Explore open access research and scholarly works from STORE - University of Staffordshire Online Repository

Advanced Search

Developing Narrative Design in Role-Playing Games with Natural Language Interaction for Non-Player Characters

Haslett, Luke (2026) Developing Narrative Design in Role-Playing Games with Natural Language Interaction for Non-Player Characters. Doctoral thesis, University of Staffordshire.

[thumbnail of A thesis submitted in fulfilment of the requirement of the University of Staffordshire for the degree of Doctor of Philosophy]
Preview
Text (A thesis submitted in fulfilment of the requirement of the University of Staffordshire for the degree of Doctor of Philosophy)
Luke Haslett Thesis.pdf - Submitted Version
Available under License Type All Rights Reserved.

Download (10MB) | Preview
[thumbnail of EThOS Agreement] Text (EThOS Agreement)
EThOS-Deposit-Agreement (Luke Haslett).docx - Other
Restricted to Repository staff only
Available under License Type All Rights Reserved.

Download (63kB) | Request a copy

Abstract or description

This thesis investigates how natural language interaction (NLI) systems can support narrative design in role-playing games (RPGs), with a focus on player experiences of non-player character (NPC) interaction. While dialogue systems in games commonly rely on pre-scripted responses and predetermined dialogue options, contemporary natural language technologies provide opportunities for players to communicate with NPCs using free-form written or spoken input. This research examines how such interaction influences player perceptions of enjoyment, conversational freedom, narrative agency, realism, immersion, intuitiveness, and engagement.

The study was guided by four research questions and adopted a mixed-methods approach comprising a literature review, pilot study, prototype development, and main experimental evaluation. An original RPG prototype, Turing, was developed using Unreal Engine 4.27 and integrated OpenAI's GPT-4 for generated NPC dialogue alongside Google Cloud Speech-to-Text and Text-to-Speech services. Twenty-two participants interacted with the prototype during controlled playtesting sessions before completing a post-play questionnaire. Structured responses were analysed using frequencies and percentages alongside selected exact inferential tests, while open-text responses were examined using reflexive thematic analysis.

The findings indicate generally positive participant responses to NLI-driven NPC interaction. Conversational interaction was associated with enjoyment, exploration, worldbuilding, characterisation, narrative discovery, personal expression, and immersion. Participants also evaluated NPC realism and naturalness positively, although these perceptions were influenced by contextual knowledge, situational awareness, game-state responsiveness, and consistency with the characters' narrative roles. The findings further distinguish conversational freedom from wider narrative agency, where participants more consistently perceived freedom in how they could communicate than control over the broader narrative. Response delays, repetitive or lengthy dialogue, limited contextual knowledge, and interaction friction also demonstrated limitations in the implemented system.

The thesis contributes theoretical, methodological, empirical, and practical knowledge concerning the integration and evaluation of NLI within narrative RPG design. It provides participant evidence concerning free-form NLI in a playable RPG, documents an approach for designing and evaluating NLI-enabled NPC interaction, and identifies design considerations relating to conversational freedom, contextual control, NPC knowledge, responsiveness, and narrative coherence. Rather than positioning NLI as a replacement for established dialogue systems, the findings suggest that its value depends on how conversational interaction is integrated with the wider narrative, gameplay, and interface structures of a game.

Item Type: Thesis (Doctoral)
Faculty: PhD
Depositing User: Library STORE team
Date Deposited: 20 Oct 2026 16:21
Last Modified: 20 Oct 2026 16:21
URI: https://eprints.staffs.ac.uk/id/eprint/9838

Actions (login required)

View Item
View Item