FUNii: The Physio-Behavioural Adaptive Video Game

This paper investigates the use of physio-behavioural detection of fun to model players’ preferences in real-time in the context of an adaptive game. To do so, a Physiological and Behavioural Model of Fun (PBMF), previously trained on 218 players, was used to model players’ preferences based n gameplay events. As a proof-of-concept, we leverged the PBMF to generate a simple player’s preference profile tailored to our test-bench game: Assassin’s Creed: Odyssey, an open-world, action-adventure game. This model associated every player to one of 3 predetermined stereotypical types of player, namely Fight, Stealth and Explore, which are closely tied to mechanics of the Assassin’s Creed series. Using the inferred preferences, we compared an adaptive vs a non-adaptive version of the same game and tested whether the adaptive version was perceived as more fun than the non-adaptive version by the 39 participants of this study. The results point to the creation of an accurate player’s preference profiles during a baseline mission, with profile matching both a “ground truth” Fun Trace – a continuous, subjective rating of a player’s fun – and a self-reported profile with an accuracy of \(69\%\) and \(72\%\) respectively. This, however, did not translate into a measurable difference in reported fun between the adaptive version of the game and the non-adaptive version in neither Fun Trace ratings nor questionnaire answers. Theses findings support that stereotypical preference modelling can be achieved successfully through a physio-behavioural model of fun, but that further investigation on adaptation strategies to those preferences are needed in order to reach the adaptive game’s promise of maximizing player’s enjoyment.