OCAD University Open Research Repository

Carmen Won’t Talk to Me: Facilitating Expressive Qualities in Games Through Natural Language Interfaces and AI NPCs

Yamamoto, Tamika (2024) Carmen Won’t Talk to Me: Facilitating Expressive Qualities in Games Through Natural Language Interfaces and AI NPCs. Masters thesis, OCAD University.

Item Type: Thesis
Creators: Yamamoto, Tamika

Given the broad popularization of AI creative tools in contemporary society, this thesis explores the use of natural language interfaces by game designers as a primary gameplay mechanic to facilitate the expressive goals of the underrepresented game-maker.

The creative contributions of underrepresented game-makers hold cultural significance,hence, this research strongly advocates for the democratization of game-making tools, a need further underscored by the emergence of AI development. Such democratization is crucial as it addresses the exploitation of their creative labor within AI development, which frequently disregards the languages and cultures of marginalized communities in language models.

Drawing from J.L. Austin’s Speech Act Theory and Suzanne Keen’s Theory of Narrative Empathy, the research aims to provide a framework for game-makers to engage players in meaningful conversations with AI Non-Player Characters (NPCs) using large language models.

The hypothesis suggests that language shapes how we perceive our relation to others, and aims to animate expressive qualities by holding players accountable for the words they use within the game. This project uses a Research-through-Design methodology which entails making a game, documenting the process and findings in a developer’s journal, and finally synthesizing results to share best practice guidelines with other game-makers. These guidelines offer suggestions for implementing a natural language interface that takes into consideration the AI language model, the game-maker, and the player. Additionally, they provide a framework for facilitating a game-maker’s expressive goals using the theories outlined in this project, and offer strategies for holding players accountable for their words by fostering meaningful actions in a game utilizing large language models.

The importance of this research lies in its aim to provide tools to underrepresented game-makers, via this document and publishing online, enabling them to leverage the affordances of large language models and allowing for the creation of their own unique stories.

Date: 8 May 2024
Uncontrolled Keywords: artificial intelligence, games, language models, narrative-driven games
Divisions: Graduate Studies > Digital Futures
Related URLs:
Date Deposited: 08 May 2024 17:57
Last Modified: 08 May 2024 17:57
URI: https://openresearch.ocadu.ca/id/eprint/4482

Actions (login required)

Edit View Edit View