · K7.01 The Passing-by Effect · K7.02 Discoverability of Interactivity
• Search terms: honeypot effect · social proof · public display · Brignull Rogers