Dunning-Kruger effect (actual charts)
I wrote about the Dunning-Kruger effect and attached the chart it is more commonly associated with, with the note that I’m not sure where that particular chart originated.
Here’s what the actual charts looked like from the original paper:
The dark solid line, which was what percentile everyone estimated their own ability to be, seemed to be hovering around 60 to 70% — everyone thinks that they are above average when compared to their peers. This means that the people who were below average overestimated their abilities by a huge degree, while the high performers consistently underestimated their ability (to a lesser degree, just because the baseline was above average to begin with). This translates to everyone starting with the same mindset — “I know a little bit more than your average Joe/Jane.”.
It is worth noting, of course, that the study was carried out on undergraduates so it may not necessarily carry over to other populations (e.g. Subject matter experts with work experience, people without college degrees). In fact, the paper goes on to carry out an experiment to see if the perceived scores are improved when the participants are either shown the answers from other test participants, or when they are “trained” on the topic.
In the first case, when the participants are shown other test answers, the top performers were able to calibrate their own perceptions after seeing how other participants have performed (5 answers were shown and they were told it was representative) and adjusted their perception to closer match their percentile (~70% to ~80%). Meanwhile the bottom performers made minor adjustments to be more modest but not significantly, still placing themselves as slightly above average at 60%. This shows that bottom performers, not being able to discern good or bad answers, are still “ignorant” of their ignorance.
In the second case where training is given, both the top and bottom performers had a much better calibration of their perception of their own ability, implying that the training allows them to better understand their own abilities.
Some other notes from the paper:
- The Dunning-Kruger effect applies more in the realm of cognitive skills rather than physical skills (grammar vs basketball)
- There are some domains where competence is more than just knowledge (art appraisers vs artists)
- It also only valid when there is a minimal amount of competence (few people will say they are above average in Astrophysics).
My question is: does knowing these results make you worry you over-estimate your own abilities, or does it make it wonder if you are possibly under-estimating it?
