This post by 80,000 hours struck me as more than usually relevant to my interests in developing the art of rationality. It doesn't really say anything new, but it does provide a decent summary of a frame that I think is an important subset of epistemic rationality, in the form of "good judgement". 

More practically, I think of someone with good judgement as someone able to:

  1. Focus on the right questions
  2. When answering those questions, synthesise many forms of weak evidence using good heuristics, and weigh the evidence appropriately
  3. Be resistant to common cognitive biases by having good habits of thinking
  4. Come to well-calibrated conclusions

Owen Cotton-Barratt wrote out his understanding of good judgement, breaking it into ‘understanding’ and ‘heuristics’. His notion is a bit broader than mine.

Here are some closely related concepts:

  • Keith Stanovich’s work on ‘rationality’, which seems to be something like someone’s ability to avoid cognitive biases, and is ~0.7 correlated with intelligence (so, closely related but not exactly the same)
  • The cluster of traits (listed later) that make someone a good ‘superforecaster’ in Philip Tetlock’s work (Tetlock also claims that intelligence is only modestly correlated with being a superforecaster)

Here are some other concepts in the area, but that seem more different:

  • Intelligence: I think of this as more like ‘processing speed’ – your ability to make connections, have insights, and solve well-defined problems. Intelligence is an aid in good judgement – since it lets you make more connections – but the two seem to come apart. We all know people who are incredibly bright but seem to often make dumb decisions. This could be because they’re overconfident or biased, despite being smart.
  • Strategic thinking: Good strategic thinking involves being able to identify top priorities, and develop a good plan for working towards those priorities, and improving the plan over time. Good judgement is a great aid to strategy, but a good strategy can also make judgement less necessary (e.g. by creating a good back-up plan, you can minimise the risks of your judgement being wrong).
  • Expertise: Knowledge of the topic is useful all else equal, but Tetlock’s work (covered more below) shows that many experts don’t have particularly accurate judgement.
  • Decision making: Good decision making depends on all of the above: strategy, intelligence, and judgement.

I do disagree with some of the distinctions being made in the post. As an example, just in the section above, the conception of "Intelligence" as "processing speed" is really flawed, and in-practice intelligence already measures something closer to "good judgement". But overall, the post seems decent as a potential intro into a bunch of rationality stuff.

New to LessWrong?

New Comment
16 comments, sorted by Click to highlight new comments since: Today at 11:04 PM
the conception of "Intelligence" as "processing speed" is really flawed, and in-practice intelligence already measures something closer to "good judgement".

What makes you think so? I know people from Mensa who completely lack good judgment.

Intelligence, as measured by IQ tests, seems to be mostly about processing speed and short-term memory capacity (how many complex you can simultaneously juggle in your head).

Could it be that the rationalist community is a special bubble, where intelligence and rationality are correlated much stronger than in the outside world?

Maybe you mean "processing speed" in some much broader sense of the term? Overall higher scores on IQ rarely come from stuff like "being faster at thinking". Like, one of the most g-loaded things in an IQ test is how large your vocabulary is, which really doesn't have much to do with "how fast you are at thinking". It has much more to do with how good you are at learning and remembering things. 

There are also many spacial rotation tasks for which you often have a really long time and it's much more about "can you rotate things in your head well?". Speed doesn't really compensate for lack of skill here (i.e. giving people more time doesn't lead to many more people getting it right).

Here is an image with all the test categories of the WAIS-IV, one of the most frequent IQ-test batteries: 

You are right that both processing speed and working memory are measured by these tests, but they only make up a bit less than 50% of the test battery. There are also many tests dedicated to verbal comprehension and perceptual reasoning, and if I remember correctly the g-loadedness of the working memory and processing-speed tasks is actually substantially lower than the other tasks in the battery.

The correlation with IQ and Stanovich's "Rationality Quotient" is also quite high at 0.7, so the correlation doesn't only exist within the rationality community, at least for the case where someone tried to create a separate construct that is intentionally trying to be different from IQ, which suggests there is substantial overlap also in other parts of society. 

one of the most g-loaded things in an IQ test is how large your vocabulary is

Is it? I am quite surprised, because that sounds like one of those culturally dependent things that were removed from IQ tests, because they decreased the measured IQ of minorities for not being sufficiently good with the English vocabulary.

