Methods and Tools for Active Adapation in Serious Games Based on a Rich User Model

点击翻译按钮获取中文摘要

Although the effectiveness of Intelligent Tutoring Systems has been clearly demonstrated, their widespread use has not been a success. The lacking of learning environments that keep students motivated and engaged is one of the major causes of this failure. Serious games provide the playful side, to overcome the lack of motivation but they should adapt to the player to increase the learning gain as well as gameplay experience. However, the rationale, conditions and effects of adaptability remain poorly explored and inadequately assessed in this context. The aim of my research program for the next five years is to investigate innovative methods and design methodology that can make serious games cognitively, emotionally, and socially more adaptive for increased learning gain and gameplay experience. The research program has four specific objectives: 1. To investigate and determine the main features in play for an active adaptation in serious games 2. To explore and develop novel methods for extracting new features from multi-modal user behaviour data 3. To investigate new adaptation models that could accurately predict learner-player behaviour and respond to it by using appropriate adaptation measures; 4. To investigate a methodology for designing highly adaptive serious games with optimal learning gain. Two main hypotheses will guide our studies: 1) A holistic modelling approach that considers all of the learner-player's many facets, including skills, affects, social behaviours (leading to a Rich Learner Player Model RPLM) is the basis for a successful adaptation of a serious game in terms of its educational effectiveness; 2) Mining multi-modal and multi-source interaction data collected during the game will require adapting or improving existing data mining techniques or inventing more appropriate ones. The RLPM hypothesis will be refined by exploring factors associated with every of its aspects. Behaviour multi-modal data mining techniques and appropriated machine learning approaches will be investigated in order to build an accurate adaptation model, including a user behaviour prediction engine based on interaction data collected during the game. The results of this research will open the door to more informed, adaptive and intelligent serious games, with a more accurate prediction model of the learner-player's behaviour. This new generation of games will anticipate the player's actions as well as reactions and will be able to generate positive behaviours and emotions and inhibit those deviants or negatives for pedagogical purposes. Adapting the game based on our RLPM will lead to a more relevant gameplay experience and optimized pedagogical strategies. In addition to the contribution to the advancement of knowledge in serious games, educational data mining and ITS research fields, the well-established game in