For starters, my final point in the last post (we don’t know what research will wind up fruitful until after we do the research) proves too much. Surely it makes sense to dismiss devotion to topics that seem obviously useless as irrational. We shouldn’t, for instance, consider it legit to defend an astrology nerd on the grounds that astrology might wind up valuable someday. From what we know now, astrology seems so unlikely to produce legit gains that’s it’s silly to devote much time at all to it.
This points to a solution. I feel it makes sense to evaluate someone’s commitment to a topic based on what we're justified in believing now about that topic. The more an issue seems promising and important to us now, the more reasonable it is to nerd out over it. Conversely, the more a topic seems useless, the less reasonable it is to obsess over it, even if it winds up actually being important. So, some academics are wholly justified in their pursuits, while others, not so much.
Where is the line between rational and irrational specialization? How permissive should we be? This depends, I think, on two factors. First, how large is the gap between what seems important and what actually is important? The more difficult it is to determine importance, the more permissive we should be.
Second, how much should we value contrarianism for its own sake? Presumably, there’s a danger to too much conformity, both in what society at large deems important, and even in what reliable experts deem important. As long as it isn’t just rationalization, it seems reasonable for a self-aware academic to appeal to such meta-considerations to justify her weird obsession.
I suppose my personal preference is to be fairly permissive—I bristle at academic turf wars. Mainly this is because I feel the need to be fairly confident to declare a line of inquiry as fruitless, and I simply lack this confidence when assessing most research. But there's also the worry that I'll destroy my sweater: I (perhaps naïvely) see most fields as roughly on equal footing when it comes to potential usefulness, so if I call out one specialization, I fear I'll wind up calling most of them irrational.
And that's just silly.