Beware Premature Introspection

This post is also adapted for a more general audience on Substack, under the title of Act Now, Think Later.

Let’s say that a friend of yours gives you some lottery tickets for Christmas. You swallow down your instinctive urge to start talking about expected value and instead graciously thank your friend for their gift. The draw’s in a week or so, so you open the rest of your gifts, spend the day with family and friends, and think no more of it.

The day before the draw, you run into your friend again. You start talking, and you meander onto the topic of the lottery tickets, and they ask you what you’d spend the money on, if you won. You honestly reply that you hadn’t thought about it. And if you were in this situation, that’s probably exactly how you’d act. Maybe you’d say something about donating most of it to your favorite charity, or buying a house, or retiring, but you wouldn’t give the question a thorough analysis.

But why not? After all, winning the lottery would be an extremely major event. It would change your life. You’d rethink your career. You’d rethink your donation strategies. It would be a major shift in what you should do. Surely, given that, isn’t it worth spending a few hours, or days, or even weeks, really going over the implications of that?

Well, yes, you might say - but that’s conditional on you actually winning the lottery. You have correctly figured out that, while you should do all those things mentioned above if you actually won the lottery (or even if your chance of winning was, say, 50/50) it’s a waste of thinking when, odds are, this lottery ticket will never amount to anything. This would be premature introspection. You would be devoting serious amounts of time to thinking about something you probably never needed to worry about at all. That’s how I’d react, sure in my knowledge that I’m being a rational, logical human being.

But then we look at job applications, and all that goes out the window.

Impact and Certainty

In many ways, putting in a job application is similar to the lottery example above - though hopefully your chance of success is a lot higher! If you put in an application for a job, or for a fellowship, or for a grant, you have some probability of succeeding. For an EA position especially, taking this offer might change a lot of things in your life. You might need to move. Maybe even overseas. You’d need to leave your current position, or abandon the traditional path of academia or industry. If you have a decent for-profit job already, you’ll probably need to take a cut in pay and/or career capital in the non-EA world. And is this even the best cause? Should I work on global health when building EA or AI safety might be more important?

And, at least for me, the natural thought process is that I should consider all of these things carefully before submitting any applications to these positions. After all, what if I consider all these things and then decide that I don’t actually want one of these positions after all? Shouldn’t I resolve all of my uncertainty around this and be rock-solid in my course of action first?

However, let’s look at our lottery example above. We can very clearly see in this example that it’s not just the impact of success that determines how much we should think about the result - it’s the impact of success, multiplied by the probability of succeeding. No matter how impactful winning the lottery would be, the chance of winning it is so small that it’s not worth spending more than a few seconds thinking about the outcome unless you actually win. 

Thus, for a job application, especially a very competitive one like many EA jobs are, I would propose the following heuristic: If you think about it for a few minutes and haven’t decisively made up your mind on whether you would want to take the job if it were offered to you, just go ahead and submit an application. Worry about the finer details once you’ve made it through a couple of rounds - and if you don’t make it that far, you didn’t need to worry about it at all, and have saved valuable time and mental energy.

Cost and Signals

There is one final variable that’s worth thinking about in this question, which is - what is the cost of trying in the first place? If I want to submit a job application, the cost of submitting the application is measured in minutes. On the other hand, if I want to try and learn machine learning to get involved in deep reinforcement learning research, the cost might be measured in dozens or hundreds of hours if I’m already a good software engineer, and even more if I’m not.

In that case, it would be worth thinking more before acting, but the first thing you should probably think about is “Is there a way I can cheaply get some signal, even if it’s relatively weak, that I might be suited for this path?” In the above example, if I don’t know how to program yet, I should probably check my fit for that first. For machine learning, it might be worth watching the first video of the Fast.ai course and seeing if it interests you, before committing to completing the full course or learning multivariable calculus. 

Job applications are good for this, since while each round tends to ask a bit more time of you, getting to each round increases the probability of success and makes it worth investing more time both into the application itself and into deciding whether you really want the position. They have this ideal feedback loop already built into their structure, and you should use it.

Charity Entrepreneurship also does this really well. You could put in an application, or you could take their 5-minute quiz which will tell you whether you should apply. As a case study to apply my own methods, Charity Entrepreneurship is a great one. I had no idea (and still don’t) about whether I would actually want to run a charity even if I was suited for the work. However, a 5-minute quiz was worth doing, even just out of curiosity. Then they said I’d be a good fit, so I committed a little more. I committed an hour or so of my time, during which I read over the broad picture of the offer and skimmed their top charity ideas. Then I asked myself “Could I potentially see myself doing this work?” and when the answer was “Maybe” rather than “Hell no”, I put in an application. There’s five rounds total, including the written application, so if I make it to the third or fourth round, I’ll start asking more seriously what I would do if I actually succeeded. Until then, it’s just wasted motion. 

But you can absolutely create your own feedback loop like this. Holden Karnofsky has a fantastic example of such a loop on researching highly intractable-seeming questions, from “Ask myself if I could see myself doing it” all the way to “Spend a few months on it.” If it can be done for vague, open-ended research questions, it should be doable for almost anything.

Just don’t spend too much time on creating the perfect feedback loop, or that defeats the whole purpose.

Conclusion

There are three things we want to consider when we ask how much time we should invest into thinking about something before we act. We want to think about the impact of success on your life, the chance of success, and the cost of the attempt. 

The more that success would mean, the more you should think about what success would look like for you before trying to achieve the goal.

The less likely success is, the less time you should spend thinking about it before you act.

The larger the cost of attempting something is, the more time you should spend thinking about it before committing to the attempt. Some of that time should be about finding ways to lower the cost of the attempt, if your first attempt is looking to be more than a day or two worth of work.

EA, especially EA community building, needs doers. Doing, then thinking is a valid principle. “Bias for action” is right up there in Amazon’s leadership principles, and they’ve managed to do pretty well for themselves. 

Effective altruism is a movement that draws introspective thinkers to its ranks, since we do need to think carefully, hard, and often for a long time about how to do the most good. But the important thing to remember is that doing the most good possible is a force multiplier. It’s a way to put your effort towards the best thing you can possibly be doing…but a force multiplier needs force to multiply. A thousand times zero is still zero.

So you know that thing you’ve been thinking about trying, but you’re not sure if you’re capable, or if you really want to work on it for your entire life? Figure out some way you can gain some useful information with a few hours of your time, and instead of asking whether you should devote your career to something, ask whether you should devote five minutes, or an hour, or next Saturday. If the answer is no, great! You’ve just resolved the problem. 

But if the answer is yes, take action now, and worry about succeeding when it happens.