What an EA leaders’ forum should actually decide
In the summer of 2016, the Centre for Effective Altruism (CEA), led by the Oxford philosopher Will MacAskill, invited a few dozen senior figures in the effective altruism movement to a three-day “Leaders Forum” in the Bay Area. The reply below is Alyssa’s response to that invitation. Rather than discuss logistics, she argues the format will only be worth the cost if the organizers commit it to a single, harder question — ideally, what the EA “brand” should mean — and walks through the substantive sub-questions a brand decision would force, from which kinds of people the movement is recruiting (and so de facto excluding) to whether EA should put more weight on funding, on implementation, or on media. The closing section sketches a CEA-run online newspaper as one practical move toward better internal communication.
Please be aware that GMail auto-flagged this email as suspicious and sent it to Spam. I wouldn’t have seen it if it weren’t for some dumb luck, so some other people probably won’t see it either.
In general, I’d be very interested in attending an event like this, but I think a three-day workshop of this size would have to have a higher degree of structure. An afternoon or evening event can just be a meet-and-greet, but for a three day event, it’s good to have a specific purpose and then organize the attendee list and schedule around that purpose. For example, the Puerto Rico AI conference had the purpose of building a common platform between the top people working in AI, and the various groups interested in AI safety, who had largely been isolated from each other before. That goal then drove the conference’s specifics. Eg., Rob Mather wasn’t at the Puerto Rico conference, not because AMF doesn’t do good work, but because tropical diseases weren’t a part of that focus.
The choice of focus has to be CEA’s, but one potentially powerful (though possibly overly ambitious) one is trying to reach a consensus on how to define the effective altruism brand. This is partly a PR question, but it’s also so much more than PR. How the EA brand is defined also largely determines questions like: What kind of people will be in the EA community, and what kinds of skills, demographics, inclinations, and viewpoints will they have? For example, if EA is defined in such a way as to include lots of political questions, then people with a political background will be attracted to that. They’ll be more likely to have skills in certain areas (communications, writing, organizing large groups), and less likely to have others (bioengineering, venture investing, mathematical proofs, server administration). They’ll tend to see the world in a certain way, and to solve problems in a certain way. People can change, of course, but in general the best predictor of future behavior is past behavior; someone who spent several decades on video games will tend to have a video-gamey approach to many problems (eg. Demis Hassabis). There will also be large differences in average personality type. Among people I know, at least, those with X interests tend to be very smart, but low-conscientiousness; those with Y interests tend to be very dedicated and passionate, but prone to being dogmatic and causing heated conflicts; those with Z interests tend to be good at building effective organizations, but bad at cause selection (for several different values of X, Y, and Z, come to think of it). Which skills do we want the most, and which flaws can we live with?
Which forms of morality are included in EA, and which are not? Someone who is completely selfish isn’t an altruist, by definition, but they might still wind up working closely with EAs on issues like x-risk. Someone who values current lives, but not future lives (eg. Aubrey de Grey), counts as an EA under standard definitions, but this position seems to be very rare among EAs in practice. Many people partly believe that all lives have equal value, but aren’t fully consistent, since that’s very hard for humans to do; eg. if given the choice to either shoot a randomly selected stranger, or shoot their spouse and get a free ice cream cone, I expect almost everyone would take #1 even though #2 gets more utility points under an equal-lives formulation. Does EA welcome people with inconsistent moralities? Does it welcome people with consistent moralities, if consistency requires them to equate their spouse’s life with X strangers for X > 1? What about negative utilitarianism? Does EA welcome the (several, actually) people I’ve met who want to destroy Earth-originating life to prevent suffering? Or the people I’ve met who’d like to find the smallest animal that experienced pleasure, breed as many as possible, and keep them high on opiates?
How large should the EA community be? My model here is that any community is like an arrow on a piece of graph paper. The left-right axis is the size of the community, and the up-down axis is the novelty of the community’s beliefs and practices, relative to the current baseline. It’s easy to rotate the arrow; you can keep rotating it counterclockwise, and create a very fringe group with very extreme beliefs, traditionally called a “cult”. Or you can keep rotating it clockwise, and create a buzzword that everyone likes, but that has no real content. Making the arrow longer is possible, of course, but it’s hard, and most attempts to make it longer actually just rotate the arrow instead. Having a larger group gives the movement more of many resources, but it also makes things more unwieldy and harder to manage effectively.
