Narrative Planning: Balancing Plot and Character (1401.3841v1)

Abstract: Narrative, and in particular storytelling, is an important part of the human experience. Consequently, computational systems that can reason about narrative can be more effective communicators, entertainers, educators, and trainers. One of the central challenges in computational narrative reasoning is narrative generation, the automated creation of meaningful event sequences. There are many factors -- logical and aesthetic -- that contribute to the success of a narrative artifact. Central to this success is its understandability. We argue that the following two attributes of narratives are universal: (a) the logical causal progression of plot, and (b) character believability. Character believability is the perception by the audience that the actions performed by characters do not negatively impact the audiences suspension of disbelief. Specifically, characters must be perceived by the audience to be intentional agents. In this article, we explore the use of refinement search as a technique for solving the narrative generation problem -- to find a sound and believable sequence of character actions that transforms an initial world state into a world state in which goal propositions hold. We describe a novel refinement search planning algorithm -- the Intent-based Partial Order Causal Link (IPOCL) planner -- that, in addition to creating causally sound plot progression, reasons about character intentionality by identifying possible character goals that explain their actions and creating plan structures that explain why those characters commit to their goals. We present the results of an empirical evaluation that demonstrates that narrative plans generated by the IPOCL algorithm support audience comprehension of character intentions better than plans generated by conventional partial-order planners.

Narrative Planning: Balancing Plot and Character

The paper "Narrative Planning: Balancing Plot and Character" by Mark O. Riedl and R. Michael Young advances the field of narrative generation by introducing an algorithm designed to craft narratives that exhibit both logical causal progression and character intentionality, crucial attributes for human comprehension and engagement.

Central Concepts

The authors articulate the challenges inherent to computational narrative generation, emphasizing two universal attributes of effective narratives: logical causal progression and character believability. Logical progression ensures that events within a narrative adhere to the rules of the narrative world, while character believability hinges on characters being perceived as intentional agents whose actions align with plausible, internally consistent motivations.

To address these challenges, the authors introduce the Intent-based Partial Order Causal Link (IPOCL) planner, an innovative algorithm expanding traditional planning to accommodate narrative structure. This algorithm extends Partial Order Causal Link (POCL) planning techniques by introducing frames of commitment, which represent segments of characters' intentional action sequences aimed at achieving specific internal goals.

Technical Contributions

The IPOCL planner stands out through key advancements:

1. Integration of Character Intentionality: Conventional planning approaches do not adequately address the complexity of multiple characters with potentially conflicting or distinct intentions. IPOCL introduces mechanisms for character intention formation and motivation planning, ensuring that all character actions are intentional or are marked as happenings, thereby supporting believability.
2. Frames of Commitment: A novel data structure, these frames describe temporally ordered actions characters perform to fulfill their goals. They serve as a bridge between character motivations and the actions required to achieve them.
3. Extended Plan Representation: An enhancement of POCL with additional planning layers to check intentional actions and motivate characters, expanding the planner's capacity to generate narratives that preserve intentional coherence and logical progression.
4. Algorithm Complexity and Optimization: The paper discusses the complexity, highlighting a significant branching factor due to the exploration of character intentions. Though computationally expensive, the presented heuristics guide the planner effectively within the vast search space.

Evaluation

The authors empirically demonstrate the effectiveness of the IPOCL planner through comparative studies with conventional POCL-generated narratives. Participants exposed to IPOCL-generated narratives showed improved comprehension of character intentions, evidenced by their performance in question-answering tasks compared to those who experienced POCL-generated narratives. This evaluation underlines the model's capability in more accurately reflecting human cognitive processing related to storytelling.

Implications and Future Directions

The IPOCL planner opens several avenues for further research and development:

• Incorporation of Complex Character Dynamics: Future iterations could integrate richer psychological models like emotion and personality to deepen character believability.
• Enhancements in Narrative Variety: Current limitations regarding the inability to generate narratives with character failures or adversarial outcomes suggest potential areas for enhancement, potentially through integration with machine learning to dynamically adapt narrative structures.
• Cross-Domain Applications: While focused on narrative generation, the principles could apply to interactive media, enhancing engagement in training simulations, virtual environments, or adaptive learning systems.

The paper contributes substantially to the computational narrative domain, offering foundational insights and methodologies that pave the way for more nuanced and engaging narrative generation systems. The IPOCL planner, through its rigorous approach to resolving narrative causality and character intentionality, underscores a significant step forward in modeling human-like storytelling systems.