When you already know the answer - Using your Inner Simulator

post by Neel Nanda (neel-nanda-1) · 2021-02-23T17:58:29.336Z · LW · GW · 4 comments

This is a link post for https://www.neelnanda.io/blog/42-inner-sim

Contents

  Introduction
  Pre-Hindsight
  Inner Simulator
  Examples
  Conclusion
None
4 comments

I first encountered the idea of Inner Simulators at a CFAR workshop. Credit to them, and read the relevant chapter of their Handbook to see their take. If you’re already familiar with the idea, or find this post slow, I expect you’ll still enjoy the list of examples at the end

Introduction

I am fascinated by framing effects - when phrasing a question in a different way, this can significantly change my immediate answer. Some of my favourite examples:

Personally, I find these effects are all easy to replicate and surprisingly strong. And this is super weird! Changing the way I frame a question can instantly make it much easier to answer well. And this can be applied instantly, in the moment, with little additional effort. Ways to improve my thinking on the five second level [LW · GW] are rare and incredibly powerful! I want to collect all the useful framing effects that I can, and make them reflexive parts of my thinking. 

This post is about one of my favourite framing effects, which boils down to ‘assume you already know the answer to a question, and let your intuition fill in the blanks’. This effect is insanely useful, and I use it multiple times a day. I find it hard to articulate the key mental move here, but people tend to find it fairly intuitive once they get it. So this post will consist of me attempting to describe the idea and gesturing at it in multiple ways, and then giving a long list of examples! (I consider the list of examples to be the most interesting part of this post)

Pre-Hindsight

I often send emails and, immediately after clicking send, realise I got something wrong, eg forgetting to include a point or to attach a file. This experience seems near-universal, but it’s also super weird! I have gained absolutely no new information, yet I suddenly have an insight that I lacked moments before. This is the sign of a framing effect! And it can be exploited. I imagine sending the email and then instantly realising I forgot something. And then I ask myself “what happened?”. And, often, this framing is enough to get my mind to fill in the blanks for what I forgot. This particularly example is sufficiently useful that Gmail now has an ‘undo’ feature for several seconds after sending an email.

This is an example of the CFAR technique of pre-hindsight (part of their broader technique of Murphyjitsu), and the core insight is that there is a significant difference between how we think about the past and the future. We think about the past with hindsight - things feel definite, and concrete, and it’s easy to generate explanations. But the future is vague and fuzzy - it’s easy to flinch away from it, and have no idea what’s going on. By imagining sending the email, I flip it from a future event to a concrete past event, and can now use the power of hindsight to guess at what went wrong.

I first encountered this technique at a CFAR workshop, as part of their class on Murphyjitsu, an approach to good planning (see notes for a workshop I gave on this or the CFAR handbook if you want to learn more). The idea is that any time you have a future intention, it’s hard to think about it, and to fully use your intuitions about what might go wrong. But if you concretely simulate the plan happening, and imagine it going wrong without fleshing out the details of what went wrong, you can now use hindsight to predict what went wrong. And that doing this is an effective way to use your intuitions about likely failure modes.

For example, I have an important job interview coming up, so I’m currently stressing out, preparing, and trying to ensure I know everything I might be expected to know. I imagine doing the interview and leaving feeling like I screwed up. When I ask myself ‘what went wrong?’, it feels obvious that I got stressed and anxious and made dumb mistakes. Now that I know this, I know that the bottleneck is not being unprepared, it’s getting too anxious about it. And I can instead focus my effort on relaxing and ensuring I go in in a good headspace.

I find pre-hindsight to be a super useful mindset for making good plans, but for the purposes of this post I want to focus on the key mental move of ‘assume your plan has failed, and get your intuition to fill in the blanks of what went wrong’. The key here is that it converts something unknown and abstract into something concrete, it assumes I know the answer, and it allows my intuitions to fill in the blanks. I find that this technique generalises far beyond just planning, and that it can be elicited with the right framing.

For example, I am often procrastinating because I have several things to be doing but no obvious highest priority. This means things feel vague and non-urgent, so I end up meandering. In order to start doing something I must first do the cognitive work of trading off and choosing the highest priority, which is effort, so I put it off. I can short circuit this by asking myself “suppose I picked one thing as my highest priority, and decided to focus on that. Which was it?” and then just running with the first thing to come to mind. 

Another example: I notice a problem in my life that’s bugging me, and I have vague thoughts on possible solutions but never really do anything about it and feel helpless. Eg, recently this was ‘being distracted at work’. I can resolve this by imagining myself two weeks from now and saying ‘I changed things up, and this problem no longer feels like an issue. What did I do?’, and letting my intuition fill in the blanks.

Inner Simulator

Zooming out a bit, what’s going on here? An experiment: Pick up a ball (or something similar), and toss it into the air. Watch the trajectory as it comes down, and try to catch it. When I do this, I find it easy to guess at the future trajectory and to put my hands in the right place to catch it. To do this explicitly and verbally, I’d need to solve a bunch of differential equations. But here, some part of my mind is doing all of this instantly and intuitively, without being under my conscious, verbal control. 

