The Cognitive Bootcamp Agreement

post by Raemon · 2024-10-16T23:24:05.509Z · LW · GW · 0 comments

Contents

  The Goal: Solve confusing, intractable problems
  The Curriculum
  Skills
  Multiple feedback-loops
  The Disclaimers (Important!)
    Prerequisites
    This will be very meta
    What's the evidence that this works?
      First Workshop 6-month followup
None
No comments

The Cognitive Bootcamp is intended to be fairly intense intense. I ask attendees to read through the Workshop agreement and a) doublecheck that the workshop is right for them b) let me know how we might want to tailor it to meet their needs.

The target audience is people who:

• have decision-making power on a large project, which tackles confusing problems;
• are not bottlenecked on executive function;
• and, who don’t have a sneaking suspicion they’re at risk of burnout. 

The commitment I’m asking if you come is that you spend basically the whole time either:

• aiming to think basically as hard as you can.
• napping / taking a walk / etc (preferably without devices).
• talking to me if the workshop feels off in some way.

Beforehand, you’ll send me your default plan for the next ~week, month or quarter (whatever the longest timescale you plan on). You’ll work on improving this plan.

Your primary goal, if you come, should be to learn at least one new skill, starting from "it’s too cumbersome to use productively", and aiming to reach "Juuuust fluent enough that you can start applying it to your day job, practicing so it becomes easy.”

I’ll present ~4 skills I think are valuable for solving confusing problems, and exercises that are helpful for grinding on them. You’re welcome to pick other skills, or other exercises, that you think will help you better. But, for each session, you need to clearly explain a) why I think the original exercise was important, b) why the thing you’ll do instead is better. (We’ll chat until we both understand each other and feel good about it)

Previously, I’ve found that giving people too much leeway in changing the curriculum results in the workshop losing focus, but too little leeway results in people not quite getting what they need. This is the middle ground that I've found works best.

The Goal: Solve confusing, intractable problems

Many of the most important problems in the world are confusing, intractable, and seem impossible. They are really important to solve anyway. 

But many people who try, end up in failure modes such as…

1. Rabbit-holing, fixating on an approach which doesn’t work, or takes too long.
2. Goodharting, substituting an easier problem, maybe without even noticing.
3. Despair. It’s impossible. You give up.
4. Zombified Agency, mechanically execute virtuous-seeming plans, but something inside you is dead and hollow and the plan probably won’t work and maybe you’ll hurt yourself.

I want you to leave the workshop with at least one new skill for solving impossible problems.

The Curriculum

The default curriculum focuses on “making better plans” via…

1. Generating multiple plans, so you don’t over-anchor on your first plan.
2. Tracking multiple goals, so you don’t over-anchor on your first goal.
3. Identifying your cruxes, and confidently deciding when to pivot or persevere.
4. Making cruxy predictions, so over time you can credibly, calibrated trust your intuition (and, know the limits of where you cannot)

An optional, higher level frame is  Fractal Strategy – thinking of plans and goals at multiple levels, tracking how smaller plans fit into bigger plans, and when it’s time to move up, down or sideways in plan/goal-space.

If you feel like you roughly have all these skills, you might still want to come to the workshop to have an environment that will help you do whatever cognitive practice you believe in, with good form, attending carefully to each step of the process.” (I’ll want to talk to you beforehand about your plan in detail)

Skills

The one skill I will require everyone to attempt at least twice is  generating metastrategies (see below). 

Each of the bullets in the Curriculum section has a corresponding skill. Some particular skills that will be presented that will come up if they seem helpful to a particular student are:

Generalizing Takeaways , i.e. asking “how could I have thought that sooner?”. We'll practice doing this both in-depth, and a rapid 5 minute version you can do every day.

Using externalized working memory, both because it increases the complexity of problems you’re capable of tackling, and because it makes it easier for me to see your thought process and offer advice.

Noticing metacognition, to identify when you’re in particular cognitive states so that you can learn about your mind and employ relevant skills/habits.

Grieving. Sometimes, there's a better plan available, but switching to it involves letting go of some kind of psychologically loadbearing beliefs.  Deliberate grieving [LW · GW] is a skill for that. (There will be an optional session for this Saturday night but it's more experimental)

Multiple feedback-loops

I don’t have perfect feedback-loops to tell you if this workshop is working for you. So, there are four different feedback-loop types, with different tradeoffs:

Predictions. Guess whether a given strategy will work, then see if you were right.

Toy Exercises . They only vaguely resemble your real problems, but you’ll know for sure whether you got the right answer in two hours.

Big Picture Planning. You’ll generate at least one new plan. You won’t really know if it’s good, but a) you’ll have intuitions about it, which are at least some information. And, b) you’ll make predictions about whether it’ll seem worth having thought about in a year.

Object-level work, in 1-hour block. Spend a few timeblocks doing object level work on your second likeliest plan. Each hour, you’ll make conscious choices about how to spend your time and attention. And then, reflect on whether that seemed useful. (in addition to crosstraining your skills on the practical object-level, this will help make your second-likeliest plan feel more real)

New Cognitive Strategies. Over the workshop, you will hopefully be identifying new strategies for problemsolving (or, realizing more significance of strategies you were familiar with but haven't been using nearly enough)

We'll also work on improving your daily, weekly and longer feedback loops in your real world work.

The Disclaimers (Important!)

