Complementary machine intelligence and human intelligence in virtual teaching assistant for tutoring program tracing

This study proposes a virtual teaching assistant (VTA) to share teacher tutoring tasks in helping students practice program tracing and proposes two mechanisms of complementing machine intelligence and human intelligence to develop the VTA. The first mechanism applies machine intelligence to extend human intelligence (teacher answers) to evaluate the correctness of student program tracing answers, to locate student errors, and to generate hints to indicate errors. The second mechanism applies machine intelligence to reuse human intelligence (previous hints that the teacher gave to other students in a similar error situation) to provide program-specific hints. Two evaluations were conducted with 85 and 64 participants, respectively. The evaluation results showed that the system helped above 89% of students correct their errors. The error-indicating hints generated by the first mechanism help students correct more than half of errors. Each teacher-generated hint was reused averagely three times by the second mechanism. The results also revealed that some error situations occurred frequently and occupied a major occurred percentage of student error situations. In sum, the VTA and these two mechanisms reduce teacher tutoring load and reduce the complexity of developing machine intelligence.