A New Yorker article on cognitive biases was on the reddit front page in the last day. It seems like a good opportunity to discuss what the article got right, what it didn't, how reddit reacted, and how one might better publicize the topic in the future.

reddit comments(this link doesn't work any more, see VincentYu's comment for links)

article

New to LessWrong?

New Comment
26 comments, sorted by Click to highlight new comments since: Today at 10:04 PM

Full paper (West, Meserve, & Stanovich, 2012).

The New Yorker article ("Why smart people are stupid") by Lehrer misrepresents the main point of West et al.'s paper, which is that – as the original paper title asserts – "cognitive sophistication does not attenuate the bias blind spot", instead of the article's claim that smart people are more susceptible to cognitive biases.

Three negligent misrepresentations by Lehrer:

First, contrary to the general message of the article, West et al. state twice that most cognitive biases are negatively correlated with cognitive ability (emphasis mine):

The finding that the bias blind spot is more apparent among the more cognitively sophisticated individuals is contrary to much of the rest of the heuristics and biases literature where most biases are negatively correlated (or at least independent) with cognitive abilities (Bruine de Bruin et al., 2007; Chiesi et al., 2011; Del Missier et al., 2010, 2011; Finucane & Gullion, 2010; Kokis, Macpherson, Toplak, West, & Stanovich, 2002; Stanovich & West, 1998, 2000; Toplak et al., 2011; Weller, Levin, & Denburg, 2011; West et al., 2008).

Most cognitive biases in the heuristics and biases literature are negatively correlated with cognitive sophistication, whether the latter is indexed by development, by cognitive ability, or by thinking dispositions (Bruine de Bruin et al., 2007; Chiesi et al., 2011; Finucane & Gullion, 2010; Kokis et al., 2002; Toplak & Stanovich, 2002; Toplak et al., 2011; Weller et al., 2011; West et al., 2008).

Second, West et al. refer to the bias blind spot as a metabias because it is a pattern of inaccurate judgment in reasoning about cognitive biases. In this context, the distinction between bias and metabias (which the article neglects to maintain) is vital because the whole point of West et al.'s paper is that this particular metabias does not exhibit the same negative correlation with cognitive ability that many cognitive biases exhibit. They do not contest or assert to have a counterexample to the claim that most cognitive biases are negatively correlated with cognitive ability.

There is an important metabias that remains unexamined in all of the previous work on individual differences, however. It is the so-called bias blind spot—explored in an important article by Pronin, Lin, and Ross (2002). They found that people thought that various cognitive and motivational biases were much more prevalent in others than in themselves. Bias turns out to be relatively easy to recognize in the decisions of others, but often difficult to detect in one’s own judgments.

Third, West et al. are rather cautious (in contrast with the article) in stating the correlation strengths of bias blind spots with cognitive ability. For instance, they have as their paper title "Cognitive sophistication does not attenuate the bias blind spot", rather than a stronger "Cognitive sophistication augments the bias blind spot". However, they do find a slight positive correlation (two-tailed p < 0.05):

We found that none of these bias blind spot effects displayed a negative correlation with measures of cognitive ability (SAT total, CRT) or with measures of thinking dispositions (need for cognition, actively open-minded thinking). If anything, the correlations went in the other direction.

Whatever explains the slightly positive correlation, a conservative way to characterize the findings here is to say that cognitive ability provides no inoculation at all from the bias blind spot—the tendency to believe that biased thinking is more prevalent in others than in ourselves. In our data, cognitive ability did not attenuate the tendency toward a blind spot at all. Thus, the bias blind spot joins a small group of other effects such as myside bias and noncausal base-rate neglect (Stanovich & West, 2008b; Toplak & Stanovich, 2003) in being unmitigated by increases in intelligence.


Social news posts about the New Yorker article:

ETA: OB post

The article is pretty annoying, but after thinking about this for a day the positive blind-spot/IQ association doesn't seem implausible at all. Smart people tend to know they're smart (and tend to overestimate how far above the mean they are). And lots of people associate cognitive biases with low intelligence (in part because they are associated for many biases). So it shouldn't be that surprising that smart people think they're less likely to suffer from a cognitive bias after one is described to them given that they're unlikely to be familiar with past West and Stanovich research on the relaton between that bias and IQ. Someone needs to replicate the experiment with descriptions of biases that include something like "Smart people are just as likely to suffer from this bias as anyone else." Then see if the positive correlation remains.

