This is an accounting of various thoughts I had when reading the Babble & Prune sequence.
Encouraging Babble
It is ironic, to me, that the Babble & Prune sequence ends with a call to exclude epistemic status tags from posts:
What happens to the reader when every post starts, “epistemic status: mostly true with a chance of rain?”
[...]
Instead of conditioning readers to hate us, I propose we return to a saner time, where the fact that your words are but a pale wavering shadow of the grand, mysterious truth in your heart is the default assumption about human communication. Where truth is a dance of successive approximations yet no step in that dance requires adult supervision. Where quibbles over certitude are banished to the comments section where they belong.
I interpret Alkjash as thinking that, if the standards were just lower in the first place, people would feel free to write more, which would get us further in the end (because the good stuff can bubble to the top).
Indeed, my impression is that the LessWrong team has worked to make LessWrong a space where people feel they can share raw thoughts rather than requiring everything to be carefully refined.
But in my opinion, the “epistemic status” flags, and similar tools, help rather than hurt.
The New Twitter Account Problem
There is a phenomenon—let me know if you think of a better name—which I call the new twitter account problem. My personal interaction with Twitter has been as follows:
Start a new twitter account. Follow a few people I like. Basically no followers. No pressure! I can say whatever I want! Dump some random thoughts into the new twitter account over the next few days.
Now I’m getting follows, likes, and shares. Positive reinforcement! This encourages me to keep it up.
Follows slowly climb, and I’m getting more and more interaction. Now I feel like there’s something at stake, and I have to do my best. I write some good tweets, but I gradually tweet less and less.
I make an alt account for “worse thoughts”, starting the cycle back at step 1. Gradually the alt account becomes my main account, and I have the same problem again, and need a third account, and a fourth… Meanwhile, the followers of my previous accounts wonder why I’m not active anymore.
LessWrong 1.0 had a similar problem. The frontpage got “too good” for people to feel like they could really post on. The “discussion post” was invented. A lot of activity moved to discussion. But then Discussion got too good, and people felt like they had to have something good even to post it in Discussion. Open Threads were created within Discussion, as the new low-bar-to-entry forum.
I’ve heard that a similar pattern has also played out on other discussion platforms. Increasing layers of “no really, it’s OK to post raw thoughts” are created as old layers get too respectable.
This is probably crazy, but...
In in-person discussions, a good way to lessen this problem is to preface statements with qualifiers like “This is a dumb question, but...” or “Here’s a crazy idea:” or “I don’t endorse this, but I was thinking...” and so on. This creates a kind of protection for the speaker. They still get credit for good ideas; but if the idea really was bad, they take less of a hit.
This doesn’t work as well for longform text discussions, because readers know you had time to think things through. But [epistemic status:] flags can play a similar role. If you don’t feel like an idea is good enough, you preface it with [epistemic status: super dumb] or whatever, and take some comfort in the fact that if the reader doesn’t like what you wrote, they were warned.
I think having this norm is much better than attempting to sustain a norm that everything is “epistemic status: raw thoughts” by default.
Say Random Things
OK, so epistemic status flags are one way to combat the new-twitter-account problem. Do we have any other tools?
One tool is to purposefully lower your standards. I believe the book IMPRO includes an exercise in which you point to random things around you and call them absolutely wrong things; (point to tree) “there’s a lamp-post” (point to grass) “there’s some spaghetti” etc. Perhaps saying absurd things on purpose helps prove to your s1 that nothing horrible will happen if you say something wrong. Another example of this is Allie Brosh drunk posting (notable because the explains the thought process behind it). This is commonly called shitposting. Unfortunately, Allie Brosh wrote less after that experiment, so it’s not clear that it had a positive impact. Anecdotally, I’ve heard that a friend had dramatically positive results with the IMPRO version.
Master All Levels
I don’t think we should just work on improving our babble, though. I think it’s really important to aspire to higher and higher standards. I want to become stronger! How can we reconcile this with the dampening effect high standards can create?
Meta-Perfectionism
The true answer is that if “holding yourself to a high standard” makes you do/say too little, then your concept of “standard” is broken. We intuitively reason based on blame/guilt, which makes improper inaction seem less bad than improper action. If we could free ourselves of that mode of reasoning, perhaps we could just not have the new-twitter-account problem in the first place.
It’s easy to paralyze yourself if you try to do the “right thing.” There’s always more uncertainty to be had. There’s always more information you could gather. It’s hard to become confident that you’re doing the right thing. This can lead to paralysis, and persistent inaction.
It’s much easier, I think, to stop asking “is this action the right action to take?” and instead ask “what’s the best action I can identify at the moment?”
Sometimes, the best action you can identify is “search for more alternatives.” Sometimes, it’s “study more” or “learn more.” Sometimes, it’s a specific action. The nice thing is that “what’s the best action I can find in the next five minutes?” always has a concrete answer. If you search for that, instead, you won’t get paralyzed.
