Abstract

It is a seemingly innocuous fact that people learn from the testimony of authorities. Children learn from parents, students learn from teachers, and laypeople learn from experts. What makes this appearance a little less innocent, however, is that some of these same people would have believed sources of testimony that are not authoritative, e.g., unreliable peers and charlatans. Since such hearers of testimony could form as many false beliefs as true ones, they appear to be unreliable consumers of testimony. Therefore, we might be tempted to question whether such hearers really do learn from authorities on account of their apparent unreliability. The standard response to this conflict of intuitions (henceforth referred to as “Goldberg’s puzzle”) is to affirm the reliability of the consumers of testimony by excluding problematic speakers from the type of testimonial exchange in which the hearer is engaged. I will argue that the standard response is ad hoc because it does not provide a principled basis on which to identify the type of testimonial exchange in which the hearer is engaged. In order to improve the standard response, I will argue that conceiving of the relevant testimonial exchanges as achievements of joint agency requires holding fixed the practical identities of both hearer and speaker across the relevant worlds that are being used to evaluate the reliability of the testimonial exchange according to a standard possible worlds semantics for counterfactuals and that such a characterization of the modal space provides a principled basis for solving Goldberg’s puzzle.