I'm missing something basic, here: If cognitive sophistication does attenuate other biases, how can they tell that the cognitively sophisticated believe themselves less biased than the average person to the point of being biased in that belief, rather than merely to the point of being correct?

Having read the paper, I don't think they can, and they admit this to some extent. What they can prove is that in the small selected set of biases they chose, the smart are incorrect; but to show that the smart are miscalibrated in general about their resistance to biases, one would need to do a lot more work. (There are scores of biases, after all.)

They studied the bias blind spots of cognitive biases that show no correlation with cognitive ability.

The finding that the bias blind spot is more apparent among the more cognitively sophisticated individuals is contrary to much of the rest of the heuristics and biases literature where most biases are negatively correlated (or at least independent) with cognitive abilities [...] Given the rarity of the result, it is important to consider alternative explanations for it, and an obvious one suggests itself. Perhaps more cognitively sophisticated individuals actually are less likely to display these classic cognitive biases. Such a finding would in fact cause a positive correlation if more cognitively able individuals correctly perceived that they were less prone to bias. This alternative explanation has diminished likelihood, however, when we reflect on the fact that these cognitive biases have been found in another study to be independent of cognitive ability (see Stanovich & West, 2008b, Table 1).

We explored the obvious explanation for the indications of a positive correlation between cognitive ability and the magnitude of the bias blind spot in our data. That explanation is the not unreasonable one that more cognitively sophisticated people might indeed show lower cognitive biases—so that it would be correct for them to view themselves as less biased than their peers. However, as the analyses in Tables 3 and 5 indicate, we found very little evidence that these classic biases were attenuated by cognitive ability. More intelligent people were not actually less biased—a finding that would have justified their displaying a larger bias blind spot.

ETA: On further thought, the lack of correlation merely simplifies the study, and is not a necessary assumption – for each subject, they administered tests on the selected cognitive biases, obtained several measures of cognitive ability/sophistication (CRT, SAT, etc.), and asked for subjective assessments on the extent of the biases:

To what extent do you believe that you are likely to commit base-rate neglect?

To what extent do you believe that the average JMU student is likely to commit base-rate neglect?

They used the subjective assessments along with the results of the administered tests to derive the extent of bias blind spots, independent of the measures of cognitive ability. These are then combined in a standard correlation study, and your hypothesis can be tested.

VincentYu has it right. They 1) used biases that had previously been shown to be uncorrelated with intelligence and 2) checked and confirmed that the biases actually were uncorrelated with intelligence in their sample.

If that were true, and if the relation between measures of "cognitive sophistication" and of bias were continuous, then by the intermediate value theorem there would be some level of "cognitive sophistication" which minimized overall bias. Of course, that might be an unstable point; anyone smarter than that tends to fall into a bias-blind-spot attractor and anyone less smart than that tends to fall into an object-level bias attractor.

It seems like a good opportunity to discuss what the article got right, what it didn't, how reddit reacted, and how one might better publicize the topic in the future.

Perhaps I'm being too pedantic here, but I dislike this way of framing our potential discussion of the article. The assumption seems to be that we at LW are authorities on cognitive bias, so obviously we couldn't learn anything from the article, but we could fact-check it. I don't know about you, but I'm pretty sure Jonah Lehrer knows a lot more than me about psychological research on cognitive bias. When I read one of his pieces, I'm not usually thinking, "Well, let me figure out what he got wrong and what he got right."

I apologize if I'm reading too much into your post. I guess one of the reasons I'm sensitive to this kind of thing is that there is a kind of collective intellectual arrogance in the LessWrong culture that I find very distasteful, and your phrasing in the post seems like a symptom.

Also, both your links go to the New Yorker article.

I don't know about you, but I'm pretty sure Jonah Lehrer knows a lot more than me about psychological research on cognitive bias.

He's a pretty knowledgeable dude, but he has major incentives to overstate research - the incentives which literally just days ago got him a job at The New Yorker, which is a pretty big plum for any journalist. When you read the original papers (I jailbroke the OP paper in question), you find he overstates things.

For example, this paper: the biases were specifically chosen to be resistant to intelligence, and the intelligent performed as well as the stupid - they were just wrong in predicting they'd perform better than the stupid (as they do in every other area of life). This is interesting if you're trying to formulate an integrated theory of IQ and the two processes, as Stanovich is doing (Stanovich's 2010 book is very good reading), but does it bear the spin Lehrer is putting on it? No.

