Hammertime Day 10: Murphyjitsu

post by alkjash · 2018-02-07T18:40:00.679Z · LW · GW · 11 comments

Contents

  Day 10: Murphyjitsu
    Inner Sim
    Welp Mentality
    Murphyjitsu
  Daily Challenge
None
11 comments

This is part 10 of 30 in the Hammertime Sequence. Click here for the intro.

Like, so pessimistic that reality actually comes out better than you expected around as often and as much as it comes out worse. It’s actually really hard to be so pessimistic that you stand a decent chance of undershooting real life.

Later in the day I will put up an open thread about the first cycle of Hammertime.

We finish up the first cycle with another post on planning. Murphyjitsu is CFAR’s method for planning which asks us to try to be so pessimistic as to undershoot real life.

Day 10: Murphyjitsu

Murphy’s Law states that anything that can go wrong will go wrong.

For our Mandarin-speaking readers, here’s a useful mnemonic: Murphy transliterates as 墨菲 (mo fei), which is homophonous to 莫非, “what if?” That’s why I think of Murphy’s Law as the What If Law.

In the course of making plans, Murphyjitsu is the practice of strengthening plans by repeatedly envisioning and defending against failure modes until you would be shocked to see it fail. Here’s the basic setup of Murphyjitsu:

  1. Make a plan.
  2. Imagine that you’ve passed the deadline and find out that the plan failed.
  3. If you’re shocked in this scenario, you’re done.
  4. Otherwise, simulate the most likely failure mode, defend against it, and repeat.

The first important sub-skill of Murphyjitsu is Inner Sim – the ability for System 1 to simulate failure modes.

Inner Sim

I have the suspicion that everyone is secretly a master at Inner Sim, the ability to instantly simulate failure. Imagine a friend declares to you their New Year’s Resolution: to write a novel, to go on a keto diet, to write a month-long sequence on instrumental rationality.

Now, listen for that internal scoffing – your System 1 instantly proliferates the future with all manner of obstacles. That’s Inner Sim at work.

If you’re anything like me, Inner Sim is better at predicting other people’s failure modes than your own. The mental move that helps apply Inner Sim introspectively is essentially Outside View: take your plan and imagine another person made them. What will go wrong?

Welp Mentality

Inner Sim does surprisingly little on its own.

I had a conversation with a rationalist friend (let’s call him “Alex”) that went something like this:

Alex: What’s bothering you?

Me: I’ve been terribly unproductive. I’m procrastinating on fellowship essays … they’re due in two weeks, and every time I think about math these essays pop up in my head.

Alex: Why?

Me: I essentially finished them, but I still have to edit it. Copy-editing is so tedious, and every run I make through my writing, it looks even more awkward than it did the previous time.

Alex: What do you predict will happen?

Me: Well … I’m going to put the essays off until two days before the deadline, edit for ten minutes when I start feeling the pressure, and submit them. Until then, I won’t get any research done.

Alex: So…?

Me (shrugs): Sucks, right?

Alex breaks down in laughter.

I call this Welp Mentality. Welp Mentality is noticing that your plans are likely to fail catastrophically, or run overtime, or take 10x as much effort as you thought, and then shrugging noncommittally. Welp.

Welp Mentality is knowing and accepting as a fact of life that every build will release two months late. That you’ll end up half-assing problem sets and essays starting midnight before the deadline. That your current exercise plan will probably peter out. I had an old motto for Welp Mentality: “Due tomorrow? Do tomorrow.”

Murphyjitsu

Murphyjitsu is the astounding notion that if you can predict a failure mode, you can do something about it!

If your builds release two months late every time, you can move the release date, or cut features, or hire more engineers. If you know you’re only going to spend six hours on a problem set the night before the due date, at very least you might as well just set a six-hour Yoda Timer for it, do it at a convenient time, and submit whatever you end up with.

In my fellowship essay case, I decided to spend ten minutes editing and submit the thing immediately. The relief of getting two weeks of my life back was palpable.

Pick a plan you have for the near future. Murphyjitsu it. Pull out all the stops: Arrange social pressure to keep you on track. Double the amount of time you spend. Set calendar and phone reminders. Murphyjitsu only stops when you would be shocked if the plan fails.

Daily Challenge

Murphyjitsu a central life goal. Are there glaring failure modes you haven’t defended against?

11 comments

Comments sorted by top scores.

comment by SquirrelInHell · 2018-02-07T20:14:50.285Z · LW(p) · GW(p)
I have the suspicion that everyone is secretly a master at Inner Sim

There's a crucial difference here between:

  • good "secretly": I'm so good at it it's my second nature, and there's little reason to bring it up anymore
  • bad "secretly": I'm not noticing what I'm doing, so I can't optimize it, and never have
comment by tcheasdfjkl · 2018-08-07T05:24:48.039Z · LW(p) · GW(p)

When I first tried to learn to do Murphyjitsu a few months back, I kept having the problem that the most likely failure scenario of any plan was "I just run out of energy and focus and/or get really anxious and can't do anything", and this seems impossible to completely reliably prevent, and so I could never get to a point where I would actually be shocked if a plan failed.

