In a widely cited, and provocatively titled, article in 1999, now approaching 1,400 citations on scholar google, Kruger and Dunning seemed to provide evidence that “difficulties in recognizing one’s own incompetence lead to inflated self-assessments”. In other words, the less skilled (that’s what the authors really meant to say, alas a provocative title tends to sell) were argued, on average, to be more unaware of the absolute and relative quality of their performance. In fact, the less people were skilled the more they seemed “miscalibrated”.
In an earlier article, Krajc and yours truly have provided in response a simple model and some exploratory computational exercises that suggested that the less skilled may simply face a more complicated signal extraction problem. Our argument hinged on the distribution of skills in the environments that were typically studied being highly asymmetric, often resembling J – shaped distributions. Simply put, it is easy for the A++ student to figure out where s/he stands but much more difficult for those ranked towards the bottom of the class.
In an article just published, Ryvkin, Krajc and yours truly provide evidence in favor of a conjecture formulated in the earlier article: that with fairly little feedback self-assessment biases can be overcome. There was certainly a distinct literature on calibration that suggested that much (e.g., Juslin, Winman, & Olsson 2000 or Koehler 1996 – see here and here). There was, however, also some evidence that suggested otherwise (reviewed in our just published article.)
We hence set out to study whether, and to what extent, the difficulties of the less skilled in recognizing their own incompetence (largely, overconfidence) can be reduced by feedback. We report the results of two studies, one in a natural setting of a two-months graduate orientation and screening semester (and there particularly the micro-economics course by instructors not part of the research team), and another in the same environment but using tasks and stimuli materials that were better under our control. We document initially the same strong miscalibration that Kruger and Dunning also documented but also show that over the course of the two months this miscalibration almost completely disappears with the notable exception of those at the very bottom of the skills distribution and there only for their relative self-assessment which might provide support for the conjecture of Krajc & Ortmann (JoEP 2008; for reference see above) or self-image (e.g., Koeszegi 2006).
So, are the less skilled doomed to be unaware? It seems, no, not really. Learning goes a long way. Which should make us happy. But, of course, it all depends (on the strength and type of feedback, for example.)