It provides an interactive learning environment which arm to provide a customized learning experience and motivate learners to further learning and achievement. It has been widely used in many different field, including algebra, physics, and computer science. There are four basic components of ITS, including the student knowledge, domain knowledge, tutoring knowledge, and communication knowledge. Student knowledge is also known as student model. It is a database used to diagnose student 's understanding, learning characteristics, and even the level of concentration. It is used to modify the teaching strategy after the test. Domain knowledge is also known as expert model. It contains the comprehensive theories and problem-solving strategies for the domain which student is learning. It is used to provide the teaching resource. Tutoring knowledge is used to select a customized teaching strategy based on the result of student knowledge and domain knowledge. It gives explanation on concept, including example, quiz, question, display, and analog. It also provides a test and give hints. After the test, the result will be analyzed and a new teaching procedure will be given. Communication knowledge is a user interface that is used to interact with students. ANDES physics tutorcite{ANDES5,ANDESLL} is the most well-known ITS for university physics. It provides an interactive learning …show more content…
Finding a good student model that can precisely match with student 's pattern is very crucial for a better learning experience for ITS, since it provides useful information on choosing the teaching process and offer helps in transferring knowledge on problem-solving. The traditional way to construct the student model is through student interviews, and relational analysis, etc. Those methods are time-consuming and often are subjective. Many researchers are trying to find an effective way to train the AI tutor. Currently, the best method of building the student model is the Bayesian network which is adopted in ANDEScite{BN1}. Each problem is represented by an independent Bayesian network. In the network, it has a solution graph in which the top level of the graph is the goal of the problem. When students make a mistake, the AI tutor can locate the corresponding node and use the known probability to identify student 's problem and give students a useful hint. It has been proven to be helpful in improving student 's problem-solving skill. In additional to the difficulty of building an effective student model, traditional ITS may have problems on the engagement of students. The problem-solving might be fluctuating. This is where the principle of the game based learning comes in. However, the current game based learning is lacking the effective feedback like traditional ITS. Many researchers concern that a student