Alessio Paone
Addressing the Narrative Paradox: Leveraging Large Language Models to Enhance Users' Perception of Freedom in Interactive Narrative Systems.
Rel. Luca Cagliero. Politecnico di Torino, Corso di laurea magistrale in Data Science And Engineering, 2024
Abstract: |
Interactive Narrative Systems (INSs) have emerged as a powerful medium for creating engaging and immersive digital experiences, allowing users to influence a dramatic storyline through actions and decisions. However, giving users more freedom within these systems comes at the cost of authorial control, leading to the well-known Narrative Paradox, which makes these systems less suitable for specific contexts such as training and education. This thesis introduces a novel approach that leverages Large Language Models (LLMs) to address this paradox and enhance users’ experience with INSs. We developed a Dynamic Interactive Narrative System Integrating LLMs (DINSIL), which exploits this technology to generate contextually relevant and creative responses to users’ actions, expanding their possibilities for interaction while the narrative unfolds, making each experience specifically tailored to the player. We tackle the drawbacks of integrating LLMs by reducing hallucinations with truth- grounded information and maintaining narrative coherence through mechanisms that regulate and guide generated content to preserve predefined story arcs. This study compares different models, including a Static-INS and an LLM-Only approach, using a sample of 83 participants who interacted with the system for about 45 minutes. The results demonstrate the superiority of DINSIL across multiple metrics, achieving up to + 34% and +30% Likert scores on Fun and Agency while maintaining a superior balance between the latter and authorial control, ultimately leading to more engaging and dynamic narratives. This research demonstrates how LLMs can be used to enhance the users’ perceived sense of freedom within Interactive Narratives (INs) without altering the authors’ predefined story arcs. In this way, our system is suitable for both entertainment and educational purposes, leading to better training outcomes through increased user engagement. |
---|---|
Relators: | Luca Cagliero |
Academic year: | 2023/24 |
Publication type: | Electronic |
Number of Pages: | 92 |
Additional Information: | Tesi secretata. Fulltext non presente |
Subjects: | |
Corso di laurea: | Corso di laurea magistrale in Data Science And Engineering |
Classe di laurea: | New organization > Master science > LM-32 - COMPUTER SYSTEMS ENGINEERING |
Aziende collaboratrici: | UNSPECIFIED |
URI: | http://webthesis.biblio.polito.it/id/eprint/31848 |
Modify record (reserved for operators) |