EA sexual harassment: Where's the evidence we're worse than baseline?

One overall thought: I wonder if what you're picking up on isn't an implicit belief that EA is uniquely bad, but the reality that people are very invested in EA. The reactions are strong and strongly worded partly because people in this community want this community to be good. They care about it, many are morally scrupulous, they're invested in its health and ability to function, they know the people in it. I don't think they're implicitly thinking "this is so much worse than anything comparable." My guess is that if you asked directly whether EA is worse than a comparable community, most would say "I'm really not sure" or "probably not? I don't know." That's been my personal experience. 

I also think the comparative framing is somewhat beside the point here. EA isn't trying to reduce sexual harassment around the world, but it is strongly trying to community-build, and the relevant question for community-building is whether this is a place people can thrive in. Given that trying to answer the question of whether EA is "better or worse" is somewhat difficult, it's worth asking how important that question really is.

One thing I love about my current organisation is they are incredibly conscientious in how they build community and support their members. As someone who has now done event planning for this community for more than 3 years, I can't even imagine trying to ask if the baseline of harassment at my events is like other comparable events. I would much rather put my time and effort into making my events safer and more welcoming. In the same way that, I don't ask if the baseline satisfaction with the content at my events is the same as the average event, because I would rather put my effort into improving the content, and so on. (Edit: I was glib with my events comparison & wording, I like data & base rates are worth knowing and can help provide context and grounding, but I hope my directional point is clear)

I'll also say this, though I suspect it won't be well-received: there's something strange about a community that advocates for reducing suffering and encourages moral ambition retreating into "well, what's the baseline?" when sexual harassment comes up. What's the baseline of people who eat meat? What's the baseline of how much the average middle-income person in a developed country donates? I understand it's not exactly the same thing, but it's disheartening to see the baseline framing applied seemingly only here. 

(I made an anon account to post this; I'm a longtime EA and Forum user).

The discussion in the comments so far focuses on two claims:

1. We can't easily tell from this data whether EA is better or worse than baseline
   1. I agree, though I do think it's some evidence we're not much worse.
2. Why does it even matter whether EA is better or worse than baseline? Community members are welcome to hold EA to a much higher-than-baseline standard, if they want to.

I want to discuss this second point from my perspective, as a woman who's been in EA for some years.

Here's an analogy:

Often, I hear the criticism of EA: 'man, they have bad epistemics and huge problems with groupthink.' 

I sort of agree. I can list many instances where I think EAs are driven by groupthink and social incentives in our beliefs, rather than rigorous epistemic standards.

But, I sort of disagree.

That's because I can't think of any community [1] that does better on epistemics and groupthink than EA. Everywhere else I am vaguely involved, the groupthink is way worse, and they have very narrow standards for "acceptable" beliefs. In fact, in most communities, you can't even discuss epistemic standards at all without getting yelled at or dunked on!

I feel like it would be very reasonable for someone to approach EA like: 

Ooh, I should be looking out for places where there's potential groupthink, and views driven by status rather than evidence.

But I think it'd be unreasonable for someone to approach EA like: 

Damn, EA is a community where there's rampant groupthink — I should be suspicious of views I hear from EAs, moreso than views I find elsewhere.

Partly, that's cause the views you find elsewhere will also have those problems — and are actually likely to be even worse on the very metric of groupthink!

... So to bring it back to sexual harassment:

The reason it matters to me whether EAs believe that EA is above or below baseline on the rate of sexual harassment is that if EA is above baseline, then it'd be very reasonable for people to approach this community like:

Damn, EA is a community where there's rampant problems with sexual harassment — I should be suspicious of whether there are sexual harassment issues at play here & how they're being dealt with, moreso than in everyday interactions.

But if EA is in fact below baseline, then that's an importantly unreasonable response. That is, everyday interactions would have a higher likelihood of sexual harassment and poor responses to it.

I think it's good for people to be cautious about the risks of sexual harassment. But [2] I don't think it's good for people to feel like EA is a place where there are very high risks. 

I do think there are unacceptably high risks — in the sense that I do not think any harm is "acceptable", and we should strive to create a world free from suffering — but it doesn't follow that an appropriate response is suspicion of EAs in particular. 

