What is of increasing marginal value, and what is of decreasing?

See also: this comment

There’s a key question that should govern the way we engage with the world: “Which things have increasing marginal returns and which things have decreasing marginal returns.”

Sometimes people are like “I’ll compromise between what I like doing and what has impact, finding something that scores pretty good on both.” Or they’ll say, “I was already planning to get a PhD in [field]/ run a camp for high schoolers / do personal development training / etc. I’ll make this more EA on the margin.”

They are doomed. There’s a logistic success curve, and there are orders of magnitude differences in the impact of different interventions. Which problem you work on is by far the most important determinant of your impact on the world, since most things don’t matter very much at all. The difference between the best project you can find and some pretty good project, is often so large as to swamp every other choice that you make. And within a project, there are a bunch of choices to be made, which themselves can make orders of magnitude of difference, often the difference between “this is an amazing project and this is basically worthless.”

By deciding to compromise, to only half-way optimize, you’re knowingly throwing away most of your selection pressure. You’ve lost lost pretty much all your hope of doing anything meaningful. In a domain where the core problems are not solved by additive incremental improvements, a half-assed action rounds down to worthless.

On the other hand, sometimes people push themselves extra hard to work one more hour at the end of the day when they’re tired and flagging. Often, but not always, they are pwnd. Your last hour of the day is not your best hour of the day. Very likely, it’s your worst hour. For most of the work that we have to do, higher quality hours are superlinearly more valuable than lower quality hours. Slowly burning yourself out, or limiting your rest (which might curtail your highest quality hours tomorrow), so that you can eke out a one more low quality hour of work, is a bad trade. You would be better off not worrying that much about it, and you definitely shouldn’t be taking on big costs for “optimizing your productivity” if “optimizing your productivity” is mainly about getting in increasingly low marginal value work hours.

Some variables have increasing marginal returns. We need to identify those, so that we can aggressively optimize as hard as we can on those, including making hard sacrifices to climb a little higher on the ladder.

Some variables have decreasing marginal returns. We want to identify those so that we can sacrifice on them, and otherwise not spend much attention or effort on them.

Getting which is which right, is basically crucial. Messing up in either direction can leak most of the potential value. Given that, it seems like more people attempting to be ambitiously altruistic should be spending more cognitive effort, trying to get this question right.