Investigating the impact of platformer game levels on game character believability

Abstract

Game character believability is often understood as the ability of a game character to convince a human observer that it is being controlled by a human rather than by an artifi cial agent. Behaviour is generally assumed to be the main contributor to the character's believability. The hypothesis that the game's level design also influences game character believability is the main focus of this study. Preference learning was used in an attempt to model this hypothetical relationship and thus determine whether the hypothesis holds and to what extent. A series of experiments was conducted whereby hypothetical models of this mapping were inferred using diff erent algorithms. The highest accuracy obtained is that of 72.83%. Game level parameters were shown to have an indirect eff ect on believability as modelled in this study. Further work on this study opens up the possibility of optimising game character believability using level parameters.