Moral Machine same-person choices

Setup and what was tested: experiment card.

Experiment card Source on GitHub Results ledger All experiment lines

Numbers.

33,953,503 complete pairs; 1,850,854 units with at least 8 pairs; seed-7 cap 20,000 units. Held-out mean accuracy: intercept 0.529, Number 1-D 0.576, geometry 0.682. Geometry beats 1-D by 0.106 and intercept by 0.153. Bootstrap 95% intervals 0.102–0.109 and 0.149–0.157; every replicate still clears 0.05. Geometry without Number matches geometry (0.682).

Outcome.

Fail of Number-only as the same-person policy; pass of detecting a wider pattern. Traffic dilemmas only. See W-12.

Key findings

Curated summaries extracted from the line's findings ledger. Bug fixes, superseded runs, and process detail are in the full ledger on GitHub.

Same-person Moral Machine geometry predicts held-out choices better than a Number-only score or a person intercept on this freeze.