I'm sure it could, but of course you can always anonymise data first. In any case it seems a bit rich for facebook to restrict access due to privacy concerns, in reality the issues is likely that reason is simply that academics can't pay as well as ad agencies.

Yes, there are substantial privacy implications from social media research. And they're particularly problematic on a platform where privacy controls restrict access to certain data/conversations to particular friends (and Facebook employees, of course).

Not everything is like this, though. For example, there's been next to no public data about how the news feed algorithm works and no way of tracking how ads are being targeted. Both of these are commercially sensitive (and it's clearly in Facebook's interest to keep its cards to its chest) but a royal pain compared with, say, newspaper research where you can see what is being served to whom along with which adverts on a day-by-day basis.

And there's essentially no public API even to get those things to which you do potentially have access - so there's no way of getting the membership of public groups without web scraping or recording by hand, AFAIK. And certainly the first of these is prohibited by TOS, which means many university ethics committees will reject the research even if the researcher does have the capability to get it.

It's basically a large black box. And, as a political scientist, black boxes with political salience concern me.

That's certainly something to be careful about. Although one would think that in at least some data about advertising could be made available without much risk - outside of specific targeting lists, how much personal data is gonna be in there?