Combined probabilistic and dichotomous modelling of scientific and quotidian epistemic processes

Abstract

Probabilistic and all-or-nothing beliefs are difficult to combine in formal epistemology, since propositions with probability 1 cannot be given up with Bayesian probability revision. This project develops new models in which empirical full beliefs are provisionally undoubted but doubtable and revisable.

In phase 1, the properties of a recently introduced model, using infinitesimal probabilities as a technical device to “remember” presently undoubted possibilities, will be explored. It will also be used to investigate mechanisms of learning and truth- approaching. In phase 2, a highly general model of combined probabilistic and all-or-nothing beliefs will be used to develop more specific models with different properties. In phase 3, multi-agent processes, in particular consensus formation, will be investigated, and relations to several central issues in the philosophy of science will be explored. In phase 4, the impact of values on epistemic change will be introduced in some of the models.

The project will result in at least 13 international journal articles and one book. The research will be performed by the applicant, but ad hoc co-operations in his wide international network are to be expected. This project will develop a new approach to one of the long-standing central problems of formal epistemology. It will also provide a new and more realistic formal modelling of scientific progress and of the mechanisms by which we can learn from experience and approach truth.