Présentation d'une communication acceptée dans ICERI2018.

Editor's Notes

• #2: Good morning. My name is Thouraya Daouas, I’m presenting here a paper intitled Learning evaluation in a persuasive serious game. It consists on a research work belonging to a part of a thesis of one of my students whose name is Sammadi. I’m a profesor in HEC of Carthage in tunisia and for this work we collaborated with Tunisia for all association.
• #3: The context of this work consists on using serious games in education. We can classify serious games in three big categories. First SG of message which deliver a message in an educational, informative and persuasive aim. Second SG of training which improves the cognitive or driving performances of the users and third SG of simulation or serious play which offers an open panel of practices.
• #4: We are concerned here with the big part of SG which is less studied, upon researchers, called Persuasive SG. It is defined as a media of persuasive communication for a wide public. In this type of SG, the Player- Learner is put right into the action he is not a passive observer. The aim of persuasive SG is to provoke on PL behavioral changes at a cognitive level, affective level and behavioral effects.
• #5: We propose a modelling using Bayesian networks in order to model the Player Learner and Influence diagram for the formalization of the decision problem.
• #6: Bayesian Networks, also called belief networks or probabilistic networks. They are Directed acyclic graphs used to formalize knowledge. Each node of the graph represents a variable which is a knowledge. And the arc represents a probabilistic dependency between knowledges or variables.
• #7: The influence diagram is an extension of the Bayesian networks. So we can found three kinds of nodes: the chance node which is the same in bayesian networks. The decision node which models the different alternatives available to the decision maker. And the Utility node which allows evaluating the alternatives by identifying a criterion to optimize. Arcs here depend on the node kind, if it is an origine node or a destination node of the arc. We can have 4 possibilities.
• #8: We propose an approach following 3 steps. First identifying PL behavioral chacracteristics. Then Modelling and finally Evaluating.
• #9: In order to identify the PL behavioral characteristics we use a psychological test. This test represents an initial behavioral state of the PL before playing. It is based on random couple of images displayed. And each couple corresponds to a PL characteristic, One image represents a positive aspect of the characteristic and the second its negative aspect. After that the PL begins playing the SG. Its a Persuasive SG with several sessions or missions. When playing we have 2 outcomes, positive or negative behaviour of the PL. These outcomes represent an intermediate behavioral state of PL.
• #10: The second step is modelling. Using Bayesian Networks and Influence diagram, we have these elements. Initial characteristics given by the psychological test and Characteristic outcomes given by the game. We model transforming these elements into variables having relationships between each other.
• #11: The third and last step is composed by two floors probabilistic and decision floor. We use principles of influence diagram to provide a behavioral evaluation of the PL in each session of the game. The model calculates the probability of occurrence of behavioral chacteristic as well as the chance of occurrence of results found in different sessions. Then, uses the probabilities just calculated and the concepts of utility and decision of ID to assess behavior.
• #12: In order to apply the model, we colaborated with an association named Tunisia for All. In the framework of a project representing an action against corruption, this association developed a Persuasive SG raising awareness against corruption called Corjalean Planet. In this SG player plays with flows of money which go towards a corrupt monster. The aim of this SG is to demonstrate the Player the utility to divert these flows towards a better use in the interest of all and consequently change his behavior positively.
• #13: This is the first screen of the SG. These are sessions with different levels, each one represents a situation, for example in a school, in the bank, in the post office, etc.
• #14: Based on messages the SG uses storytelling. The player-Learner choses one scene or session, he reads a story explaning the situation and giving him two alternatives of playing, trying to gess if the player is more against corruption or not. After playing and depending of his choice and playing, he will read another message at the end of the session explaining him in both cases the reason of his results.
• #15: This is an example of a scene in a school. When entering to this scene, the player reads a story. He is supposed to be in a period of exam.
• #16: As a conclusion this model evaluates the PL progress using a persuasive SG. Second, It gives an indicator about the effectiveveness of the SG in itself. When applying the model to the SG of anticorruption, we found predictive results about probabilities of occurrence of behavior change. These are promising results. The next step in this work is what we are preparing now. Which is doing experimentation with real data of PL samples.