Author ORCID Identifier

https://orcid.org/0009-0000-4312-4896

Date Available

9-6-2024

Year of Publication

2024

Document Type

Doctoral Dissertation

Degree Name

Doctor of Philosophy (PhD)

College

Engineering

Department/School/Program

Computer Science

Advisor

Dr. Brent Harrison

Abstract

An important challenge in game design is understanding and maintaining player engagement. This is particularly crucial in both entertainment and educational games, where the player's commitment to the game directly impacts their experience and learning outcomes. However, quantifying engagement proves challenging due to the diverse interests of players. This dilemma is addressed through adaptive game design techniques where the game world is personalized to suit player preferences. In computer games, this personalization is facilitated through Experience Management and Player Modeling, where an intelligent agent gathers information on the player, including their preferences and actions, and takes actions to modify the game world to better suit the player.

This approach to adapting games to players currently hinges on the assumption that player preferences remain relatively stable over time, allowing the intelligent agent to refine its player model accordingly. However, this assumption becomes problematic when player preferences undergo significant shifts during gameplay. It is difficult to model every aspect of a player's preferences and thus predict such shifts, but these shifts can be handled without being directly anticipated. In this dissertation, I create a system that allows for the detection and recovery of shifts in player preferences. This technique makes use of a two step process where the experience manager either watches the player for behavior that is anomalous compared to their previous behavior, or attempts to recover the player's preferences and form a new player model once it detects that the player's current preferences are no longer similar to their past model. This process makes use of a new type of game object that I call \textit{distractions} which help with testing the player's preferences without relying on the existing environment as it may be limited by previous experience manager interventions.

My research contains three parts, the detection of player preference shifts, the recovery of the player model, and the verification of the system on human data. I use an artificial testing environment that mimics aspects of human behavior to evaluate these systems in an automated fashion and a human evaluation of the artificial testing environment to validate the claims made with the automated tests.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2024.394

Download

Share

COinS