Repository logo
 

Playing with Autonomy: Exploring the Autonomous Behaviour of Non-Player Characters in Computer Roleplaying Games

Files

Thesis(374.34 MB)

Date

2023

Authors

Supervisor

Clear, Tony
MacDonell, Stephen

Item type

Degree name

Doctor of Philosophy

Journal Title

Journal ISSN

Volume Title

Publisher

Auckland University of Technology

Abstract

What does it mean for non-player characters (NPCs) in computer roleplaying games to be autonomous, or to exhibit autonomous behaviour? Can the autonomy of game agents be defined, classified, and compared? Can their scope for autonomous behaviour be captured, and quantified, in any meaningful sense? Although autonomy and autonomous behaviour are common themes in NPC-centric research, these questions remain largely unanswered in the literature. This thesis reports on a novel approach to identify, describe, classify, and quantify the autonomous behaviour of game NPCs in a way that facilitates cross-game comparison. It also explores how the type of autonomous behaviour exhibited by NPCs may relate to the nature of the game worlds they inhabit. This work began with a large-scale systematic review of NPC-centric research literature, one significant contribution of which was a hierarchical map of themes and research areas within the overall domain. From this map emerged a clear division between the behaviours and capabilities expressed by NPCs in-game, and the technological approaches employed to implement those behaviours and capabilities. Based on that finding, and practical limitations regarding access to the implementation of contemporary commercial games, a methodological approach was proposed that focuses on the behaviour expressed in-game by NPCs rather than their underlying technical implementations. The primary means of investigation was ethnographic, based on the author’s prior work to define and field-test an “NPC ethnography” for the study of non-player characters through in-game participant observation. A fieldnote-centric approach to data collection and analysis was supported by a custom Fieldnote Tool, realised in this study as a working prototype. The approach itself, and insights arising from the iterative design of the tool, offer methodological contributions applicable to the study of NPCs in general. This methodological approach was applied through two case studies, conducted in single-player computer roleplaying games The Elder Scrolls V: Skyrim Special Edition and Mass Effect: Andromeda, each involving an extensive campaign of in-game ethnographic fieldwork. During the first case study, limitations identified in existing qualitative coding methods led to the development of a novel “Keyword Expression syntax” for the structured encoding of agent behaviour in both human- and machine-readable form. This syntax represents another major methodological contribution of this work. This thesis reports findings from the two case studies, including a comparative analysis between the two cases, regarding the nature and extent of autonomous behaviour exhibited by NPCs. Those findings are synthesised into an early conceptual model of NPC autonomy, a further important contribution of this work, which facilitates the classification and comparison of behaviours across game contexts. Finally, a proof-of-concept approach is proposed and demonstrated for mapping observed examples of concrete in-game behaviour—captured using the Keyword Expression syntax—to that abstract conceptual model for the purpose of quantitative analysis and cross-case comparison. This conceptual model and associated methodological approach for the study of NPC behaviour and interaction lay the foundation for extensive future work to explore the nature and extent of NPC or agent autonomy in other games, virtual worlds, and real-world settings.

Description

Keywords

Source

DOI

Publisher's version

Rights statement

Permanent link

http://hdl.handle.net/10292/16922

Collections

Doctoral Theses