Skip to content

Roadmap

What exists, and what is deliberately not written yet. A topic listed here has no page yet; it is a claim about direction, not a stub.

Written

01_Knowing_What_You_Know — how do you know that you know? Three lessons after McGuire's Teach Yourself How to Learn, chapters 3 and 4: metacognition, count the vowels, studying vs learning.

02_Making_It_Stay — how do you learn so that it lasts? Spaced retrieval, interleaving, cognitive load, and, from Oakley and Sejnowski's Learning How to Learn, focused and diffuse thinking.

03_Believing_You_Can — what does a belief have to do with the evidence? Learned helplessness and day or night, after Zakrajsek's The New Science of Learning, chapters 1 and 2.

04_Planning_the_Work — one lesson, the buffer hour, and two graded pages, goals and schedules and learning with AI.

05_Body_and_Brain — six evidence-graded pages, no programs: how memory works, sleep, exercise, food and supplements, brain chemistry, mindset and motivation.

The rules

A lesson needs an argument and a program that demonstrates it, with no invented data or with data that says it is invented and has only the right shape. An evidence-graded page is for a claim no program can check; it needs a grade and a source for every claim, and it gives no doses, no supplement or drug advice and no medical advice. A topic that fits neither stays here.

Lessons with an obvious program, not yet written

  • Procrastination as hyperbolic discounting. A reward's value falls with delay, and falls faster near the present than far from it, so tonight's essay loses to tonight's film while next week's essay beats next week's film. A program can compute the crossing and show why a commitment made in advance (a scheduled block, the buffer hour) works when a resolution made at the time does not. Zakrajsek's chapter 2 names avoiding procrastination as a part of self-regulation, and this is the arithmetic under it.
  • Elaboration. The fourth of the big four has no lesson because it has no obvious program. A candidate: the number of retrieval routes to a fact as a graph, and how many cues survive when each edge fails with some probability. It would need honest modelling to avoid inventing the result.
  • Pretesting. A learner who guesses first against one who reads first, in the same counting style as learned helplessness. The mechanism is under-specified, so this needs care.
  • Feedback timing and the hypercorrection effect. A program can at least show what a proper scoring rule says about high-confidence errors: they cost most, so correcting them gains most. That would join metacognition to the finding.
  • Note-taking: Mueller and Oppenheimer's typists versus writers, and the replication that shrank it. A program could count how much of a lecture transcript fits at typing speed against handwriting speed, which is the mechanism they proposed.

Not planned as lessons

  • Memory palace, procrastination as habit, and the other chapters of Oakley and Sejnowski's Learning How to Learn that focused and diffuse thinking maps to existing lessons: no program shows them. Procrastination as hyperbolic discounting, above, is the one exception with a program idea.

  • Group study, office hours, test-taking tactics (McGuire's strategies 13, 27–32, 34, 35): advice with no program to show it. The strategies table lists them.

  • Sleep stages, exercise dose, diet and supplements as protocols. The evidence for anything finer than the graded pages already say is weak, and a page that told a reader when to exercise or what to take would be pretending otherwise.
  • Gestalt laws of perception (Zakrajsek's chapter 6): real, but about perception rather than study, and hard to make a claim about that a program tests.

Moved from the math library

This library began as chapter 12 of the math library ↗, Learning to Learn. The ten lessons kept their folder names, so an old URL such as 12_Learning_to_Learn/spaced_retrieval/ maps to 02_Making_It_Stay/spaced_retrieval/ here, and the math library's chapter page points across.