indicators were built to describe literatures and are now used to judge individuals.
The shift happened without any deliberate decision and has proved extraordinarily difficult to reverse.
A measures attention rather than quality, and the two correlate loosely and unevenly by field.
A methods paper accumulates citations that a theoretical contribution of equal importance never will.
Comparing across disciplines with an unadjusted number is therefore meaningless, and it is done constantly.
The journal compounds the error by attributing to an article the average of the journal that carried it.
Citation distributions are extremely , so the average describes almost no individual paper in the journal.
Using it to evaluate a person is a statistical error that has been explained repeatedly for thirty years.
University rankings such indicators with reputation surveys and produce a single position.
The assigned to each component are chosen by the publisher and are not derivable from any theory.
A small change in those reorders the table, which is itself sufficient evidence about the exercise.
Institutions nonetheless manage themselves towards the indicators, since students and ministries read the table.
Hiring concentrates on candidates with metrics that will improve the score, which narrows what is researched.
Long projects, replication and teaching all register weakly or not at all.
Reform declarations signed by thousands of institutions commit them to judging research on its content.
has been partial, because a committee reading forty applications needs a sortable number.
The indicator survives on administrative convenience rather than on intellectual defence.
Any replacement must therefore be cheaper to apply than reading the work, which is the condition that no reform has yet met.