Responsibility in (Non-)Probabilistic STIT

Attributing responsibility correctly when multiple agents are
involved is important to both philosophy and artificial intelligence. STIT
logics offer a compelling starting point for such a project. However, op-
posing proposals on how to do it have been put forward. In this paper, we
investigate two such prominent proposals and argue that the apparent
conflict between the two can be dissolved by looking at them through
a probabilistic lens. To do so, we introduce a simple probabilistic STIT
logic and two probabilistic responsibility predicates based on different
background assumptions. We show that the (non-probabilistic) respon-
sibility predicates discussed correspond exactly to the non-probabilistic
fragment of the probabilistic ones. We show that we can generate coun-
terexamples to each of the non-probabilistic proposals and that ulti-
mately, such counterexamples cannot answer the question which one is
preferable over the other.