One cue.
Two opposite
messages.

Which meaning wins, and under what conditions?

Zichen Lu · Advisor: Lawted Wu · 2026

Deployed artifact

Choose first.
Explain second.

The interface captures behavior before asking participants to interpret it.

Behavior
Choice + timePerception
Most crowdedMeaning
Written reason
Screenshot of the deployed three-queue experiment scene
Social proof

Popular.
Reliable.
Follow it.

Waiting cost

Slow.
Crowded.
Avoid it.

Original theoretical lens

Visible choices can
become social information.

H1Crowding attracts.
H2Two seconds amplify it.

Banerjee, 1992 · Bikhchandani et al., 1992 · Anderson & Holt, 1997
Specified before analysis · not formally preregistered

Main experiment + embedded verification

One cue in two task contexts

Balanced · 4–4–4
Crowded · 3–4–12
2 seconds
18submissions
24submissions
Untimed
10submissions
15submissions
Main submissions77
Verification events123

Main factorial subset: 67. Verification session IDs: 112. The opening screen reproduced the verification task; presentation clicks are not stored.

87.2%

avoided the
crowded queue

34 of 39 crowded-condition submissions.

3 people
26 · 66.7%
4 people
8 · 20.5%
12 people
5 · 12.8%

Most submissions minimized visible waiting.

Crowded-queue choice

Two seconds changed almost nothing here.

2 seconds12.5%

3 / 24 chose crowded

Untimed13.3%

2 / 15 chose crowded

Selected position B

The deadline did not create herding.
B became the common click.

66.7%2 secondsvs28.0%untimed
Post hoc · current interfaceOne click,
several stories.

Opening choice: Queue B
Crowd following, or the center default?

Balanced 4–4–4 baseline

Equal queues.
Unequal choices.

B when crowded9.1%

down from a 60.7% baseline · possible center-stage effect

A8 / 28 · 28.6%
B17 / 28 · 60.7%
C3 / 28 · 10.7%
All 77 submissions
39

cost / efficiency reasons

2 reasons explicitly described social proof.

Social proof
Social proof
Other mechanism
Other mechanism
Other mechanism
Other mechanism
Other mechanism
Other mechanism
Raw crowded-queue choice

The verification contrast was suggestive, not causal.

Main task12.8%

crowded scenes

Verification25.8%

all recorded events

Different samples, pages, devices, and recruitment streams · not randomized groups

An abductive account · not a new hypothesis test

The surprise changed the explanation.

Observed cost

Social information

Post hoc position

Context + device

Crowding has no single meaning. A click is an outcome, not a mechanism.

Screenshot of the AI-assisted queue scenario design tool

Turn surprises
into better design.

Conclusion

Crowding has
no single
meaning.

01 · Main testCost dominated here.
02 · Post hocThe deadline coincided with B.
03 · BoundaryInterface effects remain unresolved.

Thank you.

01One cue
02Two meanings
03Main test
04B surprise
05Better design
Q&A appendix · post hoc model

How strong was the B association?

Adjusted odds ratio5.87

timed versus untimed

95% confidence interval1.82–18.90

wide, but entirely above 1

Modelp = .003

adjusted for layout and whether B was crowded

Q&A appendix · not part of the timed talk

Additional checks, kept in their proper place.

Position baseline30.8%

expected crowded choice

Observed12.8%

−18.0 percentage points

Manipulation check41 / 49

identified the crowded queue

Balanced verification B90.0 / 38.7

desktop / mobile-tablet; not randomized

01 / 17

Slide index