Prediction on governance data with decision trees, random forests, and neural networks — taught on the GOV datasets and run entirely in claude.ai. Each session ends with cases, a self-check quiz, and a clickable resource table (R2D3, StatQuest, ISLR, ViEWS).
Prediction vs. causation; Kleinberg's umbrella test; live first look at GOV; visual investigation; mapping your research to ML tasks.
Splits and Gini by worked example; overfitting and its cures; one-prompt tree on military spending; interpretation drill.
Why many trees usually beat one; importance & partial dependence (and why neither is causal); forest vs. tree, live.
Six prompted tasks in pairs: interrogate, tree, forest, regression, break-the-model experiments, and a check — with a troubleshooting table and prompt log.
Layers, loss, early stopping; a small neural network on GOV; when networks win vs. forests.
Tree vs. forest vs. NN on one split; deployment role-play; the UK exam-algorithm case.
Unequal errors; drift, feedback, Goodhart, automation bias; Dutch benefits & COMPAS; framework preview.
Morning 9:00–12:00 · Afternoon 1:00–5:00 · Two labs, two synthesis sessions. Head to the Labs page to run the prompts in Claude.