Preference Elicitation and Explanation in Iterative Planning

Preference Elicitation and Explanation in Iterative Planning

Lindsay Sanneman

Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence
Doctoral Consortium. Pages 6456-6457. https://doi.org/10.24963/ijcai.2019/917

Planning for complex scenarios, particularly in which large teams of humans with distributed expertise and varying preferences share a set of resources, poses a number of challenges including integrating distributed information and accounting for context-dependent preferences and constraints. We see three key pieces to solving the problem of introducing autonomous assistance through a mixed-initiative planning system in these scenarios: preference elicitation, integrating preferences into planning, and providing tailored explanations back to the humans in the loop. The process of preference elicitation, planning, and explanation can be integrated as an iterative process by which teams can efficiently converge on the ideal schedule. Linear Temporal Logic (LTL) is a common language, readily understandable by both planners and humans, that provides a natural link between the three components of the iterative planning problem, facilitating both elicitation of expressive preferences and intelligible explanations of the system's decision-making processes. Outputs of each of the preference elicitation, planning, and explanation pieces can be expressed as LTL specifications and used as inputs to each next step in the process. We propose to explore preference elicitation, planning, and explanation using LTL specifications and the integration of these pieces into an iterative process.
Keywords:
Machine Learning: Learning Preferences or Rankings
Planning and Scheduling: Planning and Scheduling
Machine Learning: Explainable Machine Learning
Knowledge Representation and Reasoning: Preference Modelling and Preference-Based Reasoning