How should EA navigate the divide between those who prioritise epistemics vs. social capital?

As you are someone who falls into the "prioritize epistemics" camp, I would have preferred for you to steelman the "camp" you don't fall in, and frame the "other side" in terms of what they prioritize (like you do in the title), rather than characterizing them as compromising epistemics. 

This is not intended as a personal attack-I would make a similar comment to someone who asked a question from "the other side" (e.g.: "On one side, you have the people who prioritize making sure EA is not racist. On the other, you have the people who worried that if we don't compromise at all, we'll simply end up following what's acceptable instead of what's true".)

In general, I think this kind of framing risks encouraging tribal commentary that assumes the worst in each other, and is unconstructive to shared goals. Here is how I would have asked a similar question:

"It seems like there is a divide on the forum around whether Nick Bostrom/CEA's statements were appropriate. (Insert some quotes of comments that reflect this divide). What do people think are the cruxes that are driving differences in opinion? How do we navigate these differences and work out when we should prioritize one value (or sets of values) over others?"

I really dislike the "2  camps" framing of this. 

I believe that this forum should not be a forum for certain debates, such as that over holocaust denial. I do not see this as a "trade-off" of epistemics, but rather a simple principle of "there's a time and a place". 

I am glad that there are other places where holocaust denial claims are discussed in order to comprehensively debunk them. I don't think the EA forum should be one of those places, because it makes the place unpleasant for certain people and is unrelated to EA causes. 

In the rare cases where upsetting facts are relevant to an EA cause, all I ask is that they be treated with a layer of compassion and sensivity, and an awareness of context and potential misuse by bad actors. 

If you had a friend that was struggling with depression, health, and obesity, and had difficulty socialising, it would not be epistemically courageous for you to call them a "fat loser", even if the statement is technically true by certain definitions of the words. Instead you could take them aside, talk about your concerns in a sensitive manner, and offer help and support.  Your friend might still get sad about you bringing the issue up, but you won't be an asshole. 

I think EA has a good norm of politeness, but I am increasingly concerned that it still needs to work on norms of empathy, sensitivity, and kindness, and I think that's the real issue here. 

I think we should reject a binary framing of prioritising epistemic integrity vs prioritising social capital. 

My take is:

Veering too far from Overton Windows too quickly makes it harder to have an impact, and staying right in the middle probably means you're having no impact - there is a sweet spot to hit where you are close enough to the middle that your reputation is intact and are taken seriously, but you are far enough away from the middle that you are still having impact.

In addition to this, I think EA's focus on philanthropy over politics is misguided, and much of EA's longterm impact will come from influencing politics, for which good PR is very important.

I think this framing is wrong, because high standards of truth-seeking are not separable from social aspects of communication, or from ensuring the diversity of the community. The idea that they are separable is an illusion.

Somehow the rationalist ideas that humans have biases and should strive to contain them, have brought an opposite result - that individuals in the community believe that if they follow some style of thinking, and if they prioritise truth-seeking as a value, then that makes them bias-free. In reality, people have strong biases even when they know they have them. The way to make the collective more truth-seeking is to make it more diverse and add checks and balances that stop errors from propagating.

I guess you could say the divide is between people who think 'epistemics' is a thing that can be evaluated in itself, and those who think it's strongly tied to other things.

I find this framing misleading for two reasons:

1. People who support restrictive norms on conversations about race do so for partially epistemic reasons. Just like stigmatising certain arguments or viewpoints makes it more difficult to hear those viewpoints, a lot of the race-talk also makes the movement less diverse and some people feel less safe in voicing their opinions. If you prioritise having a epistemically diverse community, you can't just let anyone speak their mind. You need to actively enforce norms of inclusiveness so that highest diversity of opinions can be safely voiced.
2. Norms on conversations about race shouldn't be supported by "popularity" reasons. They should be supported out of a respect for the needs of the relevant moral patients. This is not about "don't be reckless when talking about race because this will hurt object-level work". It's about being respectful for the needs of the people of color who get affected by these statements. 

Put me down on the side of "EA should not get sidetracked into pointless culture war arguments that serve no purpose but to make everyone angry at each other."

