Bibliographic details and abstracts are available to all. Downloads of full-text dissertations are restricted to University of Portsmouth members who must login. MPhils may be accessed by all.

Abstract

There is very little research into ways to build varied, procedurally-generated artificial intelligence, but there have been some popular games that have made good use of such AI. Middle-earth: Shadow of Mordor and Crusader Kings II in particular. I have drawn upon these games to design an approach which I refer to as spectral behaviour, which provides an AI character with behaviours based on specified traits and their magnitude. In order to find out if the improvement was substantial enough to warrant future use in game development, I created a game featuring two demonstrations. The first showed what I referred to as a binary behaviour model (when a trait is present, it does not vary in magnitude) and the second featured my spectral model. I presented this game to fifteen participants in a study where they observed and examined the behaviours they were shown. When I asked them if they had noticed a distinct difference in the variety offered, 53% of the participants said they found the spectral behaviour model to be noticeably more varied than the binary model. In contrast, 6% thought the binary demonstration was more varied. 27% noticed no significant difference in variety between the two. 13% had no clear position. While spectral AI requires more development time, the majority of players are likely to notice the improvement, making spectral behaviour a viable solution for creating varied artificial personalities that are compatible with procedural generation.

Course: Computer Games Technology - BSc (Hons) - C1671

Date Deposited: 2019-11-21

URI/permalink: https://library.port.ac.uk/dissert/dis13450.html