REPRESENTING GAME DIALOGUE AS EXPRESSIONS IN FIRST-ORDER LOGIC
đź“ś Abstract
Despite advancements in graphics, physics, and artificial intelligence, modern video games are still lacking in believable dialogue generation. The more complex and interactive stories in modern games may allow the player to experience different paths in dialogue trees, but such trees are still required to be manually created by authors. Recently, there has been research on methods of creating emergent believable behaviour, but these are lacking true dialogue construction capabilities. Because the mapping of natural language to meaningful computational representations (logical forms) is a difficult problem, an important first step may be to develop a means of representing in-game dialogue as logical expressions. This thesis introduces and describes a system for representing dialogue as first-order logic predicates, demonstrates its equivalence with current dialogue authoring techniques, and shows how this representation is more dynamic and flexible.
✨ Summary
Overview
This master’s thesis proposes representing game dialogue as first-order-logic expressions rather than as arbitrary strings embedded in manually authored dialogue trees. The implementation uses Clojure, core.logic, and MiniKanren-style relational programming. The representation is intended to be computationally meaningful, unambiguous, and structurally manipulable.
The system combines three main mechanisms:
- Higher-order predicates: Dialogue utterances are represented as executable predicates, allowing agents to pose nested questions and reason about how other agents would respond. This is demonstrated with a Knights-and-Knaves scenario involving simple theory-of-mind reasoning.
- Polymorphic response functions: Message types are modeled using metadata, type hierarchies, and multimethod dispatch. This allows responses to depend on the conversational function of an utterance, such as greeting, offering, informing, or querying.
- Rules and modifiers: Agents can update internal state through assertions, retractions, and changes to stored facts. Rules connect state changes to message-based triggers, supporting emotion, memory, and social-behavior models.
Two scenarios evaluate the approach. In “Bobby and Sally,” autonomous agents use dialogue to influence emotional state, with Bobby selecting actions based on predicted effects of Sally’s rules. In “Lovecraft Country,” the system reproduces the behavior of an existing dialogue tree while also allowing dialogue variation based on agent attributes, memories, emotional thresholds, and additional agents. The thesis therefore argues that the proposed representation can subsume the expressive behavior of dialogue trees while enabling more dynamic reuse and recombination of authored data.
Documented influence
A concrete later research use appears in Modeling Autobiographical Memory for Believable Agents, which cites Wheeler’s thesis when discussing robust memory as a support system for believable agent behavior and identifies integration with Wheeler’s dialogue system as a potential direction for more dynamic interactions. (ojs.aaai.org)
The thesis is also listed in the miniKanren research community’s catalog of work related to the ecosystem, but the sources searched do not establish a broad industry adoption or a large subsequent research lineage directly attributable to it. (minikanren.org)