Attribution-based Explanations for Markov Decision Processes

Attribution techniques explain the outcome of an AI model by assigning a numerical score to its inputs. So far, these techniques have mainly focused on attributing importance to static input features at a single point in time, and thus fail to generalize to sequential decision-making settings. This paper fills this gap by introducing techniques to generate attribution-based explanations for Markov Decision Processes (MDPs). We give a formal characterization of what attributions should represent in MDPs, focusing on explanations that assign importance scores to both individual states and execution paths. We show how importance scores can be computed by leveraging techniques for strategy synthesis, enabling the efficient computation of these scores despite the non-determinism inherent in an MDP. We evaluate our approach on five case-studies, demonstrating its utility in providing interpretable insights into the logic of sequential decision-making agents.

Counterfactual Strategies for Markov Decision Processes

Counterfactuals are widely used in AI to explain how minimal changes to a model's input can lead to a different output. However, established methods for computing counterfactuals typically focus on one-step decision-making, and are not directly applicable to sequential decision-making tasks. This paper fills this gap by introducing counterfactual strategies for Markov Decision Processes (MDPs). During MDP execution, a strategy decides which of the enabled actions (with known probabilistic effects) to execute next. Given an initial strategy that reaches an undesired outcome with a probability above some limit, we identify minimal changes to the initial strategy to reduce that probability below the limit. We encode such counterfactual strategies as solutions to non-linear optimization problems, and further extend our encoding to synthesize diverse counterfactual strategies. We evaluate our approach on four real-world datasets and demonstrate its practical viability in sophisticated sequential decision-making tasks.

BedreFlyt: Improving Patient Flows through Hospital Wards with Digital Twins

Digital twins are emerging as a valuable tool for short-term decision-making as well as for long-term strategic planning across numerous domains, including process industry, energy, space, transport, and healthcare. This paper reports on our ongoing work on designing a digital twin to enhance resource planning, e.g., for the in-patient ward needs in hospitals. By leveraging executable formal models for system exploration, ontologies for knowledge representation and an SMT solver for constraint satisfiability, our approach aims to explore hypothetical "what-if" scenarios to improve strategic planning processes, as well as to solve concrete, short-term decision-making tasks. Our proposed solution uses the executable formal model to turn a stream of arriving patients, that need to be hospitalized, into a stream of optimization problems, e.g., capturing daily inpatient ward needs, that can be solved by SMT techniques. The knowledge base, which formalizes domain knowledge, is used to model the needed configuration in the digital twin, allowing the twin to support both short-term decision-making and long-term strategic planning by generating scenarios spanning average-case as well as worst-case resource needs, depending on the expected treatment of patients, as well as ranging over variations in available resources, e.g., bed distribution in different rooms. We illustrate our digital twin architecture by considering the problem of bed bay allocation in a hospital ward.