Counting citations is a rubbish metric in general. It's supposed to be a proxy
for reserach quality but it's so easy to "game" (in the sense of optimising
for citation count, rather than research quality) that a high number of
citations doesn't mean anything.
Neither does a low number of citations. For example, my field is small and
kind of esoteric, so we don't get lots of citations either from the outside or
the inside (one of the most influential papers in the field has... 286
citations on Semantic Scholar; since 1995).
With a field as small as a couple hundred researchers it's also very easy to
give the appearance of a citation mill. Given that papers will focus on a very
specific subject in the purview of the field, it is inevitable that each
researcher who studies that specific subject will cite the same handful of
researchers' papers over and over again- and be herself cited by them, since
she's now publishing on the subject that interests them.
As to self-citations, like Ioannidis himself says there are legitimate
reasons, for instance, a PhD student publishing with her thesis advisor as a
co-author. The student will most probably be working on subjects that the
advisor has already published on and in fact will most likely be extending the
advisor's prior work. So the advisor's prior work will be cited in the
student's papers.
So I'm really not sure what we're learning in the general case by counting
citations, other than that a certain paper has a certain number of citations.
Neither does a low number of citations. For example, my field is small and kind of esoteric, so we don't get lots of citations either from the outside or the inside (one of the most influential papers in the field has... 286 citations on Semantic Scholar; since 1995).
With a field as small as a couple hundred researchers it's also very easy to give the appearance of a citation mill. Given that papers will focus on a very specific subject in the purview of the field, it is inevitable that each researcher who studies that specific subject will cite the same handful of researchers' papers over and over again- and be herself cited by them, since she's now publishing on the subject that interests them.
As to self-citations, like Ioannidis himself says there are legitimate reasons, for instance, a PhD student publishing with her thesis advisor as a co-author. The student will most probably be working on subjects that the advisor has already published on and in fact will most likely be extending the advisor's prior work. So the advisor's prior work will be cited in the student's papers.
So I'm really not sure what we're learning in the general case by counting citations, other than that a certain paper has a certain number of citations.