Through Time

Learning something new: what holds up and what is legend

Which ways of learning have evidence behind them, how not to trust the feeling of knowing, and how to study with AI without it doing your thinking.

If machines keep getting faster at picking up new things, the ability to learn for yourself becomes what matters most for a person, and that is our bet, not an established fact. You will find out what research supports and what it does not, learn to notice the illusion of knowing, and draw up a study plan for a busy adult. Keeping a skill you already have belongs to the course on thinking; this one is about new ones.

So far there is only this table of contents. Lessons appear one at a time, and we promise no dates. Feed of new lessons (Atom)

18 lessons, 5 parts

Part 1 What has evidence behind it

Only a few ways of studying have a strong record, and the most popular ones are not among them.

  1. Recall instead of rereading

    Why testing yourself beats rereading, though rereading feels nicer, and how to set it up with your own material.

    coming soon

  2. Gaps between sessions

    Ebbinghaus’s forgetting curve, and why the same work spread over days stays longer than the same work done in one sitting.

    coming soon

  3. Mixing up your practice

    Mixing often yields more though it feels worse, and it matters to know which material makes it worthwhile.

    coming soon

  4. Feedback can hurt

    On average feedback helps, yet it often does harm, and much depends on whether it speaks about the task or about you.

    coming soon

Part 2 Illusions of knowing

The feeling that you have understood is a poor guide to what you can do a week later, and you need not trust it.

  1. The feeling of learning

    Active classes feel less useful than a good lecture and teach more: why a difficulty can be desirable, as Robert Bjork puts it.

    coming soon

  2. Explain how a zipper works

    Try to explain a simple device and your rating of your own understanding drops: the illusion of depth as a ready self-test.

    coming soon

  3. What Dunning and Kruger actually showed

    The meme about fools that is not in the 1999 paper, and the dispute over whether the effect is a statistical artifact.

    coming soon

  4. The myth of learning styles

    No adequate basis has been found for matching lessons to “visual” or “auditory” learners, and plenty of people still believe in them.

    coming soon

Part 3 How a skill grows

Many loud things have been said about practice, and the data in this field is more modest and still in dispute.

  1. Deliberate practice

    A clear goal, difficulty just past your level and someone who can see the error, and why mere repetition does not count.

    coming soon

  2. Ten thousand hours: what Ericsson did not say

    How a rule from a popular book drifted from what Ericsson wrote, and how much practice really explains.

    coming soon

  3. Growth mindset without the legend

    The effects turned out small, and the argument now is over whom such programs help at all: what of the idea is worth keeping.

    coming soon

Part 4 AI as a tutor

The same model can speed up learning or replace it, and the difference lies in how you use it.

  1. When AI hurts learning

    A chatbot that hands over answers can raise your score in practice and lower it afterward: how to tell it is happening to you.

    coming soon

  2. When AI helps

    What sets a model tuned to teach apart from a plain one, and why a good result right after a session does not prove anything stayed.

    coming soon

  3. The two sigma legend

    What Bloom said about tutors in 1984, what later counts find, and how to read the promises of AI tutors with that in mind.

    coming soon

  4. How to study with a model

    Your own attempt first, then a hint, questions in place of answers and a test with the screen off: an order in which you still do the thinking.

    coming soon

Part 5 A course of your own

All of this counts only if it adds up to a plan you will carry out in an ordinary busy week.

  1. How to know you have learned

    A task with no hints, a week later and on new material: a test that is hard to fool, and without it you do not know whether you have moved.

    coming soon

  2. A study plan for one

    A goal stated as a concrete action, short sessions in the calendar, and a way of getting feedback when no teacher is around.

    coming soon

  3. A look ahead: a tutor for everyone

    If a personal AI tutor becomes ordinary, the rare thing will be the will to take on what is hard: that is our hypothesis, and long trials will test it.

    coming soon