In a a recent post, "The Errant Pursuit of Quanitifcation," Lance Haun argues that we too often rely on numbers in decision-making, and that some things simply cannot be measured. I'm going to take a different tack and argue that, instead of numbers giving us a false sense of security, that we do not take quantification seriously enough.
Numerous research articles have repeatedly shown that using a blind algorithmic approach does better than human judgment in making decisions and predictions. Paul Meehl, in his 1954 book "Clinical versus Statistical Prediction," asked an excellent question: Are the predictions of human experts more reliable than the predictions of actuarial models? With his colleagues David Faust and Robyn Dawes, Meehl has explored over a hundred studies comparing statistical or computer formulas with human judgment in predicting the likelihood of to criminal recidivism, and in virtually all cases, statistical thinking was better than subject matter experts giving advice.
Why is this the case?
- Bounded rationality - as the organizational theorists March and Simon argue, humans are limited in their ability to process, interpret, and act on information. We are easily influenced by others, the order in which information is presented, recent experience, distractions, and how information is framed.
- Incomplete information - there are time constraints and information costs that limit our ability to make an optimum decision. Further, in this internet age, we are bombarded with information, and it becomes more and more difficult to distinguish between signal and noise (for example, the recent thwarted bombing on the NWA flight from Amsterdam to Detroit ).
- Prior Hypothesis Bias - We tend to stick to our beliefs even when presented with evidence those beliefs are wrong. For example, if I am a conservative, I am more likely to listen to what Fox News has to say than Rachel Maddow.
- Escalation of Commitment - we tend to commit more resources to a course of action even when it might be better to cut our losses and run. For example, a gambler at a casino might go to the ATM and withdraw more funds to try to win back his initial stake.