An empirical evaluation of a serious simulation game architecture for automatic adaptation

This paper presents an architecture designed for serious simulation games to automatically generate game scenarios adapted to player's level and knowledge. We detail two central modules of the architecture: (1) the player model and (2) the adaptation module. The player model estimates the current knowledge of the player using a Bayesian Network (BN). The evidence variables in the BN are assigned through the observation of player's actions and the current state of the simulation. Considering the estimated player's knowledge and skills, the adaptation module uses automated planning algorithms to dynamically adjust the parameters of the simulation, in order to generate scenarios that will be well suited to improve player's knowledge and skills. We implemented our proposed game architecture in a simulation serious game named Game of Homes. The purpose of this game is to teach the basis of real estate. The player is a virtual real estate broker in a city who has to seek for brokerage contracts, estimate the value of houses, fix asked prices, perform visits, and close the deals. The player competes with other brokers driven by artificial intelligence (AI). We conducted a pilot experiment with human participants (N=10) to validate our architecture in Game of Homes. On day 1, participants were asked to take a pre-test about real estate skills taught in our game. On day 2, participants played Game of Homes for approximately 90 minutes and then filled up a motivation questionnaire. On day 3, participants took a post-test. Preliminary results show that in addition to induce strong motivation among the players, Game of Homes significantly improved real estate skills between pre-tests and post-tests. Results suggest that our serious game architecture allows (a) to induce learning process by providing content adapted to the player progression and (b) to keep the player motivated and interested during the game by adapting the challenge and providing new content.