[This comment is already really long, so I'm loath to add length, but the main response I anticipate is 'people aren't saying they're suspicious of EAs moreso than others, they're saying they're suspicious of EAs simpliciter.' I agree, but I think the net effect of these comments and discourse and the meme that EA has a sexual harassment problem is to make people suspicious of EAs moreso than others on this metric, and to make women in this community very afraid of sexual harassment, and to make men in this community very afraid of being a sexual harasser. I just think that's unreasonable if they don't also have that attitude everywhere else in their life — which I think is what's happening, based on anecdote / personal experience hanging out with EAs]

[1] Any community of a vaguely comparable size — obv a community of a dozen, or even a hundred or so could do much better. Also maybe rats are better but TBH I'm not totally sure.

[2] And now I show my hand... I do in fact guess that EA is better than baseline, based mostly on my personal experience dealing with sexual harassment inside and outside of EA :)

I think the fact that these issues are more openly discussed in EA than in comparable sectors can skew the discourse. If you don’t regularly attend EA events, it’s easy to think that all of the community health resources suggest an extreme problem with sexual harassment, as opposed to a proactive approach to preventing it. That said, these resources being available can make it more surprising when specific cases are not handled well. 

I work in DC. The bar is so low here that it took months of public pressure for one person to see consequences for harassment that led to his staffer committing suicide by lighting herself on fire. In day to day work, it’s normal to get groped by tipsy old men at work events, and dealing with it gracefully is basically part of the job. 

This is anecdotal, but something similar happened once at an EA-ish AI policy event (though by a sober, less elderly person). I mentioned it to the organizers in case other people were having issues, and the response was much more helpful and proactive than anything I would have expected in industry.

I would rather have people expect more from this community and be shocked when harassment happens than have people become jaded because other parts of society are willing to put up with worse working conditions. 

I am not currently convinced by the "EA is unusually bad" hypothesis which many commenters seem to believe.

I disagree with you. In my experience, EA has historically been unusually bad about this. And even though I think your quantitative analysis is poor, for reasons I'll get to, I agree there may be an anecdote/data discrepancy. I think this stems from ambiguity about what counts as sexual harassment within the EA community, which suppresses reporting of potential instances.

Within the EA community, there's historically been ambiguity regarding what counts as appropriate/inappropriate behavior. This ambiguity stems from the fact that EA is both a social and professional community, and it's not obvious what should count as sexual harassment in the context of a community that is both social and professional.

My view on this is that some behaviors that should be considered sexual harassment at an EAG should not be considered sexual harassment at an EAG afterparty. But these lines can be hard to draw. As a result, I think much of the sexual harassment issue within the EA community essentially boils down to: people applying the social rules of engagement to professional contexts, where those rules aren't appropriate. For instance, I've been hit on (e.g., flirtatious DMs on Swapcard, aggressive eye contact across the room, 1-on-1s where the other person seemed to only want to talk about personal matters, conversations involving inappropriate questions ("what's your type?"), walk-meetings that turned into one-sided flirtations) at > 50% of the EA conferences I've attended (n ~ 10). To be clear, I think these kinds of behaviors can be appropriate at EAG afterparties (assuming you aren't being rebuffed), but at a conference, they aren't appropriate. And as far as I can recall, these things have literally never happened to me at a non-EA professional conference (n ~20). Thus, in my experience, the EA community has been much worse with respect sexual harassment compared to other professional communities, though I do think that things have improved in recent years. (And kudos to the Community Health Team for their work to clarify rules around this; e.g., regarding Swapcard DMs.)

Importantly, I have reported zero of these instances to the community health team; there are several reasons for this, including some standard ones: (1) each instance felt minor, (2) it felt like reporting would be more trouble than it was worth (time-consuming, burdensome), and (3) I want to maintain relationships with some of these people. But there's also a big reason that I think is somewhat specific to the EA community, and likely the biggest driver of my failing to report, which is (4) it was not clear to me (at the time, and in some cases, even now) that certain behaviors were inappropriate, because given the social/professional blurring of lines, I wasn't sure what the rules of engagement were. In most cases, the offending party was also more embedded in the EA community than I was, so I assumed that what they were doing was allowed. In other words, I assumed the EA community had a relatively restricted definition of sexual harassment for its professional events that didn't include things that would be considered sexual harassment in other professional contexts. I think that I was probably right about this, but that we've decided as a community in recent years that our definition of sexual harassment should be closer to the standard definition employed in other professional contexts. (Importantly, if these kinds of behaviors had happened at a non-EA conference, I would've immediately known they weren't allowed, and thus would've been more likely to report.)

