Wednesday, June 2, 2010

XP 2010 Keynote by Scott Page

Scott Page delivered an impressive talk entitled "Leveraging Diversity in Parallel: Perspective, Heuristics, and Oracles". The mechanics for his talk were excellent--he started with a personal story about his youth, drawing us in to the first point of his talk: how new perspectives can make hard problems easy. He then expanded upon it by showing how multiple heuristics can be more valuable predictors. He noted that this flavor of the wisdom of crowds only works when we have a sufficiently difficult problem and all contributors have a minimal competency to evaluate the problem--but at the same time they must have sufficiently different backgrounds and training that when they're working on a problem they see it from different perspectives. He also spoke about oracles--something used to evaluate whether we've got the right answer or not--and said that wisdom of crowds is most important when there are no oracles. When we have oracles, we learn faster by incorporating their feedback early and often. Thus, multiple perspectives on a team are part of the magic that makes a team greater than the sum of its parts. On a larger scale, the wisdom of crowds increases the intelligence of groups too, when everyone is at least competent enough to make an educated guess and they offer different perspectives on the problem. He noted that some companies are now using a swath of employees/algorithms to more accurately predict markets or elections or even Netflix recommendations. He also explained that experts are often horrible predictors of the future because they work predominantly with a small set of models. These models bias their processing of data, and make them poor "oracles". Scott provided mathematical models for all of his arguments, but I don't buy them. As he said himself, reducing people to one statistic (such as an IQ score) is overly limiting. I think the same can be said of equations as a measure of diversity in a team... it is overly limiting to see what effect that equation will have on the accuracy of predictions on the team. Maybe that wasn't his point--he simply wanted to use math to back up his arguments. In any case, I have to admit that he used a diversity of perspectives to tell his story, and I can appreciate that. It was a very captivating talk, and one that would be worth watching online.

No comments: