On Taleb, and what counts as a scientific theory
Nassim Nicholas Taleb — the former options trader and author of The Black Swan and Antifragile — was, by 2015, frequently invoked at effective-altruism gatherings as a heterodox thinker on risk. He had also begun a high-profile campaign arguing, on “precautionary” grounds, that genetically modified crops posed a hidden civilizational risk — a position most working geneticists considered confused. Oliver Habryka had cited Taleb approvingly the previous evening; the response below disputes both the framing of Taleb as a “scientist” and, more broadly, what should and shouldn’t qualify as a scientific theory in the first place.
I don’t know if we actually disagree on anything, since I don’t remember exactly the phrasing used. But because this is a really important issue for our community’s memetic ecosystem, I did a quick review of Taleb’s biography on Wikipedia (and elsewhere). And Taleb just isn’t a scientist. He isn’t remotely any kind of scientist in any way, shape, or form. He has never done science, either formally or informally. He has never studied science, either formally or informally. I doubt he even took any science classes in high school (he grew up in Lebanon, and at the time Lebanon was in a massive civil war). He’s very smart, certainly, but so is Eliezer, and it’s no more accurate to call Taleb a scientist than it is to call Eliezer a brain surgeon. In fact, IMO it’s somewhat more accurate to call Eliezer a brain surgeon, since at least Eliezer has an amateur’s knowledge of neuroanatomy. I wouldn’t say it’s impossible for a scientist to have made Taleb’s mistakes on GMOs, since all kinds of weird things can happen once in a while, but it would be vastly less likely. (Of course, scientists in turn are more likely to make other categories of mistakes.)
Taleb has lots of ideas, of course, but AFAICT none of them would remotely come close to qualifying as scientific theories, even assuming he’d never said anything false. (FWIW, the same could be said for eg. a lot of the stuff on my blog, or a lot of Eliezer’s writing.) A scientific theory is much more narrowly defined than the space of ideas in general. Broadly speaking, a scientific theory should have: A well-defined regime of applicability. Quantum electrodynamics only applies to electrons and photons, not protons; it certainly doesn’t apply to “particles” like raindrops or grains of sand. The theory of evolution is about the change in genotype frequencies over multiple generations, not the “evolution” of corporations or stars or romantic relationships, even though people might use the same word. Eliezer wrote about this type of thing in his excellent essay The Virtue of Narrowness . The regime-of-applicability constraint instantly rules out a lot of pop sociology, since many words that people use to talk about groups don’t correspond to physical categories which it’s meaningful to generalize about. It’s probably impossible, for example, to have a scientific theory of “corporations”, since a “corporation” is like a hard drive “folder” that can be willed into existence at any time for any reason (a small hedge fund, for instance, might consist of ten different “corporations” despite only having three employees).
Within its regime of applicability, a scientific theory should be universal. “I own many colorful socks” is true, but it isn’t a theory; it’s a statement that only applies to me (and only to me in 2015, not many past or future mes). A lot of political “theories” fall down here, since they will often try to “explain” only one or two events, and just ignore similar events that had different outcomes.
Of course, it should make clearly-specified, falsifiable predictions. However, the requirements are really stricter than that. A trader like Taleb predicts stock prices, but those predictions don’t qualify as a theory; a theory should be a well-specified machine that anyone can use to predict the world. (Which is why it’s more-or-less impossible to have a “science of trading”, since any theory of trading would become obsolete right after publication.) To quote Sean Carroll: “Your discovery should be the same way. If it’s a revolutionary new theory, it should be a theory that anyone can use. That means it needs to be clearly expressed and unambiguous. I’ve had more than one long and fruitless discussion with alternative scientists who would say ‘You tell me the experimental result, and I will explain it with my theory.’ That’s not the way it works. Your theory should have a life of its own; it should be a machine that I (or anyone) could use to make predictions.” ( The alternative-science respectability checklist .)
It should exist within a surrounding framework of concepts, past theories, and known experimental results. This is critical, because it means anyone can easily tell where work agrees with past research, where it contradicts past research, and where it makes its own inventions. In some of the less-healthy fields in the humanities, for example, every new researcher makes up their own mesh of terms, concepts, observations, and predictions de novo . A quick test for this is how well you have to know who the past researchers in your field are. I know about the Stark effect in physics, for example, but I don’t have the foggiest idea who Stark was (in fact, I just looked him up, and it turns out he was a Nazi!). But in a less healthy field, since each researcher makes up all of their own ideas, studying past work becomes a hagiography where you have to look at a person’s life and their beliefs as a whole before you can understand what they’re saying.
I’d recommend the reviews of Wolfram’s book A New Kind of Science by Shalizi and Weiss as a quick example of how someone can be extremely smart, mathematically adept, say things that sound surprising and reasonable about fascinating topics, and yet not be doing science in any real sense.