Study sessions

Worked study sessions that show a student running the Learning Spiral on a genuinely messy problem, with AI as a study partner rather than an answer machine. Each session annotates the moves: where the student keeps the thinking, asks for hints instead of answers, and checks the AI rather than trusting it.

These extend the annotated dialogues in Part I of the book, with longer, messier problems across subjects.

Sessions

• Statistics / data reasoning: is this relationship causal? (statistics-causal-claim.md). A student decides whether an observed relationship supports a causal claim, and catches the AI’s agreeable answer.
• Coding / debugging: finding a bug yourself (coding-debugging.md). A student locates a bug with the AI’s help rather than its fix, then catches a new bug the AI’s own fix introduces.
• Physics: which principle applies? (physics-which-principle.md). A student catches the AI solving a collision with the wrong conservation principle, a clean answer that is physically wrong.
• Mathematics: is this count right? (mathematics-counting-argument.md). A student audits a confident counting rule from the AI, catches that it double-counts by testing a small case, and finds the failing step.

These four sessions are the core library. Runnable notebook versions of the code-based sessions are planned as a separate track once the core is set.

How to use these

Read one for the moves, not the answer. Then try the same pattern on your own work: make a first attempt, ask the AI for a hint or a stress-test instead of a solution, do the work yourself, and check anything the AI states confidently.