Separately, I also don't think we should draw ~any conclusions from your quantitative analysis. Because the definitions and measures of harassment differ across the data sources you cite, charting them on the same axis ("annual rate") constitutes bad analysis. You're not comparing the same things. As an analogy, this would be like plotting "poverty rate" on the X-axis but not accounting for differing poverty thresholds across countries, let alone how information on poverty is collected (surveys vs tax documents vs news reports).

One thing I'll add which I haven't seen in the comments is some of the vibe of 'EA is unusally bad' is that EA is probably unusually bad in some ways due to the makeup and culture of our community, and less bad in others. Bluntly I think much of this comes down to things like being worse at reading social cues, a high male to female ratio, and a community that is an unusual mix of professional and social network. 

I get the impression from female friends that being in EA means tolerating, for example, a lot more being hit on in unprofessional situations (e.g. EAG) down to creepier contexts, with the men missing being rebuffed multiple times.

This doesn't discount your main argument Nathan, but to the extent that people complain about EA being worse, this might be what they're pointing at some of the time.

I also agree with others' comments that a lot of this conversation is about how EA is 'bad' rather than 'worse', and that as community builders we want to do better, even if we're already above average.

I agree that often "EA is unusually bad" isn't grounded in data (often instead in first person accounts, although I'd note that if we assume reporting infrastructure won't capture things well, that's actually where most of the signal lives). I'm pretty skeptical that we should conclude from this data that EA isn't unusually bad. 

Others have made similar points but just at a high level, my concerns when comparing to the other data (e.g. Title IX) would be that we are biasing both the numerator and denominator of the reporting rate in a way that biases us to think we are doing better than we are. 

• Numerator: There are lots of different reporting mechanisms and places and so we shouldn't expect CEA CH to capture all the cases as it is one of many parallel endpoints. Something like Title IX sits at the end of a mandatory reporting funnel. If I tell my RA that my professor harassed me, they are mandatory reporters of that. The same isn't true for if I tell someone at org X that someone at org Y harassed me (it is true within a single workplace but because EA is quite fractured, this isn't a mandatory reporting funnel to CEA CH).
• Denominator: For a great deal of "EAs" what this means is just engaging online and donating to GiveWell. The Title IX denominator is people who are co-located on the same campus. The exposure-per-person is likely wildly different to other contexts. 

I basically think errors here aren't random so your number here is a reasonable lower bound but I'm not sure it provides insight into the overall rate. 

I'm not really convinced by your evidence that there's a widespread belief that EA is worse than other communities. Your first quote clearly says this, I grant. The other quotes seem to me to all be saying something more like "this community has this problem", and "I wish this community was better / had hoped that it would be better", without saying "we are worse than relevant comparisons", much less "we are worse than relevant comparisons because of intrinsic aspects of our culture".

FWIW, my belief, and I'm sure I'm not the only one, is more like: EA handles this badly, but also handling this badly is widespread. I don't think I've ever been in a community that I was confident had a good answer for this. EA should address this within its own community, because that's where we understand what's happening best and have the best tools for addressing it. This is true regardless of whether EA is worse, the same, or better than its peers: it's true for as long as it seems feasible to do better than we're doing.

In the end, interventions should be decided on their merits, and I think we should tip the balance away from the abstract and towards the concrete questions of what specific things we should or shouldn't do.

Thanks for writing, Nathan! I think there are two separate arguments: one says that EA is doing worse than baseline, and the other is that EA is doing worse than our shared -- sometimes implicit -- community standards. It's fair to ask people to chart their expectations against baseline for almost anything, but it's also fair to establish community norms above baseline. Some of the comments you quoted (inc. one of mine) are critiques relying on the higher standard. 

 

One of my comments is cited in the section about there being a wide belief in the EA community, so I can provide my personal perspective, as well as comment more broadly.

