Scientific citations as an overwhelmed communication channel

Peter Dorman writes:

This looks like it’s up your alley. I find it interesting, but I’m not convinced by what he says about innovation. Of course, the issue is empirical, but I suspect it may be different for different sciences. In the work you’ve flagged in the social sciences, people are fishing for citations by cranking out apparent anomalies: “interesting if true“. it might be different in more theoretical fields. Overall, this is the sort of stuff I was hoping science and technology studies would be generating when I first stumbled on it a few decades ago. Instead it got hijacked by critical theory, but that’s a topic for another day. P.S. I’ve wondered why, in more scientific fields, it’s common to reference assertions with lots and lots of cites, not just one or two. My impression is that it’s less pronounced in the social sciences; at least, I started noticing it when I dove into the natural science literature for my climate book. Is that true, and if it is, why?

The above link is to a post, “How citations ruined science,” by David Oks. Here’s the summary:

There is nothing inherent about AI that makes scientists use it to produce slop; the essential problem is the incentive structures. The tidal wave of AI slop that is now threatening to engulf the institutions of science is, I suspect, a symptom of the citation regime. It would make no sense for a scientist in the 1950s or 1920s, when scientific communities were defined by personal reputation and mutual acquaintance, to submit the type of slop that is being submitted in huge quantities to scientific journals. But it does make sense within incentive structures that heavily reward publishing and getting cited.

This seems accurate to me.

I googled David Oks and he appears to be involved in public relations in some way. I wonder what he thinks of “AI slop” in that context. I ask because in advertising and public relations, I think the usual goal is not to produce anything novel or even interesting but rather to get noticed. This is related to what is called “the attention economy” but it’s centuries old, right? You put up big posters in the town square or newspaper ads or TV ads or whatever.

Another way of putting it is: For any communication channel there is some balance, some optimal level of flow of material. If the channel is unused, what’s the point; if it’s used too much, it will be manipulated.

One example is Pubpeer, the site for commenting on published scientific papers. It’s my impression that Pubpeer “works” in the sense that the comments there are often of high quality. But Pubpeer doesn’t get used very much–one way of seeing this is to note that a controversial paper that has been cited hundreds of times might only have a few comments on Pubpeer, or none at all. On the other hand, if Pubpeer did get used all the time, then I have a horrible feeling it would end up like Yelp and be swamped by fake reviews–or, to put it more precisely, reviews written for instrumental purposes to boost or knock a paper without supplying any scientific argument.

Every once in a while you can have a space like our comment section here which facilitates open discussion, but, again, once enough people are using a communication channel, there are increasing incentives to manipulate it, and the bad actors move in.