Nevertheless it's a good technique. But I've decided that my goal with it isn't to make my plan actually failproof but to make it as close to failproof as I actually can - which is actually closer to what it achieves, I just find I have to make that explicit for it to not feel like I'm fooling myself. And of course it also helps to think about what are the things that make it more likely that I'll run out of energy (e.g. too many other plans, staying up late the night before) and try to minimize those. (That part is easier to do for me now than it was a few months ago, since I am currently not swamped with more plans than I can carry out and have more freedom/ability to arrange my plans in a way that's likely to succeed.)

Replies from: Raemon, Regex
comment by Raemon · 2018-08-07T07:45:37.662Z · LW(p) · GW(p)

Just wanted to say I’ve appreciated reading along with your exploration of this sequence.

Replies from: tcheasdfjkl
comment by tcheasdfjkl · 2018-08-07T17:08:04.655Z · LW(p) · GW(p)

Thanks :)

comment by Regex · 2020-06-22T01:10:12.518Z · LW(p) · GW(p)

I expanded 'shocked at failure' into:

The plans you make work.

When they fail, it is because of one of the following reasons:

  • a predicted reason (you took a risk / made a tradeoff, saw it as low probability or unavoidable)
  • violation of an explicit assumption (encryption used is secure, won't crash on the way to the airport)
  • a black swan (coronavirus, 9/11, stock market crash)

When they fail for reasons other than these, you are extremely surprised and can point to exactly what about your worldview and anticipations misled you.

comment by theme_arrow · 2021-01-12T07:06:38.011Z · LW(p) · GW(p)

This was a super interesting exercise! I chose doing well in the next class in my master's degree as the goal. I have generally done well so far, but I'm working full-time while going to school (because work is paying for it), so every semester is a bit of a struggle. What was unexpected about murphyjitsuing this goal was the way in which there were basically two categories of possible problems I found. The first were the easy, well-contained problems (like "I generally submit assignments at the last minute, making it easy for technical difficulties to mess things up"). That's the kind of problem that can be solved via yoda timers, reminders etc. The second kind is the big life problems, like "I could lose motivation," or "I could have a serious health problem." These are much harder to defeat through planning, and can only be hit more indirectly. 

I found the second category of problems to be disheartening in a way, since there's not an easy way to fight them off. But they're also at least a little bit motivating, because it helps me see the connection between everything else in the bug list and this specific goal. Getting more regular sleep, going for more walks, and doing a better job of keeping up with friends will all help in little ways in guarding against these bigger risks. 

comment by Qiaochu_Yuan · 2018-02-07T20:25:24.064Z · LW(p) · GW(p)

I wonder if planning fallacy (or rather, the incentive to not do anything about it by default) is mostly a social phenomenon. Telling a group optimistic estimates of how long it will take a project to get done could help improve the group's commitment to the project. Telling your teacher and your parents optimistic estimates of how long it'll take to finish your homework keeps them off your back. Etc. etc.

(Testable prediction: autism should correlate with lack of susceptibility to planning fallacy. Found a random link in support, for what that's worth.)

Replies from: DoubleFelix, alkjash
comment by DoubleFelix · 2020-08-13T00:35:55.653Z · LW(p) · GW(p)

People are much better at not committing the planning fallacy (in the underestimating direction, at least) when estimating other people's timelines for a given task, so I'd say absolutely it's a social thing. There's immense social pressure to give a rosy view of yourself; you wouldn't want to disappoint, would you?

comment by alkjash · 2018-02-07T21:11:09.663Z · LW(p) · GW(p)

This sounds partially true, but I think optimism helps people get started on things even in the absence of groups. Replace "group commitment" by "your own commitment" and "telling other people" by "self-deception" and it still should work.

comment by itavero · 2020-12-02T01:03:07.803Z · LW(p) · GW(p)

This technique has stuck with me since I read about it a year ago. It's very useful to do a quick pass with it - a single iteration can make a plan ~1.2x more likely to succeed. I tend to do the "pre-mortem" in my head quickly when using it.

I know I'm not using it to its full power. This time, for my central life goal, I'm going to fully write it out and iterate until I would be shocked. It might take a long time but this is worth it. Plus, I'll probably learn a lot about this technique and what does or doesn't work.

I already started but I'll probably finish up tomorrow. Since I resolved (in the spirit of Sunk Cost Faith) to comment on this every day, I'll leave this here for now and edit my comment tomorrow to include what I've learned.

edit: I did about 10 iterations of my Very Important Plan today but I'm still not done. I'll update tomorrow.

comment by silentbob · 2020-10-26T08:07:01.573Z · LW(p) · GW(p)

Where I find Murphyjitsu most useful is in the area of generic little issues with my plans that tend to come up rather often. A few examples:

  • forgetting about working on the goal in time, due to lack of a reminder, planning fallacy etc.
  • the plan involving asking another person for a favor, and me not feeling too comfortable about asking
  • my system 1 not being convinced of the goal, requiring more motivation / accountability / pressure
  • the plan at some point (usually early on) requiring me to find an answer to some question, such that the remaining plan depends on that answer; my dislike for uncertainty ironically often causes me then to just flinch away from that whole plan as opposed to just trying to find that one important answer

It's arguably more of a checklist than real "by the book Murphyjitsu", but still, taking a goal and going through these things, trying to figure out the most trivial and easy to fix issues with the plan, often allows me to increase the likelihood of achieving a goal by 10-20% with just a few minutes of work.