Monday, May 27, 2019

Dunning-Kruger by Proxy

I don't generally go around citing a lot of biz-school research except when it goes along with making fun of David Brooks, but there's some very interesting stuff going on in the conceptual orbit around the so-called Dunning-Kruger effect—the way incompetent people lack the skill to recognize their incompetence and tend to believe they're particularly good at the stuff they can't do, in spite of the obvious evidence. And there's an elephant in the room where the discussion is taking place, or shall we say a 400-pound guy on a sofa, that nobody in the research world wants to talk about, but I of course do.

The main findings of the paper I'm looking at, "The Social Advantage of Miscalibrated Individuals: The Relationship Between Social Class and Overconfidence and Its Implications for Class-Based Inequality" by Peter Belmi, Margaret A. Neale, David Reiff, and Rosemary Ulfe, which arrived last week in Journal of Personality and Social Psychology: Interpersonal Relations and Group Processes (May 20, 2019), are two:
  • that Dunning-Kruger overconfidence actually confers an advantage in interpersonal relations on the subject, because self-confidence is charismatic, and interlocutors commonly misinterpret it as competence and give the subject the nod, or the trust, or the job; and
  • that relatively high social class reinforces overconfidence in those who have it
which offers a quick explanation for why social immobility can remain so pervasive (enough members of the upper class are brought up confident that they deserve the job and that confidence convinces the interviewer that they do), and why the worst bosses you've ever had, almost always men, have all believed they knew how to do the job better than you did, even though they were totally and often catastrophically wrong.

First, the authors were able to work with an enormous sample of small-business owners in Mexico, over 150,000 people applying over the period 2015-17 for microcredit from the international firm Reiff and Ulfe work for, reported how high or low they thought their status was and performed a flashcard test as part of the application procedure, and were then asked to assess how well they had done on the test. It turned out that those who believed they had done much better on the test than they actually had tended very strongly to be those who reported themselves as high in class. Whereas they were not necessarily those who were high in class by objective measures (educational attainments and wealth)—by one scoring method they were, but another scoring method found that the objectively high-class subjects tended (less strongly) to have assessed their performance on the flashcard test more accurately.

That is, the study found that your belief that you are high-class, accurate or not, is correlated to your inaccurate faith in your competence, but your actual class status may or not be.

A second study of 500 or so (very cheaply) paid subjects who answered an ad in Amazon's Mechanical Turk replicated the results of the first with a standard test of cognitive ability and a more elaborated set of class variables, in the strong form—reported and objective class both correlated with misplaced confidence—and a possibly interesting wrinkle: there was one objective class variable that didn't correlate significantly, that of parental education. It also found that the desire for higher class status, which was more prevalent among those who were relatively high class, also correlated with misplaced confidence. A third study of over 1000 non-students obtained the same way using a "fun trivia game" as the test again found a very robust correlation between social class (self-described and objectively determined) and overconfidence.

Finally, the same trivia test and confidence assessment were administered to 200-some university students and then they were brought back a week later for a mock job interview and speech on "how they would handle a difficult situation, and these were videotaped. Afterwards, 900-odd independent judges recruited on MTurk were brought in to assess the "candidates" for competence, morality, and warmth. Again, subjects from higher social classes were significantly more overconfident than subjects from lower classes, and higher overconfidence went with higher ratings for competence from the judges, as the hypothesis predicted (though the direct association between high class and perceived competence wasn't statistically significant):

Being from the upper class tends to make you overconfident, overconfidence makes people expect you to be competent, and people judging you competent makes them want to give you a job; as Peter Belmi told Microsoft News,
Dr Belmi said overconfidence in people of higher class could be partially due to differences in values between the middle and working classes.
"In the middle class, people are socialized to differentiate themselves from others, to express what they think and feel and to confidently express their ideas and opinions, even when they lack accurate knowledge," he said.
"By contrast, working-class people are socialized to embrace the values of humility, authenticity and knowing your place in the hierarchy.
"These findings challenge the widely held belief that everybody thinks they are better than the average."
Which is a pretty precise model of what Marxists refer to as the "reproduction of class relations", though not the model Marx had in mind, if he had any particular model in mind at all, and a very precise account of the ordinary working person's experience of people who don't know what they're doing getting promoted to the top, and of Laurence Peter's principle that people in a hierarchical organization tend to get promoted to their personal "level of incompetence", an aspect that the Belmi team seems not to have fully understood, that a system like this doesn't merely freeze social mobility, which would be bad enough, but also fills the upper ranks of society with people who don't know what they're doing, a deficit that actually makes them look to their immediate superiors as if they did.

And an excellent sketch, this is the part nobody's talking about, of how it's possible that some people think Donald Trump is competent: he's like a cartoon version of the hypothesis, a scion of inordinate wealth who develops a self-presentation so confident as to seem insane when you think about it (in fact it is insane, but that's another matter), masking a total incompetence in any matter he touches if you know anything about it, and lives off it for decades because the confidence charisma convinces enough people in the right places that he really is capable (many of them incompetent too), until he's finally all the way at the top of the hierarchy and we're scratching our heads in wonderment.

It's a confidence game in every sense of the word, and a kind of Dunning-Kruger syndrome by proxy, where a huge swathe of the population has come to believe in this man's skills simply because he seems so certain he has them, though it's obvious from the things he says, if you know anything about then (as congressional Republicans or the Washington press corps do, though they prefer not to talk about it), that he doesn't.

No comments:

Post a Comment