Epistemic status: I’m describing a problem I have, which probably has an existing solution.
1.
Let’s say you’re reading an article, and there’s one sentence in particular that doesn’t stick out to you. The sentence seems intuitive, so your eyes slide over it and continue down the page.
This happens when your intuition matches with the sentence. The sentence follows some subconscious prediction. It does not surprise you. You extract no evidence from the sentence, and you don’t update your beliefs.
The sentence is a missed opportunity. Its presence narrows the possibilities for the territory (unless you really did predict those words in that order). Some update should happen.
The immediate territory here is, which words the author wrote. But this territory is readily mapped, hard to compress, and not applicable elsewhere. There are more interesting territories. What words would you write, if you had to continue the article? What is the space of true statements on the subject? What sentences best explain the topic to other people?
This might seem like a minor nitpick. We’re humans, not AIXI. We have finite spoons. We have better things to do than questioning our intuitions all day. But this can sometimes be a bigger issue.
2.
Habits are practiced actions. They reflect some long-term optimization[1], and cost no spoons. Similarly, intuitions are practiced thought patterns. They consist of cached thoughts, both object- and meta-level. If they reside in System 1, then they cost no inferential steps.
Intuitions are powerful tools. They reflect a long period of updating. We should make sure those updates are in the correct direction.
It can be hard to gauge your intuitions. If you’re actively learning a subject, you will have plenty of thoughts cached in System 2. This makes you overestimate the extent of your intuition, and consequently, underestimate the amount you need to update.
3.
Back to that article. If you skim over a sentence, you miss an opportunity to strengthen your intuitions. Maybe that’s ok; your intuitions are already strong enough, and you’re directing effort in a different direction. But if you skim over every sentence, you might as well not read the article. You should have some policy, which sentences to skim, and which to update on. This policy should favor updates when you’re actively learning the subject.
It’s the next day, and you’re in class discussing the article. Your classmate contributes a sentence or two. Her statements seem intuitive. They sound just like the article, which seemed intuitive. It doesn’t help that there’s some nuance you’re not hearing. If you don’t update on this intuitive information, it is an even bigger missed opportunity. Direct conversation is one of the best sources of updates.
You don’t contribute much to the discussion. After all, the article was intuitive, so there’s not much to say. This could be a symptom of weak intuition: you don’t have enough thoughts cached in System 1, so you can’t formulate statements in real-time. This deprives your classmates of updates from conversation with you.
4.
I don’t know the correct policy for when to spend effort on updates. Here’s what I do know:
We have plenty of research on habits on LW. Some of this could carry over onto intuitions.
When I read mathematics, I randomly stop and prove some statement from the previous, or from common knowledge. This varies between every statement, and 1 in 10 statements, depending on how strong my intuition is.
If it seems intuitive, you’re not building intuition
Epistemic status: I’m describing a problem I have, which probably has an existing solution.
1.
Let’s say you’re reading an article, and there’s one sentence in particular that doesn’t stick out to you. The sentence seems intuitive, so your eyes slide over it and continue down the page.
This happens when your intuition matches with the sentence. The sentence follows some subconscious prediction. It does not surprise you. You extract no evidence from the sentence, and you don’t update your beliefs.
The sentence is a missed opportunity. Its presence narrows the possibilities for the territory (unless you really did predict those words in that order). Some update should happen.
The immediate territory here is, which words the author wrote. But this territory is readily mapped, hard to compress, and not applicable elsewhere. There are more interesting territories. What words would you write, if you had to continue the article? What is the space of true statements on the subject? What sentences best explain the topic to other people?
This might seem like a minor nitpick. We’re humans, not AIXI. We have finite spoons. We have better things to do than questioning our intuitions all day. But this can sometimes be a bigger issue.
2.
Habits are practiced actions. They reflect some long-term optimization[1], and cost no spoons. Similarly, intuitions are practiced thought patterns. They consist of cached thoughts, both object- and meta-level. If they reside in System 1, then they cost no inferential steps.
Intuitions are powerful tools. They reflect a long period of updating. We should make sure those updates are in the correct direction.
It can be hard to gauge your intuitions. If you’re actively learning a subject, you will have plenty of thoughts cached in System 2. This makes you overestimate the extent of your intuition, and consequently, underestimate the amount you need to update.
3.
Back to that article. If you skim over a sentence, you miss an opportunity to strengthen your intuitions. Maybe that’s ok; your intuitions are already strong enough, and you’re directing effort in a different direction. But if you skim over every sentence, you might as well not read the article. You should have some policy, which sentences to skim, and which to update on. This policy should favor updates when you’re actively learning the subject.
It’s the next day, and you’re in class discussing the article. Your classmate contributes a sentence or two. Her statements seem intuitive. They sound just like the article, which seemed intuitive. It doesn’t help that there’s some nuance you’re not hearing. If you don’t update on this intuitive information, it is an even bigger missed opportunity. Direct conversation is one of the best sources of updates.
You don’t contribute much to the discussion. After all, the article was intuitive, so there’s not much to say. This could be a symptom of weak intuition: you don’t have enough thoughts cached in System 1, so you can’t formulate statements in real-time. This deprives your classmates of updates from conversation with you.
4.
I don’t know the correct policy for when to spend effort on updates. Here’s what I do know:
We have plenty of research on habits on LW. Some of this could carry over onto intuitions.
When I read mathematics, I randomly stop and prove some statement from the previous, or from common knowledge. This varies between every statement, and 1 in 10 statements, depending on how strong my intuition is.
Be wary when everything seems very intuitive.
Hopefully, the optimization was conscious, and in a positive direction.