Some thoughts on EA outreach to high schoolers

Buck13 Sep 2020 22:51 UTC
122 points
59 commentsEA link

[A bunch of these points came from Claire Zabel. Thanks to various people who provided feedback.]

TL;DR: I think that lots of EAs have updated against outreach to high school students based on evidence that isn’t actually entirely relevant. I also think that there are some reasons to think that outreach to high school students could be competitive with outreach to undergraduates. There are a variety of downsides to outreach targeting younger people, none of which seem decisive to me.

EAs seem to be too pessimistic about high school outreach

There have been several posts about unsuccessful (according to the authors) attempts to get high schoolers into EA. E.g. see here and here. When I’ve talked to people about different recruiting possibilities (in which these posts were often mentioned if I brought up the possibility of trying to recruit high schoolers), I’ve gotten the sense that many EAs are pessimistic about trying to engage high school students. But I think these past interventions were ineffective for reasons unrelated to their target audience, and that other interventions aimed at high school students seem comparably promising to working with university students.

These posts document attempts to engage high schoolers that were relatively short and untargeted (they didn’t strongly select their audience, other than by age, and didn’t get to select from a very big group). If you imagine the analogous kind of intervention for other age groups where EA recruitment has had notable successes, we predict the results would be (and has been) similarly disappointing. E.g. if you took a random group of a few hundred university students or recent graduates, selected within that group for EA-ness, then showed them a few videos or had them listen to a few hours of talks about EA, we predict the results would usually be similarly lackluster (and, it’s my impression that they have been, when that kind of thing has been tried).

In contrast, many of the biggest EA groups are at top universities, where they can select from thousands of students, and where the students have been somewhat pre-selected for traits that seem correlated with EA-ness, like intellectual curiosity and openness.

The only somewhat-similarly-targeted analogues I know of for high schoolers are SPARC and ESPR (which recruit from people with evidence of talent in STEM fields, e.g. by looking for high school students that have done well in STEM Olympiad competitions). I know a decent number of SPARC and ESPR alum have gone on to do direct work in top cause areas, some of which seems really promising. It doesn’t seem easy to compare the hit rate of SPARC to e.g. the Yale EA group, and establishing causality is always hard, but the story of SPARC seems totally different from the lack of traction SHIC seemed to get.

Basically, I think we should treat engaging with high school EAs more similarly to how we treat engaging with older EAs: we should look for places with particularly high density of people who have a chance of contributing to high priority causes and engage them over the course of weeks or months rather than using relatively short means of engagement, and also do structured types of engagement where they can build up connections with other people interested in this. If we do that, I don’t think there’s a good reason to be more pessimistic about interventions to engage high schoolers than university students.

Benefits to engaging with younger people

I see a few big upsides to working with younger people:

  • It seems harder to recruit people the older they are, at least after people get into their thirties, and so maybe it gets even easier if you go younger than the point at which most recruiting efforts start. I’m sure this stops being true at some point — I don’t know if one could effectively recruit a 10-year-old to EA — but the lots of anecdotal experience above + my priors indicates that plausibly, a large fraction of people who were converted to EA could have been and wish they were engaged earlier. If more people could be drawn to EA that way, recruiting efforts focusing on engaging younger people might have greater counterfactual impact, because they engage before some window of opportunity elapses.

  • I think the EA and rationality communities have lots of tools that help people become overall better at thinking, and potentially vastly increase their lifetime impact. Having access to those earlier seems like it might be very useful for practicing using them and thus become more skilled; perhaps it allows people to become more skilled than is possible if one doesn’t encounter these ideas until later in life.

    • Relatedly, hearing about these ideas earlier might cause people to make some decisions better.

  • Anecdotally, a large number of the most dedicated and promising longtermist EAs I know heard about EA in high school (at a workshop I ran for a small group of newish longtermist EAs, if I remember correctly about ⅔ raised their hands when asked if they’d heard about EA before age 18), and most others I know that I’ve asked this question to wish they had, and think their impact would be greater if they’d heard about EA earlier.

  • EA Survey data implies people who are recruited earlier change cause areas more frequently (see here, the section called ‘Cause Preference Change’), which probably causes them to land in high-priority causes at higher rates (assuming EA helps with cause selection).

  • There’s less competition for the attention and time of younger people.

  • Younger people might be easier to convert for a given level of recruiter quality, because they have lower standards for whether someone is knowledgeable or interesting.

  • Younger people often have more free time, especially in the time between high school and college, to explore ideas and think broadly about their lives, compared to older people.

Downsides and failure modes:

It’s harder to select for the most talented people. There’s less concentration of intellectual talent among high schoolers than college students. So you’d probably want to put relatively less emphasis on outreach at specific high schools (though there are some high schools that are probably as dense with talent as top colleges). Also, it’s probably harder to assess talent among high schoolers than older people.

This downside might seem extremely important if you think that student groups at top colleges are by far the best intervention for recruiting undergraduates.

Getting into rationality community stuff and EA too early might be bad for people:

  • I think that rationalist community stuff can make people feel more isolated and alienated from society, which seems good in some respects but might make people less motivated to be traditionally successful by drudging themselves through various dumb systems; eg they might do worse at college admissions or getting good grades in dumb college courses.

  • People who get into EA sometimes seem to lose interest in intellectual hobbies (the stereotypical example here is someone who stops thinking about math for fun because they need to focus on learning important things like AI safety). I’m worried that people who get into EA while really young might end up not going down a variety of interesting intellectual rabbit holes and end up as worse thinkers for it.

Anna Salamon’s post Reality-Revealing and Reality-Masking Puzzles talks about how people sometimes get disoriented by thinking about AI risk; my concerns here are pretty similar to those she describes. I think that younger people might be more vulnerable to those pitfalls than older people.

Also, rationality community stuff seems to sometimes really mess people up—they go down weird rabbit holes and then decide to become monks or something—and I feel a bit of hesitation about recommending that young people get too involved with the community given the risk of that happening with them. I don’t have this same concern about EA.

It’s longer before they can do useful EA work. This is most severe if you think that EAs working now can do much more good per hour than EAs in the future. High school students are four years younger than university students on average; so if you have a 10% discount rate per year, this leads to thinking that recruiting a high school freshman is 66% as valuable as recruiting a college freshman. This factor seems likely to be offset by some of the advantages listed above; in particular it seems plausible that recruiting high school students is 35% cheaper than recruiting college students.

Two other brief points that would deserve attention among people trying to do work in this area:

  • It’s more politically delicate to do outreach to high schoolers than older people.

  • If there were more orgs doing this, there’d be the risk of abuse working with minors if in-person.

Buck13 Sep 2020 22:51 UTC
122 points
59 commentsEA link