A Goal-Based Model of Personality for Planning-Based Narrative Generation

Julio César Bahamón, Camille Barot, Robert Michael YoungAAAI 2015 · DOI

Introduction

In the field of automated narrative generation, AI planners have traditionally focused on creating logically consistent plots where characters act "intentionally" to reach specific goals. While these systems excel at maintaining causal coherence, the characters they produce often feel like interchangeable "drones." They lack the distinct behavioral nuances that define human personality. This paper addresses this limitation by introducing a computational model that treats personality as a core driver of story structure, expressed through deliberate choices among multiple alternatives.

The Problem: Characters as Plot Drones

Existing models for character personality frequently focus on surface-level attributes, such as emotional reactions or dialogue styles. However, these models rarely integrate personality into the fundamental decision-making process that shapes the story's trajectory. The authors argue that personality is best perceived by an audience when they witness a character face a meaningful choice. If a character has only one possible path to a goal, their actions are seen as plot necessities rather than reflections of their internal nature. To truly portray personality, a narrative generation system must create "narrative branching points" where multiple actions are possible and then select the one that aligns with the character's specific traits.

Proposed Methodology

The researchers build upon the Intentional Partial Order Causal Link (IPOCL) planning framework, a standard in narrative generation that tracks character goals. They extend this framework through several key innovations:

1. Personality Operationalization

The model utilizes the Five-Factor Model (FFM)—a widely accepted psychological taxonomy consisting of Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism—to define character traits. These traits serve as the parameters for character behavior.

2. Character Values vs. Planning Goals

A critical distinction is made between "Planning Goals" and "Character Values." While goals are specific states a character aims to reach to advance the plot, values are internal states that a character seeks to achieve or preserve throughout the entire narrative. This distinction allows characters to have an internal compass that guides how they navigate the requirements of the plot.

3. Alternative Possible Worlds (APW)

The planner identifies moments where multiple action schemas could satisfy a story requirement. It then explores "Alternative Possible Worlds"—different versions of the story resulting from these different choices. The system ranks these branches based on how consistently the effects of the chosen actions portray the character's assigned FFM traits.

4. Contrasting Choices

For an audience to attribute a personality trait to a character, they must be aware that the character had other options. The system specifically designs the story structure to make these "contrasting choices" evident, ensuring the audience can recognize that the character chose a specific path over less-consistent alternatives.

Key Insights and Results

The core contribution of this work is the shift from personality as a stylistic layer to personality as a structural choice. By modifying the selection mechanism of a narrative planner, the authors moved away from non-deterministic action selection toward a personality-guided selection process.

Key results include:

  • Algorithmic Consistency: The system can successfully rank potential story branches based on "personality fit," ensuring a coherent reflection of intended traits.
  • Structural Support for Expression: The model ensures the generated story structure itself supports character expression by forcing the inclusion of branching points that elicit predictable trait attributions from the audience.
  • Intentionality: By distinguishing between internal motivations (values) and external requirements (goals), the system creates characters that feel more autonomous and less like puppets of the plot.

Limitations and Future Directions

While the formal declarative model provides a robust framework for personality-based planning, several challenges remain. This research focuses on the theoretical mechanism and does not yet include a formal user study to verify if human audiences perceive the intended personalities as the system expects. Furthermore, as the number of characters and traits increases, the search space for Alternative Possible Worlds could lead to significant computational overhead. Finally, while the model handles action selection, integrating these choices with natural language generation—ensuring a character's speech matches their actions—remains a separate and necessary challenge for future work.

Summary updated

BibTeX
@inproceedings{bahamon2015goal,
  author = {Julio César Bahamón and Camille Barot and Robert Michael Young},
  title = {{A Goal-Based Model of Personality for Planning-Based Narrative Generation}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2015},
  doi = {10.1609/aaai.v29i1.9731},
}
Figure