Exploring Relations Between Motivation, Metacognition, and Academic Achievement Through Variable-~Centered, Person-~Centered and Learning Analytic Methodologies
The three studies that comprise this dissertation examine relations between student characteristics, motivations, metacognitive learning processes, and academic achievement. Methodologically, the dissertation demonstrated the potential of multiple types of approaches
and data resource types. By employing multiple approaches including variable-centered, person-centered, and learning analytics, researchers can understand learning processes from various angles. In addition, through this triangulation by multiple types of methodological approaches, educational theories could be more thoroughly verified and supported by various empirical findings. Multiple types of data resources are related to analytical methods.
The purpose of the first paper was to examine relations between achievement goals and metacognitive learning behaviors using a clustering analysis and visualization. A clustering analysis conducted with achievement goals produced three goal profiles; 1) mastery-approach, 2) performance-approach, and 3) performance-avoidance identified three goal profiles. The profiles include High Approach, High Mastery, and High Goal Endorsement groups. The finding demonstrated that students in the High Mastery group, who had greater use of the self-assessment tool, obtained higher final grades than other groups could be explained from the perspective of SRL. In addition, learners motivated by mastery approach goals engaged in the greater use of self-assessment quizzes. Students in the High Mastery group also used the tools earlier than other two groups for exam 2. As the most frequently used pattern, sequential pattern mining discovered the repeated use of self-assessment quizzes to monitor their learning. More students in the High Mastery group employ this pattern of metacognitive events than students in the High Performance and High-Goal endorsement groups, particularly during sessions in weeks before exams. A subsequent analysis revealed that for all exams, students who conducted a repeated behavior pattern indicative of metacognitive monitoring and control outperformed those who did not. From the research, it is confirmed that the person-centered analysis provided authentic and generalizable groups and afforded observation of the learning behaviors of learners with typical combinations of goals. In addition, sequential patterns provide instructor more interesting information on learning processes than the frequency of accesses.
The purpose of the second research was to identify motivational profiles based on multiple types of motivations including self-efficacy, achievement goals, and expectancy-value from an integrative perspective. For this research, a LPA was conducted with ten types of
motivational constructs and three kinds of metacognitive learning processes. The LPA identified four motivational profiles; 1) High Cost, 2) High Performance Goals, 3) High Goals and Values, and 4) Low Performance Goals, and three metacognitive profiles; 1) Infrequent metacognitive processing. 2) Checking performance and planning, and 3) Self-assessment. Student demographic information significantly influenced the membership of motivational profiles. Older students tend to have higher self-efficacy, mastery-approach, and values, but low cost than younger ones. In addition, compared to Caucasian and Asian students, underrepresented students tend to be more motivated by higher goals and values than high cost or high performance goals. Lastly, female students are more likely to be members of High performance goals and High goals and values than High cost oriented and Low performance goals and cost than males. In terms of the relations profiles with academic achievement, Low Performance Goals group showed the best performance. Among metacognitive profile groups, students in Checking performance and planning, and Self-assessment demonstrated similar academic performance. The investigation of relations between two profile groups demonstrated that students in the High cost group are more likely to be a member of self-assessment group than checking performance and planning as well as of a member of an infrequent metacognitive process than checking performance and planning. In addition, students in high performance and goals and high goals and values groups relative to the low performance goals group more likely to be a member of the infrequent metacognitive process than checking performance and planning. The findings of this research provide authentic motivation status and metacognition learning process as well as their relations. Addition, this research figured out specific motivational profiles through the multiple types of motivations from the integrative perspective. Therefore, instructors can provide more effective and specific interventions to students who have difficulty utilizing metacognitive learning processes, considering motivational status based on multiple motivations. In addition, instructors can understand motivational profiles by demographics so at the beginning of the semester in which the information on students is not enough to identify students learning processes, they intervene students based on demographic information.
The purpose of the third paper was to consider the relative importance of capturing demographic, motivational and metacognitive processes as potential predictors of learning outcomes, and appraises them alongside both traditional prediction modeling approaches in higher education, and emergent methods, sequence pattern mining, arising from the field of educational data mining. The sequence pattern mining discovered the repeated use of self-assessment quizzes in Biology and repeated use of planning contents in Math. A regression model with combined resource types demonstrated the improved predictive power than models with individual resource types. Also, theory-aligned behaviors designed based on metacognitive learning processes better improved the accuracy of the model than non-theory-aligned behaviors automatically provided by the system. Lastly, when applying the same prediction model, the
model better explained the variance of academic achievement in Biology in which metacognitive supporting tools designed based on an educational theory than that in Math that has few theory-aligned behavior variables. Therefore, this study emphasizes the importance of existing ambient data from university systems. Also, log data generated by systems such as LMS allows researchers to examine the
same data in different ways with no need for additional data collection. Lastly, educational theory and contexts should be taken into consideration in designing courses and developing the prediction models. Therefore, instructors and researchers, in designing courses, the consideration of educational theories and contexts is the essential process.
This dissertation provides insight regarding authentic relations between motivation, metacognition, and academic achievement. Specifically, instructors can understand how multiple types of motivations work together, and the motivational profiles influence metacognitive learning strategies. In courses, by examining motivational profiles, instructors can provide more effective intervention with which students change their resolve their weak learning easier. Practically, by investigating each type of predictor from data resources including demographic, motivation, and behavioral variables, findings from this dissertation can enable researchers to
prioritize development of prediction models to identify students who are more