July 17, 2014
My lab recently had the pleasure of having visiting scholar Dr. Jonas Richiardi in to tell us about a novel methodology to link genetics to the oscillations of the default mode network (DMN), with the hope of trying to figure out just what the heck is driving that sucker. During the presentation I felt a sense of excitement about research that I really haven't experienced in a while. I'll let the forthcoming paper do the explaining for me, and indeed, this isn't what this post is about. Rather, the reaction to it is what I'd like to discuss here.
All in all, people were pretty impressed. However, one of the emeritus faculty in the room did not share the same sentiment. "This is outside my area [clinical psych], but I just don't know what the point of all this is..." or some such comment. That was fair, since the presentation moved at a very high speed, so if you didn't know your cell/molecular bio it would be hard to grasp the importance of the work. So, I explained that actually figuring out the molecular mechanisms behind the more reliable findings of functional brain imaging was a pretty big deal...not to mention the incredibly cool bioinformatics involved (the project used data from the Allen Institute as well as the IMAGEN consortium). This led to some eye rolling from the emeritus prof over the concept of bioformatics and big data in general, and a few people in the room joined in, so I couldn't just write it off as a passing form of tech bashing that happens to be hip and trendy among the elders of today (I keep trying to tell myself it's just a phase they're going through...).
So I thought about it some more, and I thought about the low-tech way of doing things as a behavioral researcher (which is still pretty much the norm): You read the lit, you use your intuition, and you try to make a statement on how you think people work. And that's when I realized that every decision a person makes is based on "big data" - all your experiences and the experiences you've learned about go into the theories that you build - and there's a lifetime of information there. Your brain/mind/whatever seems to be able to organize said data somehow enough for you to craft a coherent set of sentences about it, and it's not a huge stretch to think about this as the application of some kind of learning algorithm(s) and/or pattern recognition technique(s).
But what to do you do when you want to integrate information that is unobservable to you, like brain activity or molecular mechanisms? Or when there's just too much information for you to experience, like the wealth of human interactions that go on in social media? Well friends and foes, in those cases you're just going to have to supplement your own cognitive resources if you want to get anywhere. And right now the only game in town is bioinformatics and big data.
"But what about theory?", those rebellious elders might grumble, which is a good thing to grumble about, though the answer is simple enough: it's the way you analyze the data. So this imposes some limitations on what you can speculate about - you need to be able to tie your hypothesis to a particular analytic technique. Informing yourself about machine learning algorithms and other mathematical techniques and considering these methods while building your theories will allow you to present the community with a falsifiable hypothesis.
If that sounds hard, it should, because it probably is. But we need to face up to the fact that most of the low-hanging psychological fruit has been picked, and we've got what we can out of the analytic technique of cocaine-fueled beard-stroking. The current problems of psychology and neuroscience require more than words that sound good together - we've got to express our beliefs mechanistic or mathematical terms so that others can test our predictions. A useful exercise, then, might be to introspect about where your theory came from in the first place. What did you observe when coming up with it? What kind of a pattern was it that you noticed? Maybe there's a machine learning technique that's similar that you could try on a new set of data. Imposing such constraints would certainly be a challenge, but I doubt any of us went to science school because we thought it would be easy.
If you'd like to comment on this post, use the Disqus thingamabob below or holler at me on twitter.comments powered by Disqus