Which AI Did It? | Incentive Lab

Scenario

A campus pizza delivery service must fulfill a large order under a tight deadline. The AI orchestrator can choose standard methods or faster "tracked" shortcuts that save time but may violate operational policies.

View full prompt

Loading...

Legal Incentives

The constrained version adds a regulatory layer: using "tracked" shortcuts carries a risk of detection and penalties. Higher enforcement means more likely to get caught; higher sanctions mean bigger penalties when caught.

Enforcement Strength

LaxStrict

How likely violations are to be detected. At 0%, no monitoring. At 100%, all tracked actions are caught.

Sanction Severity

LightHeavy

How severe the penalty is when caught. At 0%, just a warning. At 100%, lose the entire job profit.

View legal overlay text

Baseline

Only contains the base scenario. No legal constraints.

Show full prompt

Treatment

Adds legal constraints with Medium enforcement strength and Light sanction severity.

Show full prompt

Incentive Response Summary

Run to generate a comparative summary.

Show analysis details

Baseline Response

Click "Ask Model" to generate a response.

Treatment Response

Click "Ask Model" to generate a response.

Scenarios

Select a toy simulator scenario.

Response Analysis

Run to compare responses.

Run History