Rational beliefs are pillars of epistemology. But how does one arrive at rational beliefs given one’s evidence? How and why ought one update one’s beliefs on all the available evidence? This project aims to answer such questions by applying tools developed in formal – Bayesian – epistemology.
The paradigmatic approach in formal epistemology is subjective Bayesian epistemology. It is currently widely perceived to be the leading theory of rational degree of belief, it underpins the dominant account of decision making in science and beyond, as well as many of our statistical methods. Subjective Bayesian epistemology requires agents to make a subjective choice adopting beliefs at the start of their epistemic lives. The available consistent evidence is taken into account via conditionalisation yielding posterior beliefs. This updating recipe guarantees that agents make, in a particular technical sense, optimal decision, i.e., agents thusly provably maximise the utility they expect to obtain.
Subjective Bayesian epistemology faces unfortunately difficulties which decrease its appeal as an account of rational degree of belief. A first set of difficulties arises from the highly subjective choice of prior beliefs: due to this choice being subjective, there cannot be any guarantee that adopted beliefs are accurate; furthermore, there cannot be a guarantee that decisions taken on the basis of these beliefs are good in actual fact. A second set of difficulties arises from the recipe for updating beliefs: it requires the set of propositions of interest to be fixed for the entire time of enquiry and there is furthermore no principled way to undo belief updates and nor can inconsistent evidence
be handled systematically. Thirdly, the requirement to take all available evidence into account is based on a decision norm stipulating that rational agents have to be risk neutral.
Ideally, one would want an account of rational degree of belief which does not face these difficulties, i.e., it i) offers some guarantees that degrees of belief are accurate and endorses decisions which are good in the actual world, ii) offers a framework for updating beliefs on all kinds of evidence which is flexible enough to dynamically change the propositions of interest which also allows for undoing belief updates and handles inconsistent evidence and iii) requires the assessment of all available evidence based on a decision norm stipulating that rational agents are risk averse.
Objective Bayesian epistemology is emerging as an alternative epistemology challenging the supremacy of the subjective account. Objective Bayesians defend the view that the degree to which an agent ought to believe propositions is determined by the agent’s evidence. In particular, there is little room for subjectivity, hence the name: objective Bayesian epistemology.
The primary objective of this project is to establish objective Bayesian epistemology as a serious alternative to subjective Bayesian epistemology by showing that objective Bayesian epistemology satisfies all these desiderata thus overcoming said difficulties. This will be achieved by 1) showing that objective Bayesian epistemology compels rational agents to adopt degrees of belief based on all their evidence based on a risk averse decision norm, 2) developing a comprehensive flexible framework for evidence aggregation based on objective Bayesian principles which enables undoing belief updates and handles inconsistent evidence and 3) arguing that the foundations of evidential reasoning of objective Bayesian epistemology hence compare favourably to those of subjective Bayesian epistemology.