Page
ReadView sourceView history

Expert response heuristic for prompt writing

From Issawiki (static archive)

The expert response heuristic for prompt writing says that when writing spaced repetition prompts, the prompt should be answerable by an expert in the subject who knows the material very well but hasn't read the exact same source material you read. This forces prompts to be phrased in "timeless/contextless" ways that don't rely on a particular source material.

Pros:

Cons:

My current feeling is that maybe half of one's cards should be like this, while the other half can be more about specific source material.

There is a similar thing that sometimes happens in talks where the lecturer will ask the audience a question, and the audience sometimes gives good thoughts but the lecturer is looking for some specific answer, in a way that seems unfair to the audience. I feel like prompts that don't follow this heuristic can give this flavor.

I first had these thoughts after doing the prompts in Andy Matuschak's prompt writing guide. I read the guide pretty quickly, then just went for the prompts. I thought that since I've already been writing prompts for a couple years, and since the guide didn't seem to contain new material for me, that I could easily answer the prompts. I found that a lot of the prompts asked for pretty weird phrases/words like "vivid" and so forth, so that even though there was nothing really new in a sense, that I was being forced to stick to the very contingent facts appearing in the text.

The expert response heuristic seems more useful in "hard" topics like math and CS, and less useful in things like life advice or whatever (it's hard to write prompts about "what do you do in this situation" without referring to specific strategies that other experts might not know about? maybe the problem is that there is no agreed-upon discipline/answers here).

Some things, like notation (which can vary across textbooks) can't be phrased in this way.

With Tao Analysis Flashcards, I can try to phrase things such that people reading the textbook will retain the material well. And since I already have worked through the book, I know which pieces of knowledge will be more useful later on. But at the same time, I am also sticking closely to the text's material, so I fear the cards are not comprehensible to people who studied analysis from a different book.

See also

What links here