I call this part of my mind my Inner Simulator, and this is the part of my mind doing all the work in pre-hindsight. It’s a part of my mind that’s really good at simulating the world, and a powerful source of intuition when elicited in the right way. Using my Inner Sim creates feelings of certainty and concreteness, it’s easily able to fill in the blanks.

And learning how to use my Inner Sim right is extremely valuable! It is instant and certain, so can cut out a lot of wasted motion and wishy-washy hand-waving. It feels confident and certain, so can add motivation, decisiveness and resolve when facing uncertainty. It lets me use all of the knowledge at my disposal, helping me to identify blind spots and better understand the world.

But the reason I need to draw your attention to it is that the Inner Sim is not under your explicit, conscious control. It provides intuitive information, but to use it well, the question needs to be asked in a specific way. It’s very much part of System 1, not System 2. Some tips for how to elicit it well: (Note that these are just speculative guidelines that I find helpful, your mileage may vary. I often find that these aren’t necessary)

Examples

That was all a load of theory and speculation. But the main reason I am excited about this effect is that I use it multiple times a day, and it feels applicable to a wide range of scenarios in my life. So here’s a long list of various ways I find it helpful - hopefully at least one of them resonates!

Conclusion

Hopefully at least one of those examples resonated! I tried to strongly err on the side of completion - the key message I want you to take is that this framing is extremely widely useful, and can help in a bunch of different contexts. 

A final point: The reason I am excited about this idea is that it can be reflexive. Something that sticks, and is used in the right situations to help you reach the right answer. But for it to reach this point, it needs to practiced. I highly recommend actually trying out the examples that resonated best, and paying attention to exactly what the mental move feels like.

Further, I find it useful to have it be the default to use this framing. Eg, including questions in my weekly review like ‘what am I currently missing?’. Or having a checklist every time I start a project, that get me to do pre-hindsight.

Exercise: Pick three of your favourite examples from the above. How can you ensure that using these becomes a reflex? How can you build these into your life by default?

4 comments

Comments sorted by top scores.

comment by G Gordon Worley III (gworley) · 2021-02-23T21:33:03.132Z · LW(p) · GW(p)

One common way I "invoke the inner simulator" is by thinking "what would someone tell me I should have tried if I'm going to report to them something failed (and possibly ask for their help)?". This works surprisingly well, and ends up being similar to rubber ducking in some ways just without actually needing another person.

I honed this skill on StackOverflow, often answering my own questions in the process of writing them up to ask them, but find it useful in lots of situations.

Replies from: neel-nanda-1
comment by Neel Nanda (neel-nanda-1) · 2021-02-24T08:40:19.547Z · LW(p) · GW(p)

Nice, I really like the approach of 'write up a concrete question -> assume I received a helpful answer -> let my inner sim fill in the blanks about what it says'

comment by Shamash · 2021-02-23T23:13:09.675Z · LW(p) · GW(p)

This seems like it could be a useful methodology to adopt, though I'm not sure it would be helpful for everyone. In particular, for people who are prone to negative rumination or self-blame, the answer to these kinds of questions will often be highly warped or irrational, reinforcing the negative thought patterns. Such a person could also come up with a way they could improve their life, fail to implement it, and then feel guilty when their reality fails to measure up to their imagined future. 

On the other hand, I'm no psychotherapist, so it may just be the opposite. Maybe asking these questions to oneself could help people break out of negative thought patterns by forcing certain conditions? I'd appreciate other people's take on this subject. 

Replies from: neel-nanda-1
comment by Neel Nanda (neel-nanda-1) · 2021-02-24T08:51:57.931Z · LW(p) · GW(p)

As a single point of anecdata, I personally am fairly prone to negative thoughts and self-blame, and find this super helpful for overcoming that. My Inner Simulator seems to be much better grounded than my spirals of anxiety, and not prone to the same biases.

Some examples:

I'm stressing out about a tiny mistake I made, and am afraid that a friend of mine will blame me for it. So I simulate having the friend find out and get angry with me about it, and ask myself 'am I surprised at this outcome'. And discover that yes, I am very surprised by this outcome - that would be completely out of character and would feel unreasonable to me in the moment.

I have an upcoming conversation with someone new and interesting, and I'm feeling insecure about my ability to make good first impressions. I simulate the conversation happening, and leaving feeling like it went super well, and check how surprised I feel. And discover that I don't feel surprised, that in fact that this happens reasonably often.

Such a person could also come up with a way they could improve their life, fail to implement it, and then feel guilty when their reality fails to measure up to their imagined future. 

This seems like a potentially fair point. I sometimes encounter this problem. Though I find that my Inner Sim is a fair bit better calibrated about what solutions might actually work. Eg it has a much better sense for 'I'll just procrastinate and forget about this'. On balance, I find that the benefits of 'sometimes having a great idea that works' + the motivation to implement it far outweighs this failure mode, but your mileage may vary.