The assurance view of testimony

The assurance view holds that in telling someone that p, a speaker assumes responsibility for its truth, and that this gives the hearer a distinctive reason to believe it, one that is not merely evidence about the speaker’s reliability. Moran, Hinchman and Faulkner are its principal defenders. For Moran, the speaker’s assurance is a reason the hearer has only through recognising that it is freely offered; for Hinchman, telling is an invitation to trust; for Faulkner, the speaker invites affective trust, which presses a normative claim on its object, as opposed to the predictive trust one may place in any regular process.

The view’s attraction is that it explains why being told differs from finding evidence: a hearer who is lied to has been wronged, not merely misled, and can hold the speaker to account. Its main difficulty, pressed by Lackey among others, is epistemic. If the speaker is unreliable, her assurance seems to add nothing to the hearer’s warrant; if she is reliable, her reliability seems to do the work.

For my purposes what matters is what assurance entails. A source that is answerable for the truth of what it says is one for which that truth figures in why the assertion is made. Faulkner’s account therefore entails the Tracking Minimum, and it is one of the five accounts from which I derive it.

It also explains why the view does not admit language models as testifiers. Artefacts are proper objects of predictive trust, and on functional accounts they can be trustworthy in the sense of being disposed to fulfil the obligations their functions generate. Neither is the relation Faulkner makes warrant-grounding, and substituting either changes what is being explained rather than satisfying it. The point is not that machines are beneath trust. It is that the relation grounding testimonial warrant, on this view, is one whose existence entails a connection to truth these systems have not been shown to have.