Prerequisites

This is not an entry level rationality workshop. It has several prerequisites:

1. Executive function. You can sit down and think about a confusing problem and not immediately bounce off. Your problem should be more like “when I sit down to plan, I don’t know what to do” than “I have trouble sitting down to plan,” or “if I make a decision, I can’t follow it through.”
2. Project Ownership. You have control over an (at least somewhat) open-ended decisionmaking process i.e. you get at least some leeway to set priorities at your day job, or you have a lot of slack for ambitious hobbies.. A primary skill at this workshop is learning when to pivot. You need to be capable of deciding when you’re pivoting. When you leave the workshop, you have a project you expect to be applying the techniques to on a daily basis.
3. Self awareness. You can notice when pushing yourself hard is bad for you, instead of good for you. You have at least some awareness of when things are going subtly wrong and you need to slow down. I'll do my best to help you notice this sort of thing, but there’s a limit to how much I can help.

If you are at risk of burnout (if you’ve recently worked much harder than usual, or a nagging voice inside is worried about spending a weekend doing very intense thinking) I recommend you do not attend right now. You can register your interest for future workshops though.

This will be very meta

We are going to think about thinking.

We are going to apply feedback-loops to our feedback-loops.

While we’re doing that, I will be thinking about thinking about you thinking about those things (You don’t have to do that, just me).

Meta-level optimization is the mechanism by which, maybe, I expect people to get compounding returns on thinking, and there is a real possibility that this can be leveraged into dramatically better plans.

Well designed feedback loops are the grounding mechanism to check whether we’re doing masturbatory meta, vs useful meta. [4] [LW(p) · GW(p)]

I think there are basically two flavors of "I feel like this meta sucks". One is that you've lost track of your goals, you've spent more time meta-optimizing than makes sense, and your subconscious is (correctly) flagging that it's time to get back to work. Another is "it feels overwhelming, like you can't track what's going on."

I think it can be correct (nearterm) to stop if you're overwhelmed, but one of the skills the workshop is trying to impart is the ability to navigate complex problems without getting overwhelmed (complex metacognition included).

If you are feeling disoriented or annoyed or have a nagging feeling the current amount of meta won’t help:

1. First, pause (if  I’m the problem, say “hey Ray stop”)
2. Probably, go back to doing a more object level thing
3. But, also: consider getting out a sheet of paper / google-doc and writing things down so you can actually track what’s going on. Or, chatting with me about it and seeing if we can find a way to make it more manageable. [5] [LW(p) · GW(p)]

Part of the point of becoming fluent in “working memory extension” is that your meta processes are much easier to understand, and you can leverage them more strategically.

What's the evidence that this works?

I am talking a pretty big game. I think it’s an active ingredient for me to present content confidently, and for people to lean into a mindset where they trust it’ll work, or at least is worth trying.

But an important truth is that this workshop series does not (yet) have a strong empirical track record. If the curriculum does not intuitively make sense to you, I don’t think you should particularly believe in it.

I am trying to stick my neck out such that if the workshop is not working, it should be obvious (i.e. people will not be generating strategies that help), and over time if the workshop does not help, it should be clear that the people who take it don't end up standing on giant heaps of utility. And if I haven't gotten that to obviously work in ~a year, I'll give up.

Meanwhile: here are some facts about past participant experience:

First Workshop 6-month followup

When I ran the first workshop (which I charged $200 for), I asked the 9 participants what was the most they’d have paid for the workshop. Numbers ranged from $300 to ~$2000, on average $800.

When I asked again six months later, two “$400” people and one "$1500" changed their number to “$2000”.

• One of the $400 folk, because they changed their plans importantly afterwards. They weren’t sure how much credit to assign to the workshop, but it could be anywhere from $0 to $10,000, and "$2000 felt about right."
• The other $400 person, because they gained an major insight that was still seemed important to their worldview.
• The $1500 person explicitly changed their plan during the workshop, and after months later it seemed like the lessons had sunk in a more, and they also upped to $2000.

Meanwhile two people reported back they'd probably lowered their estimate:

• One (who originally said "$300" said "My gut says $200. My brain says potentially $0 or potentially $1000."
• Another who originally said $1500 person dropped to $0 because they realized they were too burned out to get value from the workshop (and maybe it hurt them).

When I ran the second workshop, average rating was $540, or $650 if you throw out one data point from someone (who rated it $0) who came for nonstandard reasons, and I wouldn’t have really expected to get much value out of it, if we’d talked a bit more. There was another person who also reported they were burned out [6] [LW(p) · GW(p)] and now wasn't a good time for the workshop, although I don't think they were actively hurt.

I'm raising rates now because a) I've now run four workshops, and am more confident in the curriculum, and b) I'm also baking following coaching into the program. I think the workshop *can* get someone to the point where they can practice on their own, but realistically I think most people will benefit from an environment where the skills are solidified and habits are reinforced.

I charge a fair amount of money because:

• People thinking it’s worth paying for is a crux of mine. If I didn’t think I could ultimately generate thousands of dollars worth of value for people, I would give up on the project. Charging money makes it harder to delude myself that I’m helping.
• Lightcone needs the money. And on the margin, we prefer to get money from people we are delivering value to. The cost roughly what we'd need for this to be sustainable.
• It filters for commitment. I want to filter for people who actually are going to  try to get hundreds or thousands of dollars worth of value from the workshop, and so will put more effort in.
• It soft-filters for the pre-requisites. Because the workshop isn’t a 101 workshop, I want to filter for people who have some degree of “already having your shit together”, for which “can afford to consider paying serious money for a workshop” is a proxy. (I realize that will exclude some people unnecessarily. But given the other goals, it seems like the right balance)

In general, if you didn’t feel like you got your money’s worth, I prefer to solve this by giving you some free followup coaching until you feel like you got your money’s worth.

If you earnestly tried to get a thousand dollars worth of value from the workshop, including explicitly strategizing with me on how to adapt it to you, and it didn’t really pay off and it seems like my general frame or skillset isn’t useful to you, chat with me and I’ll consider a refund.

0 comments

Comments sorted by top scores.