Abstract-Intelligent Tutoring Systems (ITSs) are intended to help in tutoring students in specific domains, typically by improving their problem solving skills. An important aspect of such ITSs is the ability to solve problems in the same manner that the student would, in addition to interpreting student actions and providing relevant feedback and help in case of any errors. Cognitive models, that mimic the way procedural knowledge is represented in human minds, are excellent means toward achieving this goal. This paper discusses cognitive modeling in the MAth Story problem Tutor (MAST). MAST is a Web-based ITS that can generate probability story problems of different contexts, types and difficulty levels. A major contribution of the paper is the ability of MAST to symbolize the probability word problems and solve them in the same manner that the student would. The paper discusses the model tracing approach of MAST to interpret the student actions in symbolizing the word problems and estimating the required probabilities to provide relevant feedback and help in case of any errors. Evaluation results have shown the ability of MAST to tutor the students and considerably improve their probability story problem solving skills.