Some people have argued for the importance of maintaining a good image in order to maximise political impact. I'm all for that. But getting into big internal bloodletting sessions over whether or not so and so is a secret racist does not create a good image. It makes you look like racists to half the population and like thought police to the other half.

Just be chill, do good stuff for the world, and don't pick fights that don't need to be picked.

This is a bit orthogonal to your question but, imo, part of the same conversation. 

A take I have on social capital/PR concerns in EA is that sometimes when people say they are worried about 'social capital' or 'PR', it means  'I feel uneasy about this thing, and it's something to do with being observed by non-EAs/the world at large, and I can't quite articulate it in terms of utilitarian consequences'.

And how much weight we should give to those feelings sort of depends on the situation:

(1) In some situations we should ignore it completely, arguably. (E.g., we should ignore the fact that lots of people think animal advocacy is weird/bad, probably, and just keep on doing animal advocacy). 

(2) In other situations, we should pay attention to them but only inasmuch as they tell us that we need to be cautious in how we interact with non-EAs. We might want to be circumspect with how we talk about certain things, or what we do, but deep down we recognize that we are right and outsiders are wrong to judge us. (Maybe doing stuff related to AI risk falls in this category)

(3) In yet other situations, however, I suggest that we should take these feelings of unease as a sign that something is wrong with what we are doing or what we are arguing. We are uneasy about what people will think because we recognize that people of other ideological commitments also have wisdom, and are also right about things, and we worry that this might be one of those times (but we have not yet articulated that in straightforward "this seems to be negative EV on the object level, separate from PR concerns" terms).

I think lots of people think we are currently in (2): that epistemics would dictate that we can discuss whatever we like, but some things look so bad that they'll damage the movement. I, however, am firmly in (3) - I'm uncomfortable about the letter and the discussions because I think that the average progressive person's instinct of 'this whole thing stinks and is bad'...probably has a point? 

To further illustrate this, a metaphor: imagine a person who, when they do certain things, often worries about what their parents would think. How much should they care? Well, it depends on what they're doing and why their parents disapprove of it. If the worry is 'I'm in a same-sex relationship, and I overall think this is fine, but my homophobic parents would disapprove' - probably you should ignore the concern, beyond being compassionate to the worried part of you. But if the worry is 'I'm working for a really harmful company, and I've kinda justified this to myself, but I feel ashamed when I think of what my parents would think' - that might be something you should interrogate. 





 

I think it is trivially true that we sometimes face a tradeoff between utilitarian concerns arising from social capital costs and epistemic integrity (see this comment).

But I don't think the Bostrom situation boils down to this tradeoff.  People like me believe Bostrom's statement and its defenders don't stand on solid epistemic ground.  But the argument for bad epistemics has a lot of moving parts, including (1) recognizing that the statement and its defenses should be interpreted to include more than their most limited possible meanings, and that its omissions are significant, (2) recognizing the broader implausibility of a genetic basis for the racial IQ  gap, and (3) recognizing the epistemic virtue in some situations of not speculating about empirical facts without strong evidence.

All of this is really just too much trouble to walk through for most of us.  Maybe that's a failing on our part!  But I think it's understandable.  To  convincingly argue points (1) through (3) above I would need to walk through all the subpoints made on each link.  That's one heck of a comment.

So instead I find myself leaving the epistemic issues to the side, and trying to convince people that voicing support for Bostrom's statement is bad on consequentialist social capital grounds alone.  This is understandably less convincing, but I think the case for it is still strong in this particular situation (I argue it here and here).

One meta-question: Is this "divide" merely based on different assumptions and beliefs about what strategy will maximize impact, or is it based at least in part on something else for one or both sides?

How should we navigate this divide?

I generally think we should almost always prioritize honesty where honesty and tact genuinely trade off against each other. That said, I suspect the cases where the trade-off is genuine (as opposed to people using tact as a bad justification for a lack of honesty, or honesty as a bad justification for a lack of tact) are not that common.

Do you disagree with this framing? For example, do you think that the core divide is something else?

I think that a divide exists, but I disagree that it pertains to recent events. Is it possible that you're doing a typical mind fallacy thing where just because you don't find something to be very objectionable, you're assuming others probably don't find it very objectionable and are only objecting for social signaling reasons? Are you underestimating the degree to which people genuinely agree with what you'd framing as the socially acceptable consensus views. rather than only agreeing with said socially acceptable consensus views due to social capital? 

