Thursday, February 16, 2006

The Scandal of Prediction (a.k.a We think we're purty smart)

Just listened to an interesting program on IT Conversations dealing with the problems of predictability and why the various models and systems of predictors fail by Nassim Nicholas Taleb entitled, The Scandal of Prediction.

What I gathered from the program is that the overall problem with accuracy in predictions and forecasting is systemic arrogance. We fail to acknowledge what we do not know and inflate the little we do know. We should not stop attempting to utilize our forecast models and systems , but let us do it with an honest look at the vast amount of knowledge we are yet to discover.

An interesting note is that Taleb refers to the concept of Yesterdays Weather with which agile developers should be familiar. Simply put, this concept is based on the understanding that there is a 70% chance that today’s weather patterns will be a repeat of yesterdays. Statistically, we know this but still we employ vast machinery to "predict" what the weather will be like in the future. Admittedly, there is more value in the use of models and formulas the further out one attempts to predict. In any event, I humbly predict (without any meteorological training or know how) a 70% chance that tomorrows weather is like today’s.

No comments: