Modeling student learning behaviors in ALEKS: A two-layer hidden markov modeling approach

This study analyzed 25,783 log data entries of student learning activities on a self-paced online intelligent tutoring system. The behavior patterns between the high- and low-achievement students and of different mathematical topics were compared using two-layer hidden Markov model. The results showed that high-achievement students exhibited more effective learning behaviors, such as asking for explanation and practicing after making an error. In contrast, low-achievement students tended to make consecutive errors without seeking help. Moreover, students’ learning behaviors tended to be more effective when learning simple topics. Our findings implied that intelligent tutoring systems could track the behavior patterns of students and detect ineffective learning states, so as to provide learning support accordingly