Conservation of weirdness, and other principles for movement-building
Sent to Tara Mac Aulay — then a senior figure at the Centre for Effective Altruism — the day after her talk at EA Global 2016. Tara had argued for a broader conception of the kinds of roles people in EA could take on. Alyssa picks up that thread to lay out three of her own working principles for thinking about a growing movement: the trade-off between size and intellectual distinctiveness, the prevalence of neutrals over allies and enemies, and why naive head-counts are a misleading measure of progress.
The first principle I use might be called “Conservation of Weirdness”. Roughly speaking, it says that groups obey this equation: E = S * W where E is the amount of effort it takes to maintain the group, S is the size of the group, and W is the level of “weirdness” — how different the group’s ideas are from the mainstream. To do a really large amount of good, S has to be moderately large (the group can’t be too tiny), and W has to be at least moderately large too; if the world’s problems could be solved by doing everything the standard, boring, socially-approved way, they’d have been solved already. Therefore, the leaders of the movement have to maintain good levels of both S and W.
Unfortunately, if we hold E constant, then we get this equation: W = E/S that is, S and W are reciprocally related; increasing one tends to decrease the other. As you increase weirdness, size will “want” to decrease; likewise, if you increase size, weirdness will “want” to decrease, in the same way that water “wants” to flow downhill and fluorine gas “wants” to set things on fire. You can overcome this with enough effort, but since you’re fighting against the natural tendencies of groups, you have to make sure there are no “leaks”, no places where S or W will try to sneak out from under you. If you try to push water uphill, it will flow out under the cracks in your fingers; if you pump it up with a hose, it will try to spray out the side; only when there are no holes anywhere will it flow to the top.
As an example, suppose you think all buildings ought to be painted yellow. Relative to current consensus, this has a pretty high level of W. You form a group of Yellowists, and start advocating for your position, lobbying housing developers, and so on. Soon, a billionaire comes along, and they offer to pay for a big PR campaign — you’ll blanket the whole country with messages about Yellowism. The PR campaign works so well that a year later, Congress passes the Yellowist Act of 2019, which allocates $500 million in federal funding for the painting of school buses, and in extra subsidies to the lemonade industry, which has nothing to do with buildings. What happened was that your hose was leaky — you didn’t pay close enough attention, and all the W leaked out when “Yellowist” got redefined, right under your nose.
There are several different strategies to combat this, all of which have their own tradeoffs (sadly, no free lunches here). First, try to keep S smaller when you can. Only grow when there’s a strong reason to grow; don’t grow just for the sake of it. Second, try to attract people with high levels of Openness, in the Big Five personality model. You’ll be able to maintain a higher level of W, without having to put in so much effort. Third, try to encourage honesty, and clear language with specific, narrowly defined terms. This doesn’t mean enforcing ideological conformity; for example, if someone thinks Nick Bostrom is a crackpot, I might work with them on other issues, even though I strongly disagree. But they can’t think Nick Bostrom is a crackpot, and then talk about how they want to fight existential risks, where by “existential risks” they mean right-wing politicians being elected and people eating GMO foods. That’s not what “existential risks” means; people can’t be allowed to hijack words for their own pet projects. Say no to word piracy. :)
The second principle is that the vast majority of people are neutral, rather than good or bad. They aren’t working against you, and might cooperate with you on specific issues (if that fits their needs), but they also aren’t really your friends. This is very important in fields like science, where the vast majority of research doesn’t pan out. Suppose Professor Bob Smith shows up, with a new plan for a miracle battery that stores twice the energy in half the space. Smith (and Smith’s supporters) will try to push people into a “good” or “bad” judgement — either Smith is a fraudster, or he’s an idiot and has deceived himself, or his battery must really work. Most people are (understandably) reluctant to make the first two claims, especially if Smith is an established scientist with a good track record, as many battery inventors are. Therefore, they’re pushed into accepting the third claim. Unfortunately, this is a bad idea too, because the overwhelming majority of battery startups fail. It’s important to take a neutral position — to say that Smith’s battery doesn’t work, and you shouldn’t invest in it or spend time working on it, but this doesn’t mean Smith is evil, stupid, or otherwise bad. (Religious apologists also do this — you might have heard the phrase “lunatic, liar, or Lord”, in reference to Jesus.)
Therefore, instead of thinking in terms of “values”, I often prefer to think in terms of “superpowers”. Values divide everything into good and bad — you’re either altruistic or selfish, either hard-working or lazy, either skeptical or gullible, either rational or emotional. But it seems more accurate to say that, for example, a few people have the Competitiveness superpower; they are driven to excel in contests, to out-do others or themselves, and will overcome any obstacle to winning. A few other people have the Diplomacy superpower; they are very friendly and socially savvy, and good at forging deals between competing parties with different interests. These powers are somewhat opposed, but that doesn’t mean everyone falls into one bucket or the other; most people have neither of these superpowers, they are neutral. It’s a bad idea to make any given superpower a pre-requisite for joining, since almost everyone won’t have it, and since you need many different powers to run a big organization well. (You might be able to get away with one, like how Google required the Hacking superpower even for non-technical roles, but I wouldn’t want to try it.) Likewise, it’s important to try and a) recognize what powers people do have, even if they’re ones you don’t use yourself or aren’t used to seeing, and b) remember that almost everyone will lack any given power, even if they’re otherwise very capable, and you shouldn’t trick yourself into seeing things that aren’t there.
Goals and motivations — eg. things like altruism vs. selfishness — are also a thing, but I think they tend to be less important than superpowers, since many goals will converge to the same decision given the same resources. (As the most obvious example, existential risk is important no matter what you care about, but there are many smaller examples too.) There are a handful of exceptions, like the negative utilitarians, but they tend to be fairly rare. I think the most important exception is to distinguish short-term from long-term goals; people with only short-term goals tend to be much harder to cooperate with. If they have only short-term goals, but don’t care very much about them or aren’t very effective, they’ll tend to do things like play video games all day, and will add overhead to the group without contributing much. If they have short-term goals, but do care a lot about them and are effective, this can be much worse; such people might burn down organizations, even including themselves, to get what they want right now.
The third principle is that a lot of things, especially people, organizations, projects, and events, are exponentially distributed; it’s therefore a bad idea to evaluate progress by counting them. There is an overwhelming temptation to do so — I’ve done it myself a lot, since metrics are so vital, and counting is so easy. It just feels great to plot number of papers published, or number of members joined or number of events hosted, and watch the counter go up and up as you put more work in. But it must be resisted. Nor can you simply say (for example) that the odds of a project making it big are 1%, so all you have to do is count the projects, perhaps the projects that pass some quality bar you set, and divide by a hundred; the odds of something being big aren’t just a constant factor you can adjust for by dividing, since they vary so much depending on what pool you’re drawing from. Paul Graham wrote about this with respect to startup investing ( Black Swan Farming ), but I think it applies to many other things too. See also this thread on Hacker News .