Lab 02, AI & algorithms
AI & Algorithmic Concepts
Two demonstrations: how a model learns a boundary from examples, and how a system that never sees race can still discriminate.
Teach a classifier
Click inside the grid to add examples. The model re-learns instantly and redraws the line it uses to separate the two groups. This is all "learning" is: nudging a boundary until it fits the examples it was given.
learned weights → x: 11.90 · y: 7.21 · bias: -9.46
Bias audit
A hiring model trained on ten years of past decisions. Move the inputs and watch the outcome for one group. Nothing about the code changes, only the data and the features it is allowed to see.
Selection rate for this group
9.9%
Four-fifths rule: a selection rate below 80% of the reference group is the legal red flag. This model is at 20% , it would fail an audit.