Assessment is an important part of learning that should not be only to assess student's knowledge, but also any misconceptions they might make. The teacher may have difficulty in identifying these misconceptions because of the diversity of student's abilities. To solve this problem, an Intelligent Tutoring System (ITS) that offers adaptive exercises is proposed. In adaptive exercises, the next problem to be solved by a student is selected by considering her performance during learning. Unlike former ITS that is mostly based on student's knowledge, the proposed ITS uses student's knowledge and misconceptions to perform adaptation. The development of proposed ITS consists of the pedagogical model and domain model that are implemented to support adaptive exercises, and the student model that is organized and inferred to find the most appropriate exercise based on Buggy Model and Dynamic Bayesian Network techniques. The proposed ITS has been implemented on mathematical exponent and
tested by 34 first-year senior high school students. The results of the experiment show that the most common misconceptions of mathematical exponents are the rules of product, quotient, power and fraction. In addition, the average learning outcomes, on a scale from 0 to 1, for students who use the proposed ITS is 0.64, while students who use the former ITS is 0.44. The results indicate that the adaptive exercise based on a combination of the student's knowledge level and misconception can improve student's learning outcomes compared to adaptive exercises based on the student's knowledge
level only.