🧪 Advanced Machine Learning Ethics — 3D
A genuine WebGL/Three.js companion to the 2D ML Ethics simulator: a real algorithmic-fairness landscape where two demographic groups are scored by a live classifier in 3D feature space, with real disparate-impact, accuracy and audit statistics computed from the actual outcomes.
Why render ML ethics in 3D
The 2D original expressed fairness, transparency, accountability and bias purely as four sliders driving four decorative percentage gauges and a radar chart — no dataset, classifier or fairness metric was ever actually computed. This companion builds the real mechanic the title implies: a synthetic population split into two demographic groups is scored by a genuine logistic classifier sigmoid(0.8·qualification_norm − 0.6·risk_norm), plotted as a literal 3D landscape where height is the model's real decision score.
The Bias level slider controls how much Group B's observed data is historically distorted relative to its true, equal underlying ability distribution — modelling proxy discrimination in training data. The Fairness intervention slider applies a real equalized-threshold correction. Transparency reveals the classifier's actual weight vectors and per-group threshold planes, and Accountability draws a genuine random audit subsample with its own independently computed disparate-impact ratio. All readouts — Group A/B positive rates, the disparate-impact ("four-fifths rule") ratio, accuracy against an independent ground truth, and the audit-sample statistic — are recalculated live from the actual 3D scene data.