While other people have already pointed out that your comparison chart has methodological issues, which I agree with, I don't think this is a crux. While I think communities and spaces can never reach a state where there are no bad actors, or where no one does or experiences harm — I very much don't think this means harm is okay (which I'm sure you agree with), or that our response to harassment might somehow matter less if it turned out our community did better than 'baseline'. I was personally very disappointed by how CEA handled the recent harassment case, and my issue isn't that I think CEA should somehow have created a work environment where nothing bad ever occurred, or that CEA has more HR violations than baseline, but instead that it should have — bare minimum — taken reasonable steps to prevent the harassment, as required by UK law. In this case the number of incidents isn't the issue, it was how the specific incident was handled.

Regardless, I think sexism isn't something that ends at, or is only a problem in the context of, the legal definition of sexual harassment. Workplaces and college campuses have a legal obligation to prevent harassment, but the EA community exists in and goes beyond those spaces (such as an EAG afterparty, or twitter/facebook discussion threads). Sexism isn't just a failing to meet a legal standard, just like doing good isn't defined by donating to 501(c)(3) organizations. EA is a community that is dedicated to doing the most good, and has strong norms around truth seeking and respect — sexism and harassment are inherently not those things, which is why I think EA is ideologically opposed to them. As with any community, some parts will be better and some parts will be worse — but I think we have an obligation to aim for The Most Good You Can Do, and for me that obligation doesn't end once we close the cost effectiveness analysis spreadsheet. I think we as a community aspire to being the best we can.

While we could have a conversation about how harassment/sexism reduces the ability for EA to be an impactful movement, I also think we should act with virtue in our conduct, and that we have certain deontic obligations. Especially in recent years I think many within EA hold these views, even if it's not from a pure 'impact maximization' justification (I think many EAs would be quite horrified if it was somehow the case the EA movement was producing a lot of positive impact, but also was intentionally allowing harassment to occur). This is why I think the specific rate of harassment within EA is not a crux — clearly there are ongoing issues, even if the true rate of incidents was lower than whatever you think the reference class should be.

I imagine many of the people you cited feel similarly. The rate of harassment is only part of the conversation — I don't think the question of 'is the EA community bad at handling harassment' can be answered with just a comparison between reporting rates by communities. Given we are a community dedicated to doing good better, it feels almost absurd if the community were content with being at 'baseline' for harassment and our response to it.

Is it possible that subgroups within EA have sexual harrassment rates that are significantly above or below baseline? For example, maybe AI related spaces or SF Bay Area spaces have higher levels of sexual harrassment. Then people who are generally in those spaces would subjectively observe that the level of sexual harrassment is high, even though it might not be if you were to average across all of EA.

If there are parts of EA that have lower than average levels of sexual harrassment, I would not expect anyone to speak up and say, "man, this animal welfare community sure doesn't have much sexual harrassment". So there is a bias in terms of who is moved to share their experience.

The acceptable baseline amount of sexual harassment in a community that is otherwise steadfastly committed to rejecting the status quo and improving the state of the world, is zero. We are not comparing ourselves to the average workplace or university.

To clear up a few general points. 

On missing moods. I agree this is horrible. As I've said, I don't think I'm the best person to say this. But if an issue is important, I think it's important to discuss accurately. I wish I were a more gracious, more empathetic, more concise, communicator. 

Methodology:

• John Salter originally claimed the numbers were perhaps 300x off. After discussion (if I understand correctly) he thinks it's more like 2-20x
• Liv Gorton thinks the denominator may be too low and the numerator too large, though I don't currently know by how much
• AbsurdlyMax has methodological concerns
• If people wish to come up with their own numbers, I'll happily look at them. Though since my current EA numbers are 20 - 50x lower than top US schools the argument still holds. 

It might seem like I am banging on about base rates, but they are in my view, a generally useful tool. It seems important to understand how bad this problem is compared to other places. Then we can decide how easy we expect it to be to fix. 

I remain confused on the overall discussion. It feels (to me) like people are disagreeing with me about something. Some people say that commenters don't think EA is unusually bad, some argue that my methodology is wrong, some think that comparison isn't the right tool here. I don't think it can be all of these things at once. This feels to me like there should be more internal conflict than there is. Sexual harassment cannot be as bad as other communities, much worse, and not comparable, all at the same time, that's not how things work. 