I'm pretty distrustful of science reporting in general. I agree that LWers should at least act more humble (so people won't be averse to correcting them, and they won't be averse to correcting themselves).

I got the article from reddit, so I guess I'm used to articles about science being inadequate or misrepresentative in a major way. This community has a good number of members who know much more than I do about the topic so I wanted to see what they thought about it. Reading the article is probably sufficient to learn from it, I just wanted a better idea of the quality of the article.

Robin Hanson also posted on this study - his post contains a longer passage from the paper.

The New Yorker article is misleading - it makes it sound like smarter people are more biased, but that is not what the study found on the various particular biases that it measured, only on the meta-bias of the bias blind spot (and even that is iffy - more on that below). Previous research has generally found that smarter people are either less biased or equally biased, depending on the bias; I don't know of any cases where smarter people are more biased.

Did smarter people show a stronger bias blind spot in this study? The authors of the paper say (in the abstract) that the bias blind spot was not attenuated by higher cognitive ability, and "If anything, a larger bias blind spot was associated with higher cognitive ability." Presumably, "if anything" means that the association was weak and not consistently statistically significant across their various measures. Otherwise they'd just say that the bias blind spot was bigger, rather than not smaller, and if anything bigger. Similarly, in the passage that Robin quotes, they only highlight the novelty of their result by saying "the bias blind spot joins a small group of other effects such as myside bias and noncausal base-rate neglect in being unmitigated by increases in intelligence." I haven't read the paper yet (I can't find it online), it sounds like it only provides weak evidence for a positive correlation between cognitive ability and the bias blind spot.

After reading the paper (thanks for the link, VincentYu), it turns out that the evidence was stronger than I thought. The authors were being conservative to use the "if anything" language - I think more conservative than they needed to be. In study 1, their composite measure of the bias blind spot had a statistically significant positive correlation with each of their 4 intelligence-related measures (SAT, CRT, need for cognition, actively open-minded thinking), with correlations ranging from .096 (p<.05) to .260 (p<.001). It's a relatively weak correlation, but still pretty strong evidence that it's a real correlation rather than noise. For 4 of the 7 particular cognitive biases that they studied, the bias blind spot had a statistically significant correlation with SAT scores (see Jack's comment for more). In study 2, the correlation between the CRT and the bias blind spot composite was almost identical (r=.102, compared with .096 in study 1), but not quite statistically significant (p=.10) because of a smaller sample size. So I think it would be fair to state the claim more strongly than they do - smarter people do tend to think that they're less biased (at least within this study).

However, I don't agree with calling this a "larger bias blind spot." The bias blind spot is an instance of the better-than-average effect. It was measured by describing a specific bias to university students, asking them how susceptible to that bias they think that they are, and asking them how susceptible to that bias they think that other students at their university. People tend to say that they are less susceptible to the bias than other people are. This cannot be true on average (since people must be average on average), so it is a bias.

But studying individual differences in the bias blind spot is tricky because any particular individual who claims to be less biased than average may actually be less biased (rather than being more metabiased about how biased they are). I'm not sure if any other research on the bias blind spot has looked at individual differences; the main focus has been to look at differences between biases (which biases do people freely admit to having, and which do they have the largest blind spot for?). The bias blind spot measures don't have any clear standard for accuracy, other than the fact that self and other must be equally biased on average as long as they're sampled from the same population.

When you're comparing people who have high SAT scores with people who have lower SAT scores (as this study did), you don't have that because you're drawing from two different populations, and people with higher SAT scores do tend to be less biased (on many particular cognitive biases, though not all of them). If you had a really good, thorough measure of how susceptible to cognitive biases people are (one which covered a wide range of different biases, and which included lots of questions so that it could reliably identify individual differences among the noise), then you would presumably find that smarter people tend to score as less biased. So it's no surprise that smarter people tend to think that they're less biased - that seems to reflect accurate self-knowledge (and maybe also accurate theories, e.g. "This seems like the kind of thing that smart people would be better at"). It may be the case that smart people have trouble distinguishing the biases that their intelligence helps them avoid from the biases where their intelligence doesn't help, but that's not the same thing as having a larger bias blind spot (i.e., a stronger better-than-average effect with respect to how biased you are).

