Carrot or stick?
If you’re running a lab, will you get more from people if you praise their successes or criticise their failures?
There’s a great anecdote in Daniel Kahneman‘s “Thinking, fast and slow”, in which he recounts a class he once gave to some Air Force instructors. One of the instructors becomes irate when Kahneman suggests that they should offer praise and encouragement to their cadets - “Every time I praise someone for doing well, they get lazy and mess it up on the next try“, he insists, “But if I chew the son of a bitch out for screwing up, you can bet your ass he gets it right the next time“.***
Kahneman, being the erudite soul that he is, immediately spots the fallacy behind the instructor’s logic. Rather than observing a fundamental truth about human nature, he is in fact observing a fundamental truth about statistics, namely the law of regression to the mean.
Regression to the mean is one of those things that seems so intuitively obvious that it’s amazing how often we fail to see it. Basically, it means that if your first observation/attempt is far from the average, your second go will tend to be closer to the average. And vice versa. A first observation close to the average is likely to be followed by one farther from it. Over time and repetition, the observations/measurements/attempt will coalesce around the mean value, because that’s…well, that’s what the mean value is.
Kahneman demonstrates this by getting the instructors to each throw two coins at a target on the floor. After the first round of throws, the distance from the target of each coin is measured. Sure enough, on the second round of throws the ones who were farther from the target have tended to improve, and those closer to the target have tended to do worse.
Consequently, the real explanation behind the pilot cadets’ performance reflects the complexity of the system in which they are operating. They are carrying out a task at such a high level, that luck and random factors will determine which side of excellent their performance falls. The cadets who do badly on their first try have been unlucky, and therefore likely to improve on their second go (regardless of whether they are yelled at or not); the ones who do well on their first try have been lucky, and are unlikely to replicate that same brush with perfection in the next round (regardless of whether or not they are praised).
Taken further, this explains one of those perversities of life. Statistically speaking, if we praise a colleague we are likely to be rewarded with a worse performance in their next attempt, and if we are nasty to them for falling short then they will probably improve. But this does not mean that nastiness is the means to motivating people - in fact, numerous studies have demonstrated the opposite. Creating a positive atmosphere (this need not be a mellow one!) brings the best out in people.
It’s something we would do well to remember when assessing labwork. Many experiment protocols, no matter how well controlled, have a stochastic element to their output. Immunoblots are about as regularised as they come, yet it’s impossible to avoid the occasional bad blot. With more complex protocols such as immunofluorescence or immuno-electron microscopy, there are so many variables involved that luck plays a large role in the quality of the results, whether we like it or not.
It’s something to bear in mind when trying to motivate subordinates. A bad result doesn’t mean someone isn’t trying, and a good result doesn’t mean that things will work like clockwork from that point on. We should focus on the positives, and encourage people at all times. It's regression to the mean that counts, not regressing to being mean.
***Not a verbatim quotation, just a Top Gun fan's recreation of how military flight instructors might talk
First published at Total Internal Reflection - HERE.