The purposes of this research aim at taking the time concept of lower grade elementary school students as an example to build a two-tier computerized diagnostic test based on Bayesian Networks and evaluate the teaching effects of the remedial instruction media based on the results of the diagnostic test. First, the researcher developed the knowledge structure and two-tier test in time concept of the lower grader upon the related mathematics concepts and misconceptions. After comparing the effectiveness of the three models in Bayesian Networks, the researcher combined the two-tier test and Bayesian Networks to build the two-tier computerized diagnostic test with the best model. Besides, according to the error patterns diagnosed by the test, the researcher used the cognitive conflict strategy to design the remedial instruction media. Finally, the effect of the remedial instruction of the experimental group and that of the control group will be examined. The results indicate: 1. The value of Cronbach α for the two-tier time concept test diagnosis tool is .821, and .739 for the one-tier time concept test diagnosis tool, which means the reliability of the former is better. 2. After comparing the effectiveness in Bayesian Networks diagnosis in the one-tier- and-two-tier test, the accuracy for the two-tier test in model two is .960, and .949 in the model three. Both model two and three are better than the one-tier test. 3. More than 83% of the experimental group gave their approval for the two-tier computerized diagnostic test. More than 88% of those think information technology integrated into remedial instructions could be helpful to their mathematics learning and also approve of the remedial instruction media designed by the researcher positively. 4. The effectiveness of the remedial instruction of the students in the experimental group with scores lower than 64.22 in the pre-test is significantly superior to the control group. The study of the arranged remedial instruction based on the cognitive conflict strategy confirms that the low-level students receive better recovery effect. 5. In the delayed post-test, the average score of the experimental group is 81.72, and that of the control group is 67.43. The notable difference attained by the examination reaches as high as 14.29. This indicates the experimental group is significantly better than the control group in the field of the time delay effect.