He elaborates more on similar thoughts in Deliberate once. But more relevantly to Babble&Prune, his post Half-assing it with everything you’ve got describes the mindset required to orient perfectionism at the meta-problem of avoiding overmuch perfectionism.
Maximum Payoff Per Effort
We face two problems:
Investing the minimal amount of time/effort to get the results we want.
Getting the best results we can given the time/effort.
Nate focuses on the first problem, describing how some situations call for a measured effort, while other situations call for an all-out effort. But we also want to do the second: the more favorable our payoff per time/effort spent, the less we need to spend on things calling for a measured effort, and the more results we get when we go all-out.
One thing university art classes do is force you to spend a long time on a single drawing. This teaches you how good you can get if you spend a lot of time on a single piece. It also teaches you how to fruitfully spend that much time on a piece.
But to grow as an artist, you also need to learn speed. You shouldn’t carefully plan every drawing you do. Artists also practice gesture drawings, in which you have a very short amount of time to capture the general pose of a model. So, a deliberate practice of art involves practicing at all time-scales.
There’s probably a natural slow-to-fast progression for many skills, where you need to go slow at first to be able to do it well, but can do it faster and faster after that.
Training Babble, Training Prune
Your prune should not just accept/reject. It should have degrees. You should be able to step it up or down appropriately. Furthermore, you want your prune to give you specific feedback—not just “that’s bad”, but “that’s mismatched”, etc. I suggest listening to your internal critic with a felt sense / focusing lens. You may find that you know more than you thought about what makes for good work.
You want your babble to be able to conform to the highest standards it can while remaining creative. You want it to be like GPT-3, not entirely random words. So pay attention to what your taste says about your babble—the trick isn’t to cast your standards aside, but rather to pay attention to them without letting them stop babble. Give your babble a little breathing room from your prune.
Alkjash mentioned that babble is not generating things with independent randomness, but rather, is more like a random walk in concept-space. A lot of the skill of good babble is in generating good mutations of ideas, not just good ideas. You can see this as you practice babble on longer time-scales. Give yourself more time to write a sentence, and you may see yourself go through several mutations of that sentence before settling on one to write. Part of what it means to spend more time on a drawing or painting is giving yourself more time to plan each line, and more time to fiddle with it, seeking improvements.
My main point here is that improving babble doesn’t mean reducing prune. Alkjash sometimes speaks as if it’s just a matter of opening the floodgates. Sometimes people do need to just relax, turn off their prune, and open the floodgates. But if you try to do this in general, you might have initial success but then experience backlash, since you may have failed to address the underlying reasons why you had closed the gates to begin with.
The Many Gates
I think one of the most useful models in Babble & Prune was the three gates.
Actually, after starting this section I remembered that I already wrote my thoughts on this in the Babble & Prune section of Capturing Ideas, particularly the part about developing ideas. Go read that if you want a further elaboration of why just opening the floodgates isn’t exactly the goal.
Babble & Prune Thoughts
This is an accounting of various thoughts I had when reading the Babble & Prune sequence.
Encouraging Babble
It is ironic, to me, that the Babble & Prune sequence ends with a call to exclude epistemic status tags from posts:
I interpret Alkjash as thinking that, if the standards were just lower in the first place, people would feel free to write more, which would get us further in the end (because the good stuff can bubble to the top).
Indeed, my impression is that the LessWrong team has worked to make LessWrong a space where people feel they can share raw thoughts rather than requiring everything to be carefully refined.
But in my opinion, the “epistemic status” flags, and similar tools, help rather than hurt.
The New Twitter Account Problem
There is a phenomenon—let me know if you think of a better name—which I call the new twitter account problem. My personal interaction with Twitter has been as follows:
Start a new twitter account. Follow a few people I like. Basically no followers. No pressure! I can say whatever I want! Dump some random thoughts into the new twitter account over the next few days.
Now I’m getting follows, likes, and shares. Positive reinforcement! This encourages me to keep it up.
Follows slowly climb, and I’m getting more and more interaction. Now I feel like there’s something at stake, and I have to do my best. I write some good tweets, but I gradually tweet less and less.
I make an alt account for “worse thoughts”, starting the cycle back at step 1. Gradually the alt account becomes my main account, and I have the same problem again, and need a third account, and a fourth… Meanwhile, the followers of my previous accounts wonder why I’m not active anymore.
LessWrong 1.0 had a similar problem. The frontpage got “too good” for people to feel like they could really post on. The “discussion post” was invented. A lot of activity moved to discussion. But then Discussion got too good, and people felt like they had to have something good even to post it in Discussion. Open Threads were created within Discussion, as the new low-bar-to-entry forum.
I’ve heard that a similar pattern has also played out on other discussion platforms. Increasing layers of “no really, it’s OK to post raw thoughts” are created as old layers get too respectable.
This is probably crazy, but...
