Whether we like it or not, various simplistic bibliometrics are increasingly being used by university administrators, funding agencies, and governments in a largely misguided attempt at optimization. As is well known, any kind of measurement of any part of the academic enterprise will cause some to aggressively “optimize” their score, sometimes through unethical means.

The h-index is one of the best-known and most widely used such metrics. Google Scholar makes it easy to compute, but also easy to “game” one’s h-index through merging articles. The authors want to understand both the difficulty, scope, and design space of such h-index manipulation.

After defining the problem quite precisely, including various rules for citation counting for merged papers, the authors prove a variety of hardness theorems relating to optimizing the h-index (basically: it is very hard; in some settings, even increasing it by one can be very hard). But then they also proceed to conduct an experiment: given a chosen set of artificial intelligence (AI) researchers, can they manipulate their h-index in the context of the various rule systems previously defined? The results obtained are varied and nuanced.

This juxtaposition of complex proofs of algorithmic hardness with experimental results did create a certain amount of cognitive dissonance. Each set of results was very well explained, but somehow the transition was not as well justified as it could have been. Nevertheless, they do belong together, and make the paper as a whole much more interesting.

I enjoyed reading this paper, but I am at a loss when it comes to identifying the target audience.