The variance in productivity between people is much greater in some areas than in others. The most productive cleaner is probably less than twice as productive as the average cleaner, whereas the most productive entrepreneur arguably is many times more productive than the average entrepreneur. One area where this difference in productivity is particularly great is research - especially in its more theoretical parts. A physicist once told me that Einstein put forth four theories worthy of the Nobel Prize in one year alone (1905, his Annus Mirabilis). That's a level of productivity that, on any reasonable measure, dwarfs that of whole armies of lesser physicists. The same pattern recurs in many disciplines (though it is perhaps not always quite as strong).
Such a recurring pattern cries out for an explanation. In some areas the reason might be hero-worship (postmodernist philosophy springs to mind), but the pattern is strong in many serious disciplines as well, so that can't be the whole explanation.
Instead, I would argue that the reason is that science is a social activity in a much deeper sense than cleaning is. A scientific problem is "hard" due to the fact that most scientists are not sufficiently talented to solve it given our present knowledge. In most sciences, there will normally at any given point be a lot of such problems which are too hard for most scientists, but solvable by a few really smart people.
I'm not saying that but for them, there would be no progress. Clearly there would, but we would need more knowledge - clues, as it were - to solve the hard problems. Thus progress would go much slower.
If this picture - rather an elitist one - is right, what are the consequences for policy? It seems to me that the most obvious consequence is that we need to make sure that the most talented people's time is spent in a rational way. Firstly, it does not seem to me that they should spend time teaching students - their time is far too valuable for that. Overall, it's a bit strange that researchers are expected to teach since they really are two quite different jobs.
That's not my main point, however. My main point is that they should focus on what what they're best at: producing new ideas. Research today involves so much more - applying for grants, writing up articles that have to conform to all sorts of disciplinary conventions, attending seminars of questionable value, etc, etc.
It is widely noted that the present incentive structure ("publish or perish") in the academia has led to a flood of uninteresting but publishable articles. A less-noted but to my mind more serious problem is that this incentive structure fails to make the most talented people create and (in particular) publish ideas at the pace they're capable of. I know several very talented people whose publication rates are very low even though they are positively brim full of interesting ideas.
I don't know exactly why they don't publish more but suspect that they're bored by the tedium of the peer review process (who isn't - I have two papers that haven't been reviewed after 1,5 years). Perfectionism is also to blame, I think. Many smart people are fed up with the massive amount of mediocre research that is produced, and therefore (and perhaps for other reasons) they swear to themselves never to publish anything that isn't top-notch and very well worked through. Though I share the annoyance with the massive amount of mediocre research, I think that this reaction is misguided. It would be much better for the research community as a whole if they shared their ideas, even if some of them are a bit half-baked.
Hence we need to devise a system that makes them produce more publically available research. Reading this blog - and in particular Yudkowsky's excellent posts - it strikes me that they should have a blog where they could write down various ideas. They would need some incentive in order to do that, and hence the academia needs to start recognizing the blog as a proper medium of publication (as is actually happening in some disciplines such as macro economics, if Krugman is right).
Indeed, I think that science would usually be better off if these people left the job of proving their hypotheses, or that of making them more precise, to other people, and hurried on to the next subject. We need a much more thorough division of cognitive labour in the academia: we should not waste the most able people's time (our most precious resource) by having them work on tasks that could be left to less able people. (This principle is explicitly followed in many other organizations - e.g. it is frequently argued that various tasks done by doctors could be done by less educated, and cheaper, employees such as nurses.)
Of course you do have a fair amount of this in the academia. Professors have Ph.D. students and postdocs to do carry out simpler parts of research for them. Academic stars do tend to work on the "big problems" and leave to others to fill in the blanks. By and large, this is good. But we need to go much further in this direction.
Apart from such obvious problems such as vested interests' resistance to change and sheer inertia (ever-present problems in all institutions), two things obstruct a change in this direction. The first is that people are poor at estimating researchers' skills. They don't recognize that the differences in skills are as large as I have argued here. Different professors earn roughly as much, have roughly the same social standing, etc - how could it possibly be that their productivity levels are vastly different, the seem to think. Also, there is the Dunning-Kruger effect, which says that incompetent people overestimate their skills, whereas competent people underestimate theirs. This contributes to the impression that productivity levels are more equal than they actually are.
Also, people are not very good at determining who's skilled and who's not. (This goes in particular for incompetent people.) In many cases, they go for superficial characteristics such as whether the researcher uses some respected technique (whether this technique is actually mastered, or even relevant in the case at hand, is often not regarded). To some extent, this is an unescapable problem, but possibly it could be alleviated somewhat by more sophisticated techniques for identifying good texts. Such simple techniques as explicitly counting the number and significance of ideas per page, could go some way towards making it more transparent who is actually an original thinker, and who is not. More advanced techniques could conceivably go much further. This is a topic rationalists should explore, I think.
Another problem is egalitarianism. The system I am sketching is in many ways more hierarchical than today's academia - some people would focus entirely on the hardest and most prestigious tasks, whereas others would have to do more mundane work. I also think that this could potentially be reflected in greater differences in pay (if nothing else, universities would be more reluctant to waste talented people's time if they had to pay them huge salaries). Hence people would oppose it for egalitarian reasons (most people would probably deny that there are huge differences in ability because it clashes with their egalitarian morality, but I think that some people who admit that there are such big difference in ability would oppose it nevertheless, for egalitarian reasons).
If you're not an egalitarian, this is obviously not a valid objection, but even if you are, I think that it has less force than most people would think. What is important for egalitarianism is that society as a whole is egalitarian - i.e. that the resources (including not only money but also, e.g. respect) are not too unevenly distributed - not that any particular profession is egalitarian. Why should it be more objectionable that productive professors earn five times more than an unproductive professor, than that professors earn five times more than cleaners? Now if a less egalitarian academia makes the Gini coefficient increase somewhat, this could easily be offset by introducing a bit more progressive taxes.
I leave the details of how to set up this system at the moment for now even though I think it's very important to pay great attention to the institutional arrangements of the academia, since researchers are self-interested creatures who react to incentives just like everybody else. If anyone has any ideas on this, I'd be interested in that.
I've got a lot more to say about this but this post is too long already so I leave it at that.