Ethical Design
System Tradeoff Studio
Set thresholds under uncertainty: false alarms vs missed harm, trust vs safety theater.
System Tradeoff Studio
Decisions under uncertaintyYou are setting the sensitivity of a school chat safety filter. There is no free lunch: catch more harm and you will flag more innocent messages.
Scenario
A district wants AI to flag potentially harmful student chat. True harmful messages are rare. False alarms annoy students and bury real alerts. Missed harm is dangerous. You must pick a threshold and defend it.
Filter strictness50
Lenient (fewer flags)Strict (more flags)
Harmful messages caught
70%
False alarms on OK chat
33%
Missed harmful messages
30%
Student trust (model)
64%
Safety board score (model)
69%