If I understand correctly, your question is whether EA is specifically worse than other organisations, rather than whether being on a par with other organisations is also very bad in this regard?
Because I get the impression that some people simply say that EA is bad, but don’t necessarily rule out the possibility that others are bad too.

I reckon it might be important to know whether the EA is actually performing worse, so the question does seem important to me, because one of the possible answers would be significant. 
(For example, learning more from organisations that are doing better, etc.)
Even if everyone is doing very poorly, performing worse than everyone else is still very important information, even if performing just as well is by no means satisfactory.
I imagine this might be part of your approach?

For my part, I don’t have enough reliable data to form a proper opinion on this specific issue, and indeed the question seems difficult to answer based on the data: a very high reporting rate compared to others within an organisation could mean that it handles reports much better (in terms of spaces for discussion, reception, listening, procedures, etc.), and so ultimately is a better-run organisation; but the opposite could also be true: it might mean that it performs worse, whilst managing reports in much the same way as others.

Furthermore, better care, better facilities, a better reporting policy, better responses to these reports, etc., could (in theory) well coincide with an underlying greater tendency to commit assaults. I don’t think these necessarily contradict each other? Even if they might, there may be an argument to be made for that.
And in that case, identifying the reporting mechanisms would not be enough to reduce the actual number of cases.
However, in that case, I imagine the number of reports would be particularly high, so it would still serve as a warning.

I can't help feeling that these measurement issues are skewing the data in all sorts of ways,
but I still believe it’s important to make progress on this issue; it’s not a case of saying “we can’t answer” definitively and thus stopping the discussion.

However, why are you making this post?
I don’t get the impression you’re creating this post to say “let’s ask ourselves whether EA is worse, because it would be important if it is, and here’s the data I’ve been able to gather so far”, but rather with an idea that strikes me as “hey, it’s not that bad, is it?”^[1], which is completely different and, moreover, isn’t very epistemologically humble given the complexity of the subject regarding the data.

I also think that’s why there are so many critical comments, because if you’d simply asked a specific comparative question whilst being perfectly clear about your aims and what you meant, I don’t think there would be any such comments, or at least far fewer of them. (I may be wrong; that’s just my current view.)

Because I’m not sure if you’ve realised that saying ‘others are worse’ or ‘others aren’t any better’ is often used to downplay one’s own behaviour. It’s a very common argument, and your article doesn’t seem at all cautious on this point, for several reasons mentioned here, which personally leads me to believe that this is the underlying motivation. (I may be wrong,I hope so, at least, but I’m mentioning this out of honesty and because I think this information is useful when writing posts.)

Which in no way detracts from the relevance of the issue (in the event that it is indeed worse).

1. ^{^}

   the elements in question are: the article’s title and the way it is written, mixing comments that say it’s bad with comments that say it’s worse, even though the two have nothing to do with each other.

One tension I grapple with in these conversations that I'm struggling to articulate is the difficulty between separating out the fact that bad things tend to be so much worse when you experience them, so that if the data isn't that damning, as I man, I can feel good about EA and continue with my day. Whereas I imagine for women who bear the brunt of it, EA being not that bad is still pretty shitty, and if I knew this viscerally, I would be a whole lot more empathetic and do a lot more to try and combat it. 

Maybe the point I'm trying to make is when engaging with topics like this there can be a muting of how bad 'not that bad' is. 

I don't think this invalidates or contradicts the main point of your post Nathan, more a potential missing mood to all those discussing in the comments (not calling out anyone in particular). 

It is extremely difficult to determine base rates for something like sexual harassment, because it's an offence that allows for ambiguity and plausible deniability, because there's room for retaliation, etc, and it will strongly depend on how much people trust the bodies they are reporting to. 

What we can do is look at the responses to the incidents that do get raised, and the experiences of victims, and judge whether or not they live up to the standards we want to see in a group that takes sexual harrasment seriously. I do not think the grades are very good on this front. 

Only 3 months ago we had a writeup detailing a shockingly terrible response to sexual harrassment by one of the most prominent EA orgs out there. The response is far worse than anything I've ever seen at any organisation I've ever been in. This indicates to me that the environment is nowhere the high standards that should be aimed for. 

Regardless of the actual base rates, the question that matters the most is whether there is room for improvement, and I think it's blindingly obvious that the answer is yes.