CPC2015 risky choice

A published risky-choice experiment: a many-feature pattern did not beat a one-number expected-value score or a person intercept. Recorded as a null; not retuned.

Experiments · Witness · Results

Source on GitHubResultsResults ledger

What. CPC2015 Experiment 1 is a lab risky-choice dataset with many trials per person. We ask whether a wider feature pattern beats scalar expected-value difference or a person intercept on held-out trials.

Why. The same detection pipeline as same-person Moral Machine, on a different choice class. A pass here would not have been a moral-bundle result; a miss is still a result.

Witnesses.

Host.

Zenodo CPC2015 Experiment 1 (RawDataExperiment1sorted.csv; Erev et al. 2017). Unit is SubjID. Risk/ambiguity lab class only — not a moral-bundle result.

Setup.

Frozen protocol h4-cpc2015-v1.0.0, fixture h4-cpc2015-v1.json. Checker check_h4_cpc2015.py. Subjects need enough trials for a train/test split. Frozen margins: geometry must beat ΔEV and intercept by at least 0.05. Not retuned after the miss.

Analysis.

Held-out mean accuracy for a person intercept, a one-number expected-value difference (ΔEV), and a many-feature geometry. Same detection pipeline as W-12, different choice class.

Finding.

On CPC2015 Exp. 1, the many-feature pattern did not beat ΔEV or a person intercept. Null; not retuned. Full results