More carrot, less stick: how to make research assessments fairer