The dissertation I wanted to do, and why I didn’t, and whether I could have

I’m currently going through A Beginner’s Guide to Irrational Behavior on Coursera. Last night I read Behavioral Economics: Reunifying Psychology and Economics as part of the course, which is an article discussing the relationship between economics and psychology and how behavioral economics is bringing them back together.

The article itself is interesting, but in particular the discussion about bounded rationality and, in general, the entire concept of behavioral economics, makes me realize that the reason I had so much trouble doing my dissertation (and doing it about what I actually wanted to do it about) was because I was writing for the wrong department. Or, perhaps worse, what I really wanted to be doing is in between departments.

The project I wanted to be working on was about minimal rationality for decision theory. The idea is that decision theory, as normally done, requires an awful lot of an agent: a highly detailed preference ranking for every possible state of the world, a probabilistic belief about every possible state of the world, as well as a probabilistic belief about every state of the world given every other state of the world (for example, how likely I think it is that there are space aliens on the moon given that we have been there and not seen them, but also how likely it is that there are space aliens on the moon given that I just ate a banana). These are, of course, unreasonable assumptions (surely our brain doesn’t really have all of this stuff in it), but normally decision theorists don’t care – they are interested in ideal agents, not actual agents. And insofar as they are interested in actual agents, there’s a story you can tell in which you project onto them these belief functions and desire functions based on their behavior.

But I do care about actual agents, and I wanted to see how stripped down we could make these assumptions and still get decision theory to work. In fact, the best case scenario would be a decision theory that could actually be used by somebody, and also one which accurately determined whether they had acted rationally or not, given their actual beliefs and preferences. So I was interested in figuring out how realistic we could actually make it – do we really need beliefs about every state of the world? And preferences? Or can we transform decision theory to only require some minimal set of beliefs?

In fact, it might even turn out that decision theory will work even if (as all real agents are) some of my beliefs are contradictory! My basic theory would have been that only beliefs and desires directly relevant to the decision would need to be consistent, and they would only need to be minimally ordered. It doesn’t actually matter which order I put the options I don’t want to do, as long as I can select the top one, for example.

There’s more to it, but that’s the basic thing I wanted to do my dissertation about. Unfortunately, my advisor told me that this would not be a very good topic, because basically nobody else was doing that kind of work, and so it would be inadvisable to my career (if nobody else is doing it, then it’s unlikely to be interesting to prospective universities when I’m applying for jobs). And so I ended up trying to write it on something I was a lot less interested in, but would be “better for my career”. The topic I ended up working on was decisions about information gathering, which was interesting, but not as interesting to me as my original topic. I eventually quit writing it in favor of being a programmer.

I have to wonder, though, if I was just in the wrong department. Even though philosophers would be uninterested in the topic I’d picked, from what I can tell economists (especially behavioral economists) would probably be very interested. I hadn’t even really considered that as an option, probably because I didn’t realize that what I was really working on was an economics paper in philosophical jargon. Possibly. In any case, I look forward to learning more in this field so I can see whether I can explore more of that passion I had had. I may not end up with a dissertation, but I might end up with some good material to write about anyway.