To seem popular, just add coauthors…
Delving into the bowels of citation literature for some academic chores and, well, you know, got a bit grumpy about some things.
Long ago, when there was no Web of Science and you had to go to thick paper volumes to find citing articles, academic advancement had as a principal quantitative component the number of papers, with perhaps a multiplier for perceived quality of journal. Better places ignored this in favor of directly evaluating the quality and impact of the work, but in many schools this was the main metric. With the development of manipulatable databases of citations, then counting citations was a preferred metric, and indeed it is superior to simply counting articles. With that have come fancier metrics like h-indices and g-indices and the like. So we are in the golden age of advancing the very best, no?
Here’s the problem. All of these reward people who get their names into publications with lots of other folks. Why? Because we all try to build on our own work, so we will tend to cite the stuff we were involved in before. So a paper with 10 coauthors is more likely to be cited than one where you are sole author if for no other reason than those other authors will cite it from time to time. You could in fact be in the et al on a bunch of papers with nearly no effort and accumulate some really impressive statistics (and yes, GG has several specific examples in mind). Does it mean you are a hot shit scientist? Well maybe, but more likely it means you are a master of manipulating the system.
But wait, you cry, just remove the self citations! Ah, so true, but there is the rub: it is almost impossible to do correctly. SCI’s tool (which has other problems) will only remove the self citations of the author under examination, not any other self-citations. It is almost impossible to fix this. The result is a bias against those who tend to work alone and in favor of those who work in groups. (There is frequently a double bias here: if you are in a large group, typically there are a multitude of papers from a large study, so you get more papers for less work, plus you get more citations. So this can go with the square of the number of coauthors).
Advocates of h-indices and g-indices like to argue these are less subject to such problems, but GG is not convinced–they are less sensitive only in the sense that the numbers aren’t so different because these metrics are more like roots or logs of citations than total counts. So, if you want to play the game the most aggressively possible, by all means recruit coauthors. But if you want to make science more stable and reward the very best, watch out for the self-citation inflation problem. Good luck.