Tutorials to introduce AI technology in their legal, ethical, and societal contexts. Each demo runs entirely in your browser--no installation required.
Explore how word embedding models encode human biases using vector arithmetic. Type three words; the model finds the fourth.
Watch a neural network learn step by step using backpropagation. See the weights update on every connection as the network trains — one node, one example, or one epoch at a time.
Examine the real ProPublica dataset behind the landmark Machine Bias investigation. Adjust the risk-score threshold and wave a Magic Wand to explore what it would take to satisfy competing definitions of fairness simultaneously.
Build your own fairness metrics by dragging cells from two confusion matrices into a ratio workspace. Discover how 20+ named metrics are derived from the same data — and why they so often disagree.