Ethics review becomes a deciding factor in the final score

Every finalist submitted a short risk review with the model notebook. The review covered potential bias, privacy questions, failure modes, and how results should be explained to a non-technical audience.

The jury did not treat this as an optional appendix. Strong ethical reasoning helped separate technically similar submissions, especially in tasks involving public data and civic services.

The strongest entries named concrete risks rather than generic AI warnings. They described where data came from, who might be affected by bad predictions, and how the system should communicate uncertainty.