Module 11 Summary

What this module established

Bias enters through collection decisions, each made for a sensible local reason, before any model exists - and none of them is visible in the weights. Record them at the time.

Carry forward

• Report accuracy per group with row counts. The worst group is usually the smallest, which is why the gap hides.
• Capability is not permission. Sort uses into intended, unvalidated and prohibited - the drift cases are the ones that actually happen.
• Put the baseline, the per-group results and the known gaps in the model card. The uncomfortable numbers make the rest credible.

Before moving on

Move on when your groups were defined before measurement and your permission list has three categories with stated conditions.

Mark complete