Developing Stit Models For Causal Models

Abstract

As artificial intelligence is becoming more influential, it has become desirable to incorporate formal accounts of responsibility in techniques relying on AI. Finding a relation between stit logic and causal models would therefore be a great development, since both systems can be used for defining different parts of responsibility. Causal models are a great tool for modelling the causation part of responsibility while stit logic can be used for modelling parts of responsibility that causal models cannot effectively represent. Few people have, however, studied the relation between causal models and stit logic. The primary goal of this project will be to see whether it is possible to interpret one formalization of responsibility in terms of the other. Causal models use conditional probability distributions and directed graphs to model causality among variables. They are widely used in many disciplines, including artificial intelligence and philosophy. Since causality is crucial for the formalization of responsibility, causal models can be used for this purpose. Little work has, however, been done on using causal models to formalize responsibility. Stit logic is a logic containing the “stit” operator. “Stit” is an acronym for “see to it that” and the corresponding operator is used to model the effect that an agent has on a specified variable in the future. This project intends to find out how stit logic is related to causation by interpreting stit logic in terms of causal models.