What I Think about the Dunning & Kruger Effect
The Dunning & Kruger Effect (DKE) asks participants to assign numbers to their abilities or estimate a score they are likely to get, and then compares their estimate with the actual score. The effect is that the difference between the estimated and the actual score is larger for students who score poorly and very well than it is for mid-range scoring students. In the popular version of the effect, it is only mentioned that poorly scoring students have bad estimates of their actual scores. This is supposed to demonstrate that the worse you are, the less likely you are to realize how bad you really are.
There is no such thing as the Dunning & Kruger Effect. It is a statistical effect widely understood among researchers who use rating scales. It is simply not possible to make estimates with the accuracy demanded by the research. As a result, participants at extreme ends of the scale are less likely to be accurate. This is well-known among scientists who work with rating scales – that estimates at the extreme ends of the scale are less accurate. It’s called centrality and is considered a rater bias that occurs naturally. The verbal description of the DKE makes sense. I’d say it’s almost folklore among the educated. But the experimental demonstrations of it developed by K&D have little to do with this verbal description.
Let’s put it this way. We got 100 students and we give them a test. Their scores are distributed normally. Before the test, we ask them
1. What will your score be?
2. Will your score be in the bottom quarter, second quarter, third quarter or top quarter of the class?
3. Will your score be in the bottom third, second third, or the top third of the class?
4. Will your score be in the top half or the bottom half of the class?
5. Will you pass or fail?
In fact, students have no idea what their scores will be. It is rare that anyone can guess their actual score before they take the test, especially the very low scoring ones. Students who score say 20 or 30 on the test will give you an estimate for their score of 40 or 50. There seems to be a low estimate below which students do not estimate their scores as likely to be, no matter how badly they know they will do. But the thing is, they know they’re going to fail. They know they’re in the bottom of the class. The more accurately you ask them to estimate their score, the less correct they are. In my sample, students were 100% correct about whether they would pass or fail. Bad students know who they are. They just have trouble assigning numbers to their predictions. And a greater demand for accuracy makes this even more difficult.
There are many examples of social cognition that are widely cited outside of psychology, by economists and policy scientists and such, that are rarely or never mentioned in social psychology. KDE is one of them. Outside of Kruger, Dunning and their students no one in psychology does work on this. It is widely cited particularly by economists and business theorists.
For a more detailed description of the effect, have a look at this blog post
Also, this link will take you to a Google Scholar search for the DKE
Many of these papers are free.
For more research on centrality, see the links listed here,
A lot of these papers are quite technical. An accessible paper to introduce the idea of centrality, why it is a natural problem in rating and what it means to measure it is,
Saal, F. E., Downey, R. G., & Lahey, M. A. (1980). Rating the ratings: Assessing the psychometric quality of rating data. Psychological Bulletin, 88, 413-428.
You should be able to download this for free from the Internet.