Yeah, the casual suggestion that MIT should try to badly damage the careers of most of its faculty (presumably causing them to leave) suggests that he hasn't thought about this problem very long or talked about it with many people. As much as I dislike academic publishing, the last thing I want is to have deans and their staff evaluate the merits of research on a case by case basis. It sounds nice in theory, I guess, but they're people with agendas and biases too, and even setting that aside, they don't have the time or the training to do it. (as an, I hope, neutral example: could a dean trained as an economist really evaluate the relative merits of 15 different experimental physics papers? Could a trained physicist really evaluate the contributions of 15 different economics papers?) (note that this all applies to deans at other universities. The deans at my university are of course exceptional and above reproach).

So, on the one hand, even MIT probably can't unilaterally make a lot of progress on this issue. On the other hand... almost every recent paper I've ever been interested in is available from the author's webpage (making this quote: "much of the work produced in academia is never seen by anyone outside that cloistered world, because everyone who’s not affiliated with a university is cut off from access" ring hollow to me), all of the work to produce the journals is done for free or nearly free by researchers, and (anecdotally) everyone prefers electronic copies of the journal articles anyway. It seems like open access publication is almost inevitable.

A more manageable (but less sexy) step might be for MIT and other universities to defray the admin costs for some of these flagship journals that want to go open-access. The article mentions the American Historical Review as having annual (editorial) costs of about $500k [1], which is tiny for a university.

[1] From this post: http://blog.historians.org/news/1734/aha-statement-on-schola...