Unless you would count the number of words in everyone's native language, like the English vocabulary of a native English speaker versus the Chinese vocabulary of a native Chinese speaker. Except, there is no 1:1 correspondence between English and Chinese words, so you would get wildly different results between similarly intelligent speakers, depending on how specifically the number of known words was measured. (Are plurals counted as separate words? Irregular verbs? "Good / better / the best"? If you already got points for "sand" and "box", do you get another point for "sandbox"?) Are bilinguals considered twice as intelligent, or do we take the maximum, or the union of their vocabularies?

Comparing speakers of the same language, yes, there is a correlation between IQ and vocabulary. But IQ is a biological thing, while vocabulary is relatively easy to change by education. Yes, it is easier to teach hundred new words to a smart child compared to a stupid child, but if you just teach hundred new words to one child and leave the other children alone, that doesn't give the educated child extra IQ points.

Okay, I am now really confused. But I suppose that within a homogenous monolingual community, with standardized education, comparing only with people of the same sex, using school year as a proxy for age... the correlation between vocabulary size and IQ could be quite strong.

The correlation with IQ and Stanovich's "Rationality Quotient" is also quite high at 0.7

Correlation, sure. I'd say that certain level of intelligence is necessary for certain level of rationality. But the original article hinted that these two may come apart at the extremes. You seem to dismiss this possibility without providing an argument. Correlation 0.7 is still compatible with the possibility that some of the high-IQ people are very high on rationality, but many are average or low.

(I am not good at statistics, but I made a toy model where first IQ was generated as a random number between 50 and 150, and then RQ was generated as a random number between 50 and IQ, to simulate the hypothesis that intelligence is a precondition, but not a guarantee of rationality. The correlation in this model was something above 0.6, and yet many characters with very high IQ didn't have very high RQ. Of course this doesn't prove that reality works like this, but it shows that mere correlation does not disprove the possibility.)

And then there is my anecdotal experience with Mensa, where I have also heard similar stories from other people living in different countries. I shortly joined the local Mensa Facebook group, and then quickly left it, because it was a constant source of conspiracy theories, and people claiming to have disproved theory of relativity. (I suspect that intelligence also correlates with being a contrarian, and being a contrarian correlates with believing in conspiracy theories.) Maybe Mensa specifically selects for a combination of high IQ and low RQ, dunno. Then again, maybe rationality community specifically selects for high RQ.

Yep, vocabulary is pretty g-loaded. The reasoning is pretty straightforward. The size of your vocabulary (in whatever language you have grown up in) is a pretty good measure of your learning rate over basically your whole life, since people who are better at learning will have picked up many more words over the course of many years. This table showed up early in the google search results, but I've seen the same numbers in a bunch of other places (my favorite reference is this book). Ignore the highlighting, that's just what the Google image gave me. The relevant test is obviously "Vocabulary".: 

The correlation also holds up under much less restricted conditions. Again, I recommend Macintosh "IQ and Human Intelligence" as a good reference on all this, which has multiple chapters dedicated to this.

And then there is my anecdotal experience with Mensa, where I have also heard similar stories from other people living in different countries. I shortly joined the local Mensa Facebook group, and then quickly left it, because it was a constant source of conspiracy theories, and people claiming to have disproved theory of relativity. (I suspect that intelligence also correlates with being a contrarian, and being a contrarian correlates with believing in conspiracy theories.) Maybe Mensa specifically selects for a combination of high IQ and low RQ, dunno. Then again, maybe rationality community specifically selects for high RQ.

My model of mensa is that it self-selects on the negative tail of IQ correlations. Since if you were good in more ways than your IQ would indicate, there is a good chance you would hang out somewhere else than Mensa. It also doesn't really seem to deal well at all with the problem of performance on IQ tests being highly trainable, and I think I heard many Mensa members often took multiple IQ tests in order to get in, giving rise to natural goodheart's law problems. 

But overall my bet would still be that believe in conspiracy theories and general decision-quality is still much much better at Mensa than in the average population. Mensa simply isn't that high of a bar, in particular if you get any substantial training effects.

Thanks, I appreciate you taking time to respond to my objections. Will read the book.

