Black and white swans
Nassim Nicholas Taleb, author of “Fooled by Randomness” and the more popular but less technical “The Black Swan”, gets a great deal of attention and press/media coverage over the last 6 months, mainly because people think that he predicted the current financial disaster. I think he deserves all the attention but there are quite a few misperceptions that touch onto our marketing data field:
- He did not predict the current financial mess but he was prepared for it. And he probably made some money with it. But there is a huge difference between “predicting” and “being prepared”.
- He warns strongly against a blind arrogance and trust into predictive data models but he is still using them to attempt to understand the world around him. It’s a big difference between not using data models and being realistic about their limitations and shortcomings
- He is extremely weary about believing that the past holds the key to understanding our future. But he is still studying the past, less to predict the future but more to understand human behavior and market dynamics.
- He is a big Karl Popper fan. That’s why he is using data to falsify theories not to verify them. This is a big fundamental difference that is difficult for most marketers to grasp. A particular data set is helping us in two ways: Inspiring new hypothesis and falsifying existing ones. But it does not hinder any decisive action but it puts one’s decision into a realistic framework of probabilities and size of a particular outcome that we hope to achieve with our decision.
posted by Michael Fassnacht at 7:16 AM
3 Comments:
like your review a lot - short and sweet and to the point.
Very timely Michael. What's dangerous is the dumbing down of the metrics, when underlying complexities are masked by a single score, something usually on a scale of 1 to 10 to provide the magic answer.
March issue of Wired magazine has 2 great articles on this topic. One's "Formula for Disaster" which describes David Li's Gaussian copula function that masked the complexity of CDO's with a simple rating, which later became one of the key reason why so many worthless bonds made their way into triple-A rated securities. Here're a couple of interesting excerpts:
Using Li's copula approach,meant that rating agencies like Moody's -- or anybody wanting to model the risk of a tranche -- no longer needed to puzzle over the underlying securities. All they needed was that correlation number, and out would come a rating telling them how safe or risky the tranche was.
And:
It's impossible to sum up such a history in one correlation number, but CDOs were invariably sold on the premise that correlation was more of a constant than a variable.
And Taleb himself said:
Anything that relies on correlation is charlatanism/
Studying data to understand patterns at best results in establishing correlations, not necessarily causations. Dumbing the complexity of those correlations further down to create an easily digestable score, which is not so uncommon in the marketing world with its plethora of customer scores, often masks the truth. The magnitude of the problem then depends on how big of strategic decision/bet/investment does someone make based on this "intelligence".
Taleb may be a contrarian and a real fiscal conservative and risk-averse -- but he takes these analyses at face value, i.e they are product of a mathematical formula and have their assumptions and flaws. They may help you make better decisions, but you can't take them as gospel.
Thank you for sharing a nice article.
Congratulations on your posts, not go from your site successful.
Post a Comment
<< Home