How closely should what we believe about EA match the reality of EA, and how closely should the marketing of EA match our beliefs? Most people think about this in terms of “honesty” and “dishonesty”, but it’s actually far more complex and subtle than that. For example, my friend Alice (not her real name) talks a lot about the importance of science. She runs a website which (according to her) is intended to help people become better scientists, and stimulate big scientific discoveries. She has some science experience herself, and puts quite a lot of work into this. But I noticed that the scientists I knew hadn’t heard of Alice, so I did an informal survey by taking a hundred or so of Alice’s followers, and looking up their resumes on LinkedIn. Almost none of them were scientists, even for a loose definition of “scientist” (eg. had an undergrad degree in science, or equivalent experience), which IMO was pretty alarming. But Alice wasn’t trying to fool people, or deny the truth; she just hadn’t thought the question was important enough to spend a few hours doing research. Would Alice have been more willing to do the research if she felt less pressure to be transparent? Or does a lack of transparency discourage doing research, because then you have to carry the burden of keeping whatever you find secret? I honestly don’t know. There are all kinds of weird effects here, like where you get more flack for talking about X than for doing X, even if you’re publicly known to be doing X (eg. Warren Buffett was more-or-less openly poly, and this is essentially never mentioned by anyone, while if he were an advocate for poly everyone would think he was nuts).
How should EA divide attention between funding, implementation, and media? Right now, in my own opinion, the emphasis is firmly on media; the leading figures in EA tend to be people like Peter Singer, who are not themselves donors and don’t themselves run non-profits, but who do communication and advocacy work. (An extreme emphasis on funding tends to start looking like Scientology; an extreme emphasis on implementation tends to start looking like voluntourism.) My own view is that all three are important, and that right now implementation is (relatively) neglected, but there doesn’t seem to be a firm consensus on this.
I say this might be overly ambitious, because a focus this general is liable to get sidetracked. It would be very easy to steer left, and spend the event talking in vague generalities, like “EA is about creating moral value through utilitarianism” or some such; it would be equally easy to steer right, and focus on very narrow questions like meat at the conferences. And even if consensus is reached, the media might well just ignore it if the PR strategy isn’t planned out in advance. I think that, before ever talking to any press, people should always spend fifteen minutes to answer the three questions: what am I trying to communicate?, what audience am I trying to communicate it to?, and how do I want that audience to change their behavior? Often people forget this, and wind up talking about whatever happens to come to mind, of which the media then picks out the most controversial or juicy bits and makes them represent the whole organization.
On the issue of communication… I think this is very important, but that we’re going to need a much stronger and more comprehensive solution than just holding a single event. In eg. the AI risk world, MIRI (now that it’s gotten its act together) largely acts as a clearinghouse for talking to all the professional groups interested in safety. CEA could potentially fill a similar role, to some extent. But in AI risk, the important work will essentially come from a smallish circle of experts; public impressions still matter for funding or social status, but there’s no need to maintain continuous lines of communication with ten thousand active participants. In EA there is, and that’s much harder. One proposal I thought up a few days ago was to have an actual, serious, online newspaper dedicated to EA; not just a blog or a newsletter, but a real media source with a full-time professional editor. This would be a substantial commitment, but it might only be a small fraction of the overall CEA budget (you’d know the details there), and I think something like this is sorely needed; historically, towns started getting newspapers once they hit a four-figure population, because that was the only way for everyone to keep track of things. It would also provide a convenient, if imperfect, means of tracking EA “consensus”; right now a blog or Facebook post might be what all EAs think, it might represent a small but vocal minority, or it might be the rantings of a loon. Maybe that’s not exactly the right idea, but we’re going to need something for the left hand to tell what the right hand is doing. My favorite example is that the #1 donor to AMF is Dustin Moskovitz, but I don’t think anyone knows who the #2 donor is; think of all the good this person did for the world, think of all the ideas they might have on how to make money or how to choose charities or how to support EA or how to run effective organizations in general, and yet nobody even knows who they are.