I agree that Mensa selects for "not having better ways to spend your time". I think you probably overestimate the impact of retraining -- there are many people who keep taking the test every year, and keep failing every year. (Perhaps retraining properly already requires some intelligence threshold?) I agree that Mensa isn't high enough bar -- it's like a slightly above-average university, except that you can meet dozen people in local Mensa and hundreds in local university.

Maybe Mensa specifically selects for a combination of high IQ and low RQ, dunno.

I'm curious what the actual cutoff for Mensa is. 

Googling, I get that the cutoff for getting into Mensa is 98th percentile in IQ, which means in the US this means you're in the top 6.5 million people in terms of IQ. This doesn't seem to me to be a particularly high cutoff. I'd imagine most Math PhDs form a far more selective group, given that I expect they're all in that set and also there's probably only order-of-magnitude 10k Math PhDs, and also they have been selected more on conscientiousness and competence (insofar as Mensa has no work required other than the test, whereas mathematicians have way more intellectual hoops to jump through).

(I don't know what exactly this implies, mostly just noticing things aloud here.)

But then the rationality community is full of people who think they have disproved to Copenhagen Interpretation. Maybe the rationality community deals in controversialism, but their controversies aren't so salient to people in the community.

I predict, and would be willing to bet if we can operationalize it, that: if we compare [the opinion you refer to expressed by some people in the rationalist community] to [the "this disproves relativity" opinions Villiam is refering to], the rationalist community opinion is more mainstream among physicists than the Mensa opinions.

The degree of confidence with which it is held is not usual among physicists, and a number have objected to it over the years.(Generally receiving short shrift)

I can see a difference of degree between lesswrong and mensa, but only a difference of degree. There is no need to explain why mensa are contrarian, as if you are completely different

Can you be more specific? There are many things you could mean by "difference of degree". Some of them seem quite bad if true, and some seem perfectly benign.

Like sure, both are nonzero on some scale of contrarianism. But I predict that, for example, separating the world into "zero on that scale" and "nonzero on that scale" is mostly a silly categorization. (Even if the optimal amount on that scale is zero, which I doubt.) If all you mean by "difference of degree" is that they're both nonzero on that scale, then I think it says very little.

Whereas if you mean, for example, something like... "if you asked mainstream physics PhDs for their thoughts, they'd say "this is a dumb opinion to hold" only slightly more often for the mensa opinions than for the LW opinion"... if you mean something like that, then calling it "only a difference of degree" seems like a significant claim that would reflect badly on LW if it were true. But I don't think it's true.

What I mean by "difference of degree" is "NOT difference of kind".

For comparison, this lecture was given at European Mensa meetup in Prague six years ago.

Short version, it "disproves" the theory of relativity by proposing the existence of Global Aether. The theory "avoids quantitative details to keep the difficulty level as low as possible". It also debunks quantum physics, because quantum physics literally accepts magic, defined as "forces from virtual fields with mathematical properties and without material or tangible cause". Et cetera.

The really bad part was when I later tried to explain to Mensa members why the lecture was obviously nonsensical, because we already have a lot of experimental evidence in favor of relativistic effects (basically sharing this link and providing a short summary, such as "the GPS in your smartphone must calculate using relativistic equations, otherwise it would give wrong results"), so any alternative theory would need to explain these effects too, not just ignore them and insist on flat space-time.

The consensus of Mensa members was: "He is an internationally respected author who published a lot of books" (translation: He has a website with a list of dozen self-published books), plus the usual arguments that "people deserve to hear alternative opinions" and "according to Popper, scientific theories cannot be proved anyway". With the recommendation that if I misunderstood something, I should ask the author, instead of wasting everyone's time.

...so, one of the reasons I don't go to Mensa meetups, despite technically being a (former) member.

I believe there is a huge difference between "Copenhagen vs MWI" and "relativity vs aether" controversies.

Are you a physicist?

“Intelligence” as “processing speed” is really flawed, and in-practice intelligence already measures something closer to “good judgement”

Intelligence or IQ?

Both, though of course "intelligence" without reference to IQ is just super vague and people use it to mean all kinds of things. I do think the concept of IQ is usually narrower than intelligence, and IQ already doesn't really feel well-described to me by something like "processing speed".