HeadSpace: Incorporating Action Failure and Character Beliefs into Narrative Planning

Abstract

Previous work on story planning has lacked a knowledge representation for characters that attempt actions that fail because of the characters' misconceptions about the world state. This work describes HeadSpace, a state-space heuristic search planning system that generates stories that track and manipulate characters' beliefs about the story world. The planner produces story plans with actions that are attempted but fail. We show an example story plan that contains failed-action content that cannot be generated by typical planning-based approaches to story creation, and we provide an analytical evaluation that characterizes our planner's increased expressive range relative to other narrative planners addressing character belief and/or failed action execution.

The Story So Far on Narrative Planning

“…For instance, a great deal of recent work has modeled agents who obtain mistaken beliefs and act per them such that they fail. These failed actions, like the disparities of belief used to prompt them, can be used to create irony as in the SABRE planner (Shirvani, Ware, and Farrell 2017), support the illusion of theory of mind as in the IMPRACTICAL planner (Teutenberg and Porteous 2013), and build tension as in the HEADSPACE planner (Sanghrajka, Young, and Thorne 2022). Of these, HEADSPACE would be able to generate narratives like The choice of which phenomenon to model is the major point of divergence in the narrative planning community.…”

Evaluating Reader Comprehension of Plan-Based Stories Containing Failed Actions

“…Plans and the planning systems that produce them have been widely studied in artificial intelligence research, though they are typically used to describe specifications of real-world action sequences. The plan structures we employ are those produced by the HEADSPACE planner as described by Sanghrajka and their collaborators (Sanghrajka, Young, and Thorne 2022).…”

“…More detail on the HEADSPACE algorithm can be found in the work of Sanghrajka, Young and Thorne (Sanghrajka, Young, and Thorne 2022;Young 2017a). Examples of two HEADSPACE plans used in our experiments are shown in Figures 1 and 4, and the first of these is described in more detail below.…”

“…Examples like this highlight how mistaken beliefs and action failures are used intentionally by authors to build suspense or other dramatic effects by playing with the disparities of knowledge and ability between characters within the unfolding story world. Recent advances in automated story generation have demonstrated the ability to create plot sequences with failed actions (Sanghrajka, Young, and Thorne 2022;Young 2017b;Christensen, Nelson, and Cardona-Rivera 2020), but work characterizing the efficacy of those approaches has been preliminary.…”

“…In the discussion that follows, we briefly review a plan-based knowledge representation for characterizing plans with action failures based on mistaken character beliefs. This representation forms the basis for plans produced by the HEADSPACE planning algorithm (Sanghrajka, Young, and Thorne 2022;Young 2017b). We then describe an experimental evaluation of use of these plan data structures to create textual stories.…”