The South Sea Bubble, and a survey of the EA Facebook group
Sent to a small group of EA-adjacent friends — among them Nick Tarleton, Kelsey Piper, and Tara Mac Aulay. The argument: a movement that has come to spend most of its airtime promoting itself, with no clear external goal to anchor that promotion to, looks structurally similar to history’s most famous self-referential bubble.
Last night, I watched an online documentary about the South Sea Bubble of 1720. In addition to being entertaining, it was an interesting study in how an organization can… I don’t know if there’s really a word, something like “wrap around in on itself”. The South Sea Company got started because the English government was in debt, and they needed to get better terms on their bonds. The South Sea Company would buy up bonds from the public, in exchange for South Sea stock, and the government would then pay a discounted interest rate on those bonds to the company. To make up for the lower rate, and therefore make the stock a good deal, the company was given a monopoly on South Sea trade. This was, in retrospect, dumb — the South Sea ports were all controlled by Spain, which didn’t want their trade going on British ships — but it didn’t sound dumb at the time. The East India Company was wildly profitable, and many thinkers like Isaac Newton and Jonathan Swift invested their money in the South Sea Company.
When trade revenue didn’t materialize, in order to get cash, the South Sea Company started a marketing campaign to make money by selling its own stock. This raised the share price, which created more demand for the shares, which in turn raised the share price more, and so on, starting a big wave of hype. The South Sea Company developed ever more complicated schemes to market itself, and eventually the South Sea Company would use money from selling South Sea stock to loan out to people wanting to buy more South Sea stock, which they could then pay off by selling small pieces of the stock as the price rose, while making money on the 30% annual South Sea dividends, which the company could pay from dividends being reinvested in South Sea stock. The purpose of the South Sea Company was to promote people buying into the South Sea Company. Eventually, of course, it all fell apart.
To measure how much the purpose of Effective Altruism had become the marketing of Effective Altruism, I did a quick-and-dirty survey of the EA Facebook group posts, from June 1st through today. It’s important not to attach too much importance to single quotes, but… this one in particular still stood out: “(Suppose that) the government therefore decides to keep taxation at the current level, but spends it all on EA causes such as Promoting effective altruism , AI research and policy advice, Ending factory farming, etc.” (emphasis added) Of course, categorizing posts isn’t an exact science, and I’m sure others would get slightly different numbers, but here’s what I found. There were 163 posts in total over the twelve weeks, or a rate of about two per day. Of those, 99 posts (61%) were meta-posts about marketing Effective Altruism, organizing Effective Altruism events, doing outreach to donors about Effective Altruism, the media image of Effective Altruism, and so on. 41 posts (25%) were about a bunch of miscellaneous topics, things like random stories in the news, Peter Singer’s birthday, people’s posts about their own lives, and so on. (I threw all of 80,000 Hours’ posts in here — their advice is a mix of optimizing personal happiness, optimizing earning power, optimizing social capital, finding people to work for EA outreach orgs, finding people to directly work for charities, etc, so it doesn’t seem cleanly categorizable.) 8 posts (5%) were about moral philosophy in general, which was actually lower than I expected. The remaining 15 posts (9%) were about object-level causes; either problems that the poster thought needed more attention, or solutions that they thought were overlooked. Of the 15, 11 were… for lack of a better word, pretty unoriginal. They advocated for causes that had already received huge ongoing media coverage, like water desalination, promoting vegetarianism, blood donation, disaster relief, and so on. One post seemed, as I think the poster acknowledged in the comments, to be mostly a rehash of ideas previously floated by GiveWell. The last three seemed at least somewhat new, and while I don’t really endorse them, I’ll link them here because, hey, they deserve credit for trying, and two of the three didn’t even get any comments: facebook.com/groups/effective.altruists/permalink/1086818664707823 facebook.com/groups/effective.altruists/permalink/1093892264000463 facebook.com/groups/effective.altruists/permalink/1118416878214668 One thing I suggested to Will MacAskill before EAG, but which I sadly don’t think he ever responded to, was the need to clearly and concretely define what exactly Effective Altruism was. A clear definition serves as a bulwark against organizations wrapping in around on themselves, by putting focus on a commonly shared external goal (one other than simply increasing the organization’s membership). When people define Effective Altruism, they usually use terms like “using evidence” and “maximizing impact”, but these phrases are by themselves very vague. Eg., Ben Goertzel, a very smart AI researcher I know, has a strong belief in human psychic powers. When I disagreed with him, he said that his belief was the one supported by using scientific evidence. After all, there are hundreds of peer-reviewed scientific studies, by dozens of labs over decades, that all provide proof of psychic powers. (That’s not a joke — there really are that many real studies.) It was, Ben said, the scientific community that wasn’t using evidence, since they unfairly refused to consider all this legitimate, mathematically sophisticated and detailed research. Of course, I think Ben Goertzel is wrong here, but that just means I’m using a different definition of “evidence”, and that the word “evidence” doesn’t itself differentiate between my position and Goertzel’s, or for that matter the positions of thousands of other people.
With such a vague definition, it isn’t really possible to construct external metrics for evaluating Effective Altruism’s success. If the goal were to eliminate malaria, for example, the first thing anyone would do would be to put up a big chart tracking the number of malaria cases. If the goal was to reduce existential risks, well, that’s a lot harder, but at least you can make guesses and toy models that estimate x-risk odds. But how would one evaluate progress towards goals like “evidence” and “impact”? A few years ago, OPP switched from focusing only on charities supported by randomized controlled trials to riskier charities, which might have higher returns over time. Is that progress forward, since it acknowledges the evidence showing RCTs are imperfect? Or is it backward, since it’s moving away from the solidity of hard science towards philosophy and guesswork, which are more susceptible to individual biases? Does a focus on x-risk have a greater impact, since it helps save a large number of expected future lives? Or is that a lower impact, because (as one wing believes) future lives aren’t valuable compared to present lives, or (as the other wing believes) impact is about minimizing suffering, and a post-x-risk world has no suffering at all? With no real external metrics, the only ones you can measure are the meta-metrics about the Effective Altruist cause being to promote Effective Altruism: number of members, amount of website traffic, amount of money raised, and so on. This is much more honest than the South Sea Company version — where there were external metrics, they just weren’t visible to anyone because the cashier cooked the books — but the end result is the same: without a clear external goal, the only real goal becomes to expand for the sake of expanding.