To be clear, I think there is always a degree to which some people are just doing things for social reasons, and that applies no less to recent events than it does to everywhere else. But I don't think recent events are particularly more illustrative of these camps. 

it appears to me, that those who prioritise AI Safety tend to fall into the first camp more often and those who prioritise global poverty tend to fall into the second camp.

I think this is false. If you look at every instance of an organization seemingly failing at full transparency for optics reasons, you won't find much in the way of trend towards global health organizations. 

On the other hand, if you look at more positive instances (people who advocate concern for branding, marketing, and PR with transparent and good intentions),  you still don't see any particular trend towards global health. (Some examples: [1]][2][3] just random stuff pulled out by doing a keyword search for words like "media", "marketing" etc). Alternatively you could consider the cause area leanings of most"ea meta/outreach" type orgs in general, w.r.t. which cause area puts their energy where.

It's possible that people who prioritize global poverty are more strongly opposed systematic injustices such as racism, in the same sense that people who prioritize animal welfare are more likely to be vegan. It does seem natural, doesn't it, that the type of person who is sufficiently motivated by that to make a career out of it, might also be more strongly motivated to be against racism? But that, again, is not a case of "prioritizing social capital over epistemics", any more than an animal activist's veganism is mere virtue-signaling.  It's a case of genuine difference in worldviews. 

Basically, I think you've only arrived at this conclusion that global health people are more concerned with social capital because you implicitly have the framing that being against the racist-sounding stuff specifically is a bid for social capital, while ignoring the big picture outside of that one specific area. 

Also I think that if you think people are wrong about that stuff, and you'd like them to change their mind, you have to convince them of your viewpoint, rather than deciding that they only hold their viewpoint because they're seeking social capital rather than holding it for honest reasons.

I actually think the framing is fine, and have had this exact discussion at EAG and in other related venues. It seems to me that many people at the “top” of the EA hierarchy so to speak tend to prefer more legible areas like AI, pandemics, and anything that isn’t sales/marketing/politics.

However as the moment has grown, this means leadership has a large blind spot as to what many folks on the ground of the movement want. Opening EA global, driving out the forced academic tone on the forum, and generally increasing diversity in the movement are crucial if EA ever wants to have large societal influence.

I take it almost as a given that we should be trying to convince the public of our ideas, rather than arrogantly make decisions for society from on high. Many seem to disagree.

Where do you draw the line at epistemically indefensible? Is there anything that is not epistemically indefensible? 

Also just so I understand, is doubling down on pseudoscience like, for example, race and intelligence, is being epistemically....bold? Integral? Are you willing to make space in EA for flat earth theory? For lizard people in the center of the earth? Anti-semitism? Phrenology? 


 

Do you disagree with this framing? For example, do you think that the core divide is something else?

I think this framing is accurate, and touches on a divide that has been repeatedly arising in EA discussions. I have heard this as "rationalists vs. normies," "high decouplers vs. low decouplers," and in the context of "feminization" of the movement (in reference to traditionally masculine dispassionate reason falling out of favor in exchange for emphasis on social harmony).

Additionally, I believe there are significant costs to total embrace of both sides of the "divide."

1. There are cause areas with significant potential to improve effectiveness that are underexplored due to social stigma. A better understanding of heritability and genetic influences on all aspect of human behavior could change the trajectory of effective interventions in education, crime reduction, gene-editing, and more. A rationalist EA movement would likely do more good per dollar.
2. Embrace of rationalism would be toxic to the brand of EA and its funding sources. The main animus I have seen for the Bostrom email is that he said black people have lower IQs than white people. This, though an empirical fact, is clearly beyond the pale to a large percentage of the population EA seeks to win over. Topics like longtermism and animal welfare are weird to most people, but HBD actively lowers public perception and resultant funding. The good per dollar may go up, but without a degree of tact, the amount of dollars would significantly decrease.

I must imagine that there is a utility function that could find the equilibrium between these two contrasting factors: where the good per dollar and amount of funding achieve the most possible total good.