Summary: Within a set of reasonable cause areas, in order to maximise impact, should you follow excitement and personal fit (View 1), or prioritize by expected value even when that means less interesting work (View 2)? I sketch a few arguments for each, then navel-gaze and feel uncertain. I wouldn’t recommend reading unless you’re interested in “personal, meandering first drafts”.

Here’s two views on impact, mildly exaggerated.

View One. When choosing what to work on, you should first narrow your set of considered causes to those which look good under scope-sensitive impartial altruistic perspectives. Once you’ve done this, your career choice should be majorly driven by your own personal fit and any outstanding opportunities open to you, rather than trying to continue to prioritise between these areas by their relative importance. People are most impactful when they do this individually-driven thing and think about their relative excitement, interest, and any other rare qualities they possess, rather than trying to further prioritise by all-things-considered importance.

View Two. Your career choice should be majorly driven by your assessment of the marginal returns to working on some cause area. Personal fit, etc, will obviously play some part in this process — you’ll probably be better at things you like than things you dislike — but not the predominant one. You should be wary of surprising and suspicious convergence between what you like most and what’s most impactful: some impact is about sucking it up in order to do the higher EV thing.

The second view is more sympathetic to ‘much labour is fungible’-type thinking; it’s more sceptical of self-delusion, closer to viewing doing good as the kind of thing that often involves sacrifice. The first view is more interested in people as particular individuals, more willing to work on things because they seem interesting and enjoyable.

The second view, when pressed, thinks the first view is self-indulgent, sometimes delusional, susceptible to cope; it points out how morally serious and hard-nosed and actually trying it is. The first view, when pressed, thinks the second view misses out on impact by thinking of people’s labour too much as fungible units; it thinks it tends towards self-martyrdom; and it thinks it’s too confident in its inside view. The first view views itself as appropriately epistemically modest, pragmatic, avoiding the failure mode of sacrificing overall good for legible smaller wins.

Each view has arguments to recommend it, most people follow some combination of both. Disclaimers over, I want to talk a little more about these two views.

Hashing it out

What’s the case for view 1? Three main reasons, I think:

  1. The persistence and intractability of cause prioritisation disagreements
  2. Caring about something is necessary for being excellent at it
  3. Predicting the future is hard

And, against these reasons (and for view 2):

  1. Disagreement alone is insufficient to suspend judgement
  2. Everything is like everything else, and on being the top 5%
  3. Motivated reasoning and surprising and suspicious convergence

Cause prioritisation disagreements

Lots of smart, ambitious, scope-sensitive impartial altruists have thought for a long time about how to prioritise different work. Even when these people share the same starting point, these people significantly disagree on what the most impactful work is to do. Two implications:

  1. First, we should be more pessimistic about ourselves landing on the correct cause prioritisation answer and being confident in it (the epistemic update)
  2. Second, we should update towards thinking that these causes don’t differ wildly in cost-effectiveness

Clearly, this argument alone isn’t sufficient. But I suspect people can do the right kind of “saying more” here, at least to make this argument give reason to adopt view 1.

Caring and excellence

Most things are heavy-tailed; mere 1-10x differences in abstract importance can easily be outweighed by where someone is on the distribution of impact for their role; and caring about your role is a significant determinant of impact (perhaps necessary for being one of the best people in the world at it). So, caring about what you work on really matters. Hence view 1.

The best two ways to reply, I think:

  1. Lots of work is kind of the same, and people care about properties not specifics. People care more about properties of particular problems — e.g. “involves working closely in a team”, or “has technical problem-solving” or “fast feedback loops” or “likeminded, value-aligned colleagues” — than the hyper-specific details of each role working on a problem of that type.
  2. Being the best in the world at one thing is hard, but being the best thing in the world at the intersection of 2 or 3 things is much easier. Sure, maybe loving your work gets you into the top 0.1%, but you can probably get into the top ~5% by just trying hard, and that’s all you need to be world class at the intersection of two such things.

Predicting the future is hard

Even if you think there’s consensus among some people on cause areas, the future is messy, people are bad at predicting it. But in contrast, you’re probably right about what you like, dislike, and are especially suited for. So given this set of plausibly impactful opportunities, use your personal skills and interests to prioritise more than you use predictions.

Surprising and suspicious convergence

I think the main argument against view 1 is just: it would be so surprising if the most impactful thing to do was also the most interesting and exciting!

You’re a person who cares about all kinds of different things. We care about our friends’ opinions and our job titles and prestige and location and, more abstractly, some things are just richer and more interesting to us than other things. We can learn to pay attention, but we have these preferences. And it would be so weird if they lined up with impact — what mechanism would ensure that?

So we have this “surprisingness” argument: it would be surprising if these coincided. And we have an explanation (motivated reasoning) for why we might mistakenly think they coincide. And we have the stakes — there are heavy tails everywhere. These choices really matter.

Does the god of impact eat its own young?

(stub)

Misc other notes

I don’t really know what I think here. Some notes:

Appendix

Mar 7, 2026 — add something on the deference/deference cascade problem to view 1. (and how would you know what’s good to do, anyway?)