The sometimes perceptive, sometimes bizarre, but usually interesting Eliezer Yudkowsky has a new set of posts up at lesserwrong.com, excerpts from his book Inadequate Equilibria. (Lesserwrong.com is the successor site, established in 2017, to the rationalist site Less Wrong.) The first post of interest is
Inadequacy and Modesty.
Yudkowsky has two main topics: One is when to trust one’s own judgment, when one disagrees with experts, versus going with the experts’ opinion. The link above talks about informationally efficient and inefficient situations, where we can roughly define an informationally efficient situation as one where the experts are as right as possible given currently available information.
His second topic is the set of ways that a society can get stuck in a suboptimal equilibrium. That’s an enormous topic, which I’ll take up in a later post, but the connection is this: One example of such suboptimality is when there are insufficient incentives for the discovery and spread of information. The dissemination of information – who knows what – is obviously connected to the topic of expertise.
For me, there are two main items of interest in all this. One is random walks and their links to rational beliefs, and the other is the question of expertise itself.
A random walk is a kind of variable that often arises in informationally efficient situations. Yudkowsky uses the classic exanple of stock prices, though he doesn’t use the term “random walk.” He discusses the reasons to believe that stock markets are informationally efficient, which means that all relevant information known to market participants is already incorporated into stock prices. That in turn implies that you can’t profit by second-guessing the market, because the expected (i.e., mean) change in the price is zero. That in turn is the definition of a random walk. (If you’ve ever taken a Finance class this may sound familiar; it’s the Efficient Markets Hypothesis.)
See my post The Mind Cannot Foresee its Own Advance, which presents a generalization of this point.
There are good reasons to think that stock markets are pretty damn informationally efficient:
Empirically, it’s extremely difficult to beat the market – for periods of time long enough to not just be temporary luck – even with Cray big iron and a truckload of quant PhDs on staff. This empirical regularity is the real meat of the argument. But why is that the case? Several reasons, noted by Yudkowsky (but my wording here):
1. There are enormous incentives $$$!!! to uncover patterns that other market participants haven’t uncovered, so you can second-guess them and make bug bucks,
2. There are lots of people involved in attempting to do this constantly, which tends to push securities prices in the correct direction,(*)
(*) If a security is underpriced you should buy it to profit when people eventually realize it’s underpriced and the price rises. But also, the very fact of your buying it constitutes an increase in demand for it, which tends to push the price up. A symmetric phenomenon happens when you bet on a price falling in the future.
3. There is fast feedback from empirical reality telling them whether their trading strategies are successful, so fast error correction,
4. You can bet either way in the stock markets. That means that whether everybody else is too optimistic or too pessimistic, there are bets you can place to profit when the current mispricing is eventually corrected.
At the link above Yudkowsky says, somewhat floridly,
In the thickly traded parts of the stock market, where the collective power of human civilization is truly at its strongest, I doff my hat, I put aside my pride and kneel in true humility to accept the market’s beliefs as though they were my own, knowing that any impulse I feel to second-guess and every independent thought I have to argue otherwise is nothing but my own folly. If my perceptions suggest an exploitable opportunity, then my perceptions are far more likely mistaken than the markets. That is what it feels like to look upon a civilization doing something adequately.
A very important point here is that stock markets are exceptional! As points 1 – 4 explain, there are reasons to think the experts – in this case, securities trading firms – are correct on average.
But in the generic case – consider the field of history, e.g. – none of those things is true. And the whole question of who is an expert basically scuttles this attempt to say “You should often defer to experts.”
For example, Marxist historians like to say that they’re experts on history – they’ve figured out its ineluctable laws! – but they’re actually a bunch of ideological fuckwits who can’t think themselves out of a paper bag. But they’re also professors at many a US university (all the universities, I think). They’ll tell you they’re experts, man. It says so right here on the label, “I’m an expert.”
So is that a reason to regard them as experts?
Plainly not. Well, but we don’t have to trust them on this; after all, they’re hired by universities! Universities must be unbiased; it says so right on the label! Right? No, actually universities (outside of STEM fields, and increasingly even there) are also largely a bunch of ideological fuckwits.
Or so I claim. Am I right or wrong? Before you respond, I claim that I’m an expert on this topic. Hmmmm, we’d need to assess their degree of ideological fuckwittery empirically, wouldn’t we?
Or are we going to count heads? Whose heads? The historians’? But they’re the ones whose very credibility is being questioned, so that would be going in a circle.
How about the average person? Well, the Marxists lose that one, since most people aren’t Marxists. More to the point, if we’re asking people other than soi-disant “experts,” we’ve already departed from the dictum “trust experts.”
(By the way, should we trust the pollsters who are doing the head counting? Are their polling methods unbiased? Are they competent polling experts? How do you know?)
Furthermore, there are no particular incentives to be correct in this area. Your fellow historians are largely leftists who will grant you tenure for saying, in various ways, “Socialism is nice. Capitalism is bad.” You are not betting your own money or your life by predicting that next time – in China… Cuba… Venezuela! – socialism is sure to work.
Marxism is not intended to factually describe reality anyway. It’s an ideology of power, a convenient pre-made language for people who want to seize power, and realize they need a veneer of justice to help them gain adherents, put their opposition on the defensive, etc. Or as Marx himself put it, more discreetly: “The philosophers have only interpreted the world, in various ways. The point, however, is to change it.”
Once it’s recognized that you can’t just trust anyone who says “I’m an expert” – anyone can say that – the whole epistemic question returns to the forefront. How are you going to judge who’s an expert? Hmmm, maybe we need some standards regarding the use of evidence. (Frequentist? Bayesian?) Also some procedures, like whether testable empirical claims are being made, whether they’re replicable and actually replicated, etc.
You see the problem: In order to judge who’s an expert, you have to be damn far along the trajectory of knowledge in the relevant subject, far enough along that you’d be a jack-leg expert yourself on the topic. You might as well just assess the evidence for yourself, ignoring the purported “experts.”
What I do, and I hope everyone does, is try to identify areas in which there don’t seem to be monetary or ideological incentives to be biased, and provisionally trust the experts in those areas, and ignore experts in areas where the incentives are bad. This is far from foolproof, of course. (I’m perpetually surprised by how politicized nutrition science is.) But given the impossibility of becoming experts ourselves in all topics, we use heuristics, imperfect though they are, to try to avoid getting scammed by fraudulent “experts.”
There are many reasons, not just malign intent, that this instinct to mistrust “experts” is sound. Yudkowsky mentions some and I will mention others in my next post.
Here’s a teaser: In my next post I will cite a paper by two game theorists on the topic of expertise. This paper was published in The Quarterly Journal of Economics, which is a top Economics journal in the world. A major conclusion of the paper: In general, the equilibrium outcome is that experts will deliberately not inform you perfectly. Now here’s my question for anyone who says “trust experts”: Do you trust the conclusion of these two game theory experts?