@joeposaurus

Ha!

Also good to know that the fact that I think that I fully get this argument doesn't make it any more likely that I'm an idiot.

@joeposaurus

That the test to measure effect was faulty doesn't mean the results are faulty ("broken clock is right twice a day"). DK is so easily observable in real life that it can't summed up as false, even if a test to measure it should prove to be badly executed.

@joeposaurus And even the graph from the original study pretty much says outright "people are bad at guessing their relative performance, but they tend to guess closer to their actual performance."

Which flies 100% against the way people invoke "Dunning-Kruger." It's just saying "lol this person is stupid" dressed up in phony intellectualism.

@joeposaurus sorry, but no. The author of this "debunk" is significantly confused.

His argument for autocorrelation completely fails to see that what the research looks for - and found - is not a correlation, but a lack of perfect correlation (and how divergent it is).

Their counterexample - best illustrated by figure 7 - is a good show of their confusion: yes, in a sample of people that have a completely uncorrelated notion of perceived vs actual performance, the DK effect is at full glory.

@joeposaurus It was an interesting read, but I have the impression that it fails to address what I thought was an essential point of the DK effect: i.e. that the spread between real and perceived ability was bigger for the first quartile than for the last quartile.

The spread is bigger on the real data, meanwhile it's exactly the same on the random data.

Am I missing something there? 🤔

@joeposaurus

#bookmark

@joeposaurus

/IroniemodusON

Blödsinn ich versteh von Mathe gar nix und weis aber genau das es den Dunning-Krüger Effekt gibt!

Erlebe ihn doch andauernd, jene die überzeugt davon sind das die Erde eine Scheibe ist plärren die ganze Zeit rum und wenn sie irgend wann merken das ihnen die anderen schon gar nicht mehr zu hören schweigen sie und Wählen die Hohlerdepartei.

DAS KANN KEIN ZUFALL SEIN!

@joeposaurus

/goes on with his Besserwissermode ..

And what about all those who jumped on the Dunning-Krueger-Effect train because they have no clue about Math, statistics and auto-correlation but are self proclaimed specialist who are mad about getting opposed?

They all proved the effect to be true!

you see!
you see!

@joeposaurus Seems to me that what the data says is that:
a) Everyone is bad at assesing their own skill level.

b) Everyone, with the exception of the most skillful quartile, tend to over estimate their skill level, where as if the data were totally random, then you'd expect the two lines to cross at the median point on the x-axis.

You can make the point that the original claim was that the least skillful were uniquely bad at assessing their own skill, there is still a real effect there.

@joeposaurus

> "The Dunning-Kruger Effect is Autocorrelation"

I might be wrong, but I think this analysis is based on a conceptual misunderstanding.

Not the fact that the two curves are correlated is interesting (can be replicated with random data) but the fact that they are not perfectly correlated is.

The problem of "too juicy to not be true" might apply (but the other way as intended).

This convinces me more:

https://andersource.dev/2022/04/19/dk-autocorrelation.html

(Thanks @jayalane )

Anyway: Need more thinking.