Specialists Not So Special, Experts Not So Excellent

The world has always been data, but in the past few decades we’ve started to become better at collecting it (and by better, I just mean that the cups we’re dipping into the ocean are getting bigger and less cracked). Analyzing it? Well, that’s a whole different problem.

But predictive analytics is getting bigger and bigger every year. It’s almost turning into a type of modern-day soothsaying, in whose cloudy “science” is seen the key to predicting future trends and human behavior. Large corporations in particular are interested in the field, since they “serve” massive populations that they’d like to “serve” better.

But predictive analytics is a complex thing. Takes a level of specialist beyond mere mortals.

Except maybe not.

Here’s a surprising Slate interview with Jeremy Howard, president and chief scientist of Kaggle, an online predictive analytics competition. Kaggle basically crowdsources predictive challenges around datasets that are donated by researchers and companies. Winner gets money and regular migraines.

According to the interview, Kaggle says that the people who win these competitions aren’t the specialists. From the interview:

PA: These competitions deal with very specialized subjects. Do experts enter?JH: Oh yes. Every time a new competition comes out, the experts say: “We’ve built a whole industry around this. We know the answers.” And after a couple of weeks, they get blown out of the water.

PA: So who does well in the competitions?JH: People who can just see what the data is actually telling them without being distracted by industry assumptions or specialist knowledge.

Specialization has long been both a trend and a controversy in many other realms. Education is the big example, I think. So it’s interesting to hear about it in such a complex field that by its very nature seems to need the most cloistered of expertise. So if specialist knowledge isn’t the key to accurate predictive data analysis, what is? According to Howard:

The difference between the good participants and the bad is the information they feed to the algorithms. You have to decide what to abstract from the data.

Basically, the critical difference is critical thinking. But that doesn’t mean no expertise need apply, just that such a role should be more circumscribed. Again, according to Howard:

Some kinds of experts are required early on, for when you’re trying to work out what problem you’re trying to solve. The expertise you need is strategy expertise in answering these questions.

Photo credit: James Cridland, Flickr