Inversion

Published: 2024-02-26

Inversion is necessary but insufficient to produce great outcomes. This is particularly (and possibly only) true in domains where there is some non-trivial probability mass on the far right of the distribution.

The motivating example of this phenomena was "cleaning" my personal email. When doing it, I often think of inverting as the motivating factor for doing to - the decision rule looks like: Avoid not cleaning your email, because it will cause you anxiety about tail risks to your life due to operational failures.

It's a task that is necessary to make sure the operations of my life continue to run in a relatively consistent fashion (mortgage is paid, other house bills are paid, emails from old colleagues about new opportunities are responded to etc), but by itself, it will not produce within the broader scope of my life the kind of right-tail outcomes that I wish to produce (and to some degree, that I identify my self as being able to produce).

There are many other tasks in my day-to-day life that fit the mold of Avoid not doing X, because you know doing X will make you feel better and produce marginally better outcomes or protect you from some critical downside. Using this mold is such an effective tool that it's hard not to apply it to everything - Avoid not writing code for Y hours a day, because ... or more interestingly, Avoid not doing your Anki spaced repetition practice for the day, because you know that doing it will strengthen your memory and generally increase your abilities.

To accomplish some kinds of tasks (notably, I believe, "building" new things), this model of inversion simply doesn't work well - there needs to instead be a motivating vision. Picking the correct model out of these 2 options seems like an underrated leverage point, given my personal bias for the inversion model.