The ever interesting Felix Salmon has a post today on hedgehogs versus foxes that I found quite interesting. In it, he makes a distinction between a careful and incremental approach to knowledge (traditional research) versus a fast and haphazard approach (blogging). I find this especially fascinating because I tend to be an incrementalist as a research (let's refine the current optima) whereas I am much more willing to discuss things for which I am not always an expert on a blog.
What struck me as most interesting is that a complete strategy for improving a field likely embraces both approaches. The bloggers (foxes) are racing around asking naive questions in the search for an undiscovered solution. The academic researchers (hedgehogs) are refining the known solutions to make them actually optimal. Both approaches are actually very useful, on their own terms.
The thing to keep in mind is that the fox approach is going to generate a lot of failure. When I do epidemiology methods research, I always keep in mind that just beating he standard practice in the field is very hard to do. Heck, for a lot of problems, just getting substantial improvement in estimation over a carefully thought out linear regression model (regardless of whether the outcome is continuous or not) is surprisingly challenging. It can certainly be done and for some classes of problems, like time to event, the improvement is dramatic. Even more difficult, the current tools have been carefully refined and well understood -- developing a superior alternative is hard (just look at the number of methods papers published every month in statistics in medicine).
But every once in a while, a new idea shows up that leads to dramatic improvements, one way or the other, over conventional approaches. I think bloggers do a lot of the grunt work of looking around and trying to see if there is a superior paradigm or approach to looking at the current problems. Then academics take the slow and difficult process of incremental improvement.
It's a nice way of looking at the relation between these two approaches.