The perfect surprise: a new analysis in dynamic epistemic logic

Abstract

In this article, we present a new logical framework to think about surprise. This research does not just aim to better understand, model and predict human behaviour, but also attempts to provide tools for implementing artificial agents. Moreover, these artificial agents should then also be able to reap the same epistemic benefits from surprise as humans do. We start by discussing the dominant literature regarding propositional surprise and explore its shortcomings. These shortcomings are of both an empirical and a conceptual nature. Next, we propose a philosophical solution to the problems that ail these systems, based on the notion of issue of epistemic interest. Finally, we give a formal framework to think about surprise. More specifically, we develop a probabilistic dynamic epistemic logic (called $\mathcal{SURPRISE!}$) that succeeds at formalizing the relevant philosophical concepts. This will be done through an issue management system grounded in topology. As an added bonus, the additional expressive power allows us to capture a richer variety of scenarios, and it also enables a more careful analysis of said scenarios.