Patterns of mathematics problem solving and synthetic facial expressions in a personal instructing agent

• The paper clarifies the disagreement in the literature regarding the math-problem persistence behavior differences between genders. • Motivational cues may only be useful for struggling students. • Other forms of motivational cues may be needed to engage average and excellent students. This study explored the patterns of mathematics problem-solving and synthetic facial expressions (SFEs) exhibited by a personal instructing agent named PIA. Toward this goal, 81 Grade 8 students participated in a three-day experiment where they were randomly assigned either to the facial (FG) or non-facial (NFG) group. The students’ interactions within the PIA were collected and stored as log files. The attributes extracted from the log files included types of mathematics problems solved (i.e., schema), status of the mathematics problems solved, difficulty levels of mathematics problems solved, and SFEs exhibited by the PIA. Lag sequential analysis (LSA) disclosed that there were similarities and differences in the sequence of math problem-solving behaviors among students. The Apriori algorithm revealed that struggling students tend to solve problems successfully, irrespective of their sex; however, struggling female students tend to solve more problems successfully than their male counterparts. Nonetheless, regardless of their levels of math competency and the version of software used, all students solved problems they were comfortable with and always started with easier problems, gradually progressing. Limitations and future research were also discussed.