Objective Bayesianism

Objective Bayesianism is the position within Bayesian statistics that holds prior probability distributions can be derived from the structure of the problem itself, without appealing to personal judgment or subjective belief. The programme was pioneered by Harold Jeffreys through the Jeffreys prior, and later developed by Edwin Jaynes through the maximum entropy principle and by José-Miguel Bernardo through reference priors. The central ambition is ambitious: to show that Bayesian updating is not merely consistent reasoning from arbitrary starting points, but reasoning from starting points that any rational agent would accept. Critics — especially subjective Bayesians like de Finetti and those in the subjective tradition — argue that the appearance of objectivity conceals choices about parameterization, model space, and the very definition of 'ignorance' that are no less arbitrary than declared subjective opinion. The debate is not about whether Bayesian methods work. It is about whether the word 'objective' names a genuine achievement or a rhetorical shield.