In in-person discussions, a good way to lessen this problem is to preface statements with qualifiers like “This is a dumb question, but...” or “Here’s a crazy idea:” or “I don’t endorse this, but I was thinking...” and so on. This creates a kind of protection for the speaker. They still get credit for good ideas; but if the idea really was bad, they take less of a hit.
This doesn’t work as well for longform text discussions, because readers know you had time to think things through. But [epistemic status:] flags can play a similar role. If you don’t feel like an idea is good enough, you preface it with [epistemic status: super dumb] or whatever, and take some comfort in the fact that if the reader doesn’t like what you wrote, they were warned.
I think having this norm is much better than attempting to sustain a norm that everything is “epistemic status: raw thoughts” by default.
Say Random Things
OK, so epistemic status flags are one way to combat the new-twitter-account problem. Do we have any other tools?
One tool is to purposefully lower your standards. I believe the book IMPRO includes an exercise in which you point to random things around you and call them absolutely wrong things; (point to tree) “there’s a lamp-post” (point to grass) “there’s some spaghetti” etc. Perhaps saying absurd things on purpose helps prove to your s1 that nothing horrible will happen if you say something wrong. Another example of this is Allie Brosh drunk posting (notable because the explains the thought process behind it). This is commonly called shitposting. Unfortunately, Allie Brosh wrote less after that experiment, so it’s not clear that it had a positive impact. Anecdotally, I’ve heard that a friend had dramatically positive results with the IMPRO version.
Master All Levels
I don’t think we should just work on improving our babble, though. I think it’s really important to aspire to higher and higher standards. I want to become stronger! How can we reconcile this with the dampening effect high standards can create?
Meta-Perfectionism
The true answer is that if “holding yourself to a high standard” makes you do/say too little, then your concept of “standard” is broken. We intuitively reason based on blame/guilt, which makes improper inaction seem less bad than improper action. If we could free ourselves of that mode of reasoning, perhaps we could just not have the new-twitter-account problem in the first place.
In The best you can, Nate Soares writes:
He elaborates more on similar thoughts in Deliberate once. But more relevantly to Babble&Prune, his post Half-assing it with everything you’ve got describes the mindset required to orient perfectionism at the meta-problem of avoiding overmuch perfectionism.
Maximum Payoff Per Effort
We face two problems:
Investing the minimal amount of time/effort to get the results we want.
Getting the best results we can given the time/effort.
Nate focuses on the first problem, describing how some situations call for a measured effort, while other situations call for an all-out effort. But we also want to do the second: the more favorable our payoff per time/effort spent, the less we need to spend on things calling for a measured effort, and the more results we get when we go all-out.
One thing university art classes do is force you to spend a long time on a single drawing. This teaches you how good you can get if you spend a lot of time on a single piece. It also teaches you how to fruitfully spend that much time on a piece.
But to grow as an artist, you also need to learn speed. You shouldn’t carefully plan every drawing you do. Artists also practice gesture drawings, in which you have a very short amount of time to capture the general pose of a model. So, a deliberate practice of art involves practicing at all time-scales.
There’s probably a natural slow-to-fast progression for many skills, where you need to go slow at first to be able to do it well, but can do it faster and faster after that.
Training Babble, Training Prune
Your prune should not just accept/reject. It should have degrees. You should be able to step it up or down appropriately. Furthermore, you want your prune to give you specific feedback—not just “that’s bad”, but “that’s mismatched”, etc. I suggest listening to your internal critic with a felt sense / focusing lens. You may find that you know more than you thought about what makes for good work.
You want your babble to be able to conform to the highest standards it can while remaining creative. You want it to be like GPT-3, not entirely random words. So pay attention to what your taste says about your babble—the trick isn’t to cast your standards aside, but rather to pay attention to them without letting them stop babble. Give your babble a little breathing room from your prune.
Alkjash mentioned that babble is not generating things with independent randomness, but rather, is more like a random walk in concept-space. A lot of the skill of good babble is in generating good mutations of ideas, not just good ideas. You can see this as you practice babble on longer time-scales. Give yourself more time to write a sentence, and you may see yourself go through several mutations of that sentence before settling on one to write. Part of what it means to spend more time on a drawing or painting is giving yourself more time to plan each line, and more time to fiddle with it, seeking improvements.
My main point here is that improving babble doesn’t mean reducing prune. Alkjash sometimes speaks as if it’s just a matter of opening the floodgates. Sometimes people do need to just relax, turn off their prune, and open the floodgates. But if you try to do this in general, you might have initial success but then experience backlash, since you may have failed to address the underlying reasons why you had closed the gates to begin with.
The Many Gates
I think one of the most useful models in Babble & Prune was the three gates.
Actually, after starting this section I remembered that I already wrote my thoughts on this in the Babble & Prune section of Capturing Ideas, particularly the part about developing ideas. Go read that if you want a further elaboration of why just opening the floodgates isn’t exactly the goal.