Criticism and Praise

Print Friendly, PDF & Email

drunkards walk

Do criticism and praise work to affect performance?

Leonard Mlodinow briefly addressed this topic in The Drunkard’s Walk: How Randomness Rules Our Lives. He explores the studies of Daniel Kahneman who was lecturing the Israeli air force flight instructors on behavior modification. Kahneman was trying to make the point that rewarding positive behavior works, but punishing mistakes does not.

One of the students called him out. He had praised people warmly for beautifully executed maneuvers and the next time they do worse. He screamed at people for badly executed maneuvers and they improve the next time. The other flight instructors agreed. But Kahneman’s research demonstrated that rewards worked better than punishment.

So what was going on?

Regression toward the mean. In a random series of events, an extraordinary event is most likely to followed by an ordinary one. Due purely to chance, it’s hard to have two extraordinary events in a row.

The fighter pilots have a certain level of ability. An extraordinarily good performance is most likely to be followed by an ordinary performance. So the praise would seem to fail to maintain the extraordinarily good performance. Similarly, an extremely bad performance is most likely to be followed by an ordinary performance, which in this case would be better than the bad performance. So the screaming criticism would seem to cause an improvement in performance.

So it appears that the criticism does some good and the praise does no good. What is really happening is a misconception of uncertainty and probabilities. The connection between actions and results is not as direct as we might think.

In compliance, we eschew lots of data. It’s good to step back every now and then to think about the implications of the data and the underlying assumptions.

Author: Doug Cornelius

You can find out more about Doug on the About Doug page

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.