This study didn't include a thorough test of how biased people were - it picked a few specific biases and had 1 question each testing how susceptible people were to that bias (2 for anchoring). And this study actually did find that smarter people were less biased - on 3 of the 9 questions (across the 2 studies) there was a statistically significant relationship between intelligence and the bias (those with higher SAT scores had less outcome bias, less myside bias, and less anchoring bias on one of the two anchoring questions). The authors downplayed those effects because they only accounted for a tiny percentage of the variance, but on a noisy (single-item) measure I'd only expect a variable like this to explain a small amount of variance (and the correlation between intelligence and the "bias blind spot" composite also had a small effect size). Interestingly, outcome bias and myside bias were two of the biases where SAT did not have a statistically significant relationship with their bias blind spot measure - so smarter people did seem to be less biased on the whole, and they thought that they would be less biased, but they didn't know which biases they would be less susceptible to.

Out of 28 correlations 24 were positive and 17 were statistically significant, six of the 7 biases tested showed the relationship (the one that didn't isn't really a cognitive bias). For the Bias Blind spot composite score SATs were significant at p<.001, N=482. The authors are conservative in their assessment of the results. To provide some context West and Stanovich have spent the last decade examining correlations between intelligence and cognitive biases.

Upvoted. Based on reading the abstract and Lehrer's article, I'd expect only a 50% chance or so that these results will replicate, and even less so that they will generalize to other biases.

Which results? The results of the paper, which found that cognitive sophistication did not attenuate the bias blind spot or the slight positive correlation between cognitive sophistication and bias blind spot that the journalist decided to focus his article on?

the slight positive correlation between cognitive sophistication and bias blind spot that the journalist decided to focus his article on

I read the article, and VincentYu is spot on regarding the content. Lazy and misleading reporting on science is one of the few things that make me angry; every time I see a science-y Jonah Lehrer or Malcolm Gladwell piece trending, I know it's going to be a facile and wrongheaded take on an interesting topic, that I'm going to have to talk people out of the misconception in the future, and that I want to beat my head against a wall.

I think this quote from the paper (see VincentYu's link) gives a good summary of the findings and their significance:

The cognitive primitiveness of some of the processes causing the bias blind spot might be consistent with the failure of intelligence to attenuate the bias. However, this cannot explain the (albeit modest) positive correlations of the bias blind spot with cognitive sophistication that we found (see Table 2). The most likely explanation of this finding would probably be what we might term the “justified rating” account. Adults with more cognitive ability are aware of their intellectual status and expect to outperform others on most cognitive tasks. Because these cognitive biases are presented to them as essentially cognitive tasks, they expect to outperform on them as well. However, these classic biases happen to be ones without associations with cognitive ability.

Your link to the reddit comments does not link to the reddit comments. :P

I thought i linked to the comments, but I can't find them any more. Searching reddit with the url or title worked before but it doesn't now, i wonder if it got deleted.

The positive correlation seems an over generalization of the true fact that intelligence is generally negatively correlated with cognitive bias. The bias exceptions to that generalization are where you get the positive correlation to the bias blind spot.

The ending gets into a bit of speculation that introspection is bad. The main purpose of this section seems to be curiosity stopping.

[This comment is no longer endorsed by its author]Reply

Can you elaborate on this? I don't see any curiosity stopping going on there.

"The more we attempt to know ourselves, the less we actually understand." Seemed like it was meant in that way, i could be reading into it too much

Ah - you're discussing the Lehrer article. Yeah, I can see how you got that.

The authors are discussing something called the "introspection illusion" -- see here for a very complete review: https://psych.princeton.edu/psychology/research/pronin/pubs/Pronin%202009%20Introspection%20Illusion.pdf

Basically, it's not talking about introspection as, say, Zen self-questioning, where you start with an assumption that you do not know yourself and therefore must question (and be curious). Instead, it's talking about the more-automatic process that most people go through where they think, "Oh, I believe myself to be above the effects of bias - and, therefore, I won't exhibit bias." Most people think of themselves as good, and, bias being bad, most people therefore don't think of themselves as having it. Even people who do question themselves and truly probe their own reactions don't do so 100% of the time; that would be pretty exhausting for just about anyone. I'll be the first to admit that I don't fully trust my biases, nor trust myself to always figure out when my biases are operating (though I do very well at the cognitive tasks - which I almost wholly attribute to having taken a formal logic class).

So: Lehrer may be squelching curiosity, especially curiosity about oneself, but I suspect that he may not fully understand the introspection illusion. I would suspect that West et al. would have a very different opinion - more towards being very skeptical of what answers you believe you receive from your introspection and towards constantly questioning yourself.