Task Plan verbalizations with causal justifications

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review

96 Downloads (Pure)


To increase user trust in planning algorithms, users must be able to understand the output of the planner while getting some notion of the underlying reasons for the action selection.
The output of task planners have not been traditionally user-friendly, often consisting of sequences of parametrised actions or task networks, which may not be practical for lay and non-expert users who may find it easier to read natural language descriptions.
In this paper, we propose PlanVerb, a domain and planner-independent method for the verbalization of task plans based on semantic tagging of the actions and predicates. Our method can generate natural language descriptions of plans including explanations of causality between actions. The verbalized plans can be summarized by compressing the actions that act on the same parameters. We further extend the concept of verbalization space, previously applied to robot navigation, and apply it to planning to generate different kinds of plan descriptions depending on the needs or preferences of the user. Our method can deal with PDDL and RDDL domains, provided that they are tagged accordingly.
We evaluate our results with a user survey that shows that users can read our automatically generated plan descriptions, and are able to successfully answer questions about the plan.
We believe methods like the one we propose can be used to foster trust in planning algorithms in a wide range of domains and applications.
Original languageEnglish
Title of host publicationICAPS 2021 Workshop on Explainable AI Planning (XAIP)
Publication statusPublished - 6 Aug 2021


Dive into the research topics of 'Task Plan verbalizations with causal justifications'. Together they form a unique fingerprint.

Cite this