Is the Notion of Rationality of Inference Normative or Descriptive?

Abstract

Defeasible inferences can be formalized and evaluated in many different ways. It is tempting to make a choice between them by using the argument “from people”. According to this line of thought, inferences should be described and evaluated in the same way as this would be done by people. In the article I identify the principles on which the argument from people is based, what its applications are, and where its main weaknesses lie. The central claim of the paper is that the argument should be used with extreme caution, because of its fundamental methodological shortcomings.