Action Selection for Interaction Management: Opportunities and Lessons for Automated Planning

The central problem in automated planning---action selection---is also a primary topic in the dialogue systems research community, however, the nature of research in that community is significantly different from that of planning, with a focus on end-to-end systems and user evaluations. In particular, numerous toolkits are available for developing speech-based dialogue systems that include not only a method for representing states and actions, but also a mechanism for reasoning and selecting the actions, often combined with a technical framework designed to simplify the task of creating end-to-end systems. We contrast this situation with that of automated planning, and argue that the dialogue systems community could benefit from some of the directions adopted by the planning community, and that there also exist opportunities and lessons for automated planning.