Synthetic Pilot Lab

What I’m testing

My capstone examines how faults in experimental software distort recorded outcomes. The standalone implementation pairs correct and faulty game engines, replays scripted decisions, and checks the resulting records with an independent evaluator.

The evaluator reconstructs expected states and payoffs, then compares them with the saved observations. Some sequences of decisions expose a fault; others leave it unexposed. A passing check means no violation was observed in that sequence.

Why the decision sequence matters

A fault can be present without changing every outcome. If an action is recorded twice, for example, a sequence that never reaches that action cannot reveal the problem. A useful diagnostic needs both a way to exercise the faulty behavior and an independent calculation of what should have happened.

This distinction separates detecting a known fault from measuring how well a participant or model searches for one.

Work completed

I designed the diagnostic requirements, directed development, and reviewed the implementation and analysis. The current software includes offline diagnostics, a scripted benchmark, and measurement analysis, checked against saved synthetic records.

Next question

Model-based fault discovery remains untested. The completed work uses synthetic records and scripted decisions, with no human participants.