Skip to main content

Epistemic Utility Theory: Foundations and Applications

Final Report Summary - EUT (Epistemic Utility Theory: Foundations and Applications)

In this project, we aimed to apply the powerful tools of decision theory to provide novel arguments for the epistemic norms that we take to govern what it is rational to believe; and to discover new epistemic norms. We treat the possible epistemic states of an agent as if they were epistemic actions between which that agent must choose. And we consider how we should measure the purely epistemic utility of being in such a state. We then apply the general apparatus of decision theory to determine which epistemic states are rational in a given situation from a purely epistemic point of view; and how our epistemic states should evolve over time. This allows us, often for the first time, to give formal justifications of epistemic norms without appealing to pragmatic considerations that seem intuitively irrelevant to the norms in question. These formal arguments have the great advantage that their assumptions are made mathematically precise and their conclusions are deduced from their assumptions by means of a mathematical theorem. We call their study epistemic utility theory or accuracy-first epistemology.

The main achievements of this project are as follows:

- We have offered a novel characterisation of the measures of accuracy.

- We have offered accuracy-first justifications for a number of credal norms, such as Probabilism, Conditionalization and related norms, the Principal Principle and related norms, and Principle of Indifference and related norms.

- We have offered a thoroughly investigation into the problem of peer disagreement from an accuracy-first perspective.

- We have extended the remit of accuracy-first epistemology from individual epistemology to social epistemology, where we have justified the so-called linear pooling rule for aggregating probabilistic judgements.

- By responding to a number of objections that had been made against the accuracy-first framework, we have secured the foundations of that approach.

- We have begun to explore how the notion of probabilistic knowledge might be treated in the accuracy-first framework, and we have argued for particular consequences of doing so.

- We have considered the consequence of the accuracy-first approach for the question of how categorical beliefs relate to credences.