Friday, July 02, 2010

Why Competency Self Assessments are essentially flawed beyond redemption

I'm working on an evaluation to determine if a training programme for senior government managers makes a difference - In the competence level of the managers, and in the service delivery they are able to produce within their work context. We are tracking a wide evidence base about all of the participants, but the client is insistent that a competency self-assessment be included. We agreed, on the condition that this one piece of evidence will be used together with all of the other evidence we will be collecting throughout the study. The value that the competency self-assessment will add, is something we have debated in the team. The following entertaining post by Errol Morris, however, pretty much sums it all up:

David Dunning, a Cornell professor of social psychology... wondered whether it was possible to measure one’s self-assessed level of competence against something a little more objective — say, actual competence. Within weeks, he and his graduate student, Justin Kruger, had organized a program of research. Their paper, “Unskilled and Unaware of It: How Difficulties of Recognizing One’s Own Incompetence Lead to Inflated Self-assessments,” was published in 1999.

Dunning and Kruger argued in their paper, “When people are incompetent in the strategies they adopt to achieve success and satisfaction, they suffer a dual burden: Not only do they reach erroneous conclusions and make unfortunate choices, but their incompetence robs them of the ability to realize it. Instead, …they are left with the erroneous impression they are doing just fine.”
It became known as the Dunning-Kruger Effect — our incompetence masks our ability to recognize our incompetence.

There have been many psychological studies that tell us what we see and what we hear is shaped by our preferences, our wishes, our fears, our desires and so forth. We literally see the world the way we want to see it. But the Dunning-Kruger effect suggests that there is a problem beyond that. Even if you are just the most honest, impartial person that you could be, you would still have a problem — namely, when your knowledge or expertise is imperfect, you really don’t know it. Left to your own devices, you just don’t know it. We’re not very good at knowing what we don’t know

In logical reasoning, in parenting, in management, problem solving, the skills you use to produce the right answer are exactly the same skills you use to evaluate the answer

Thursday, April 08, 2010

Can a song cause someone's death?

A quantitative evaluator, a qualitative evaluator, and a normal person are waiting for a bus. The normal person suddenly shouts, "Watch out, the bus is out of control and heading right for us! We will surely be killed!"
Without looking up from his newspaper, the quantitative evaluator calmly responds, "That is an awfully strong causal claim you are making. There is anecdotal evidence to suggest that buses can kill people, but the research does not bear this out. People ride buses all the time and they are rarely killed by them. The correlation between riding buses and being killed by them is very nearly zero. I defy you to produce any credible evidence that buses pose a significant danger. It would really be an extraordinary thing if we were killed by a bus. I wouldn't worry."

Dismayed, the normal person starts gesticulating and shouting, "But there is a bus! A particular bus! That bus! And it is heading directly toward some particular people! Us! And I am quite certain that it will hit us, and if it hits us it will undoubtedly kill us!" At this point the qualitative evaluator, who was observing this exchange from a safe distance, interjects, "What exactly do you mean by bus? After all, we all construct our own understanding of that very fluid concept. For some, the bus is a mere machine, for others it is what connects them to their work, their school, the ones they love. I mean, have you ever sat down and really considered the bus-ness of it all? It is quite immense, I assure you. I hope I am not being too forward, but may I be a critical friend for just a moment? I don't think you've really thought this whole bus thing out. It would be a pity to go about pushing the sort of simple linear logic that connects something as conceptually complex as a bus to an outcome as one dimensional as death."

Very dismayed, the normal person runs away screaming, the bus collides with the quantitative and qualitative evaluators, and it kills both instantly...Very, very dismayed, the normal person begins pleading with a bystander, "I told them the bus would kill them. The bus did kill them. I feel awful. To which the bystander replies, "Tut tut, my good man. I am a statistician and I can tell you for a fact that with a sample size of 2 and no proper control group, how could we possibly conclude that it was the bus that did them in?"

* On that note, however, I reject all the talk that the singing of a song CAUSED the death of a man. It may have been a contributor, but I doubt it was either a necessary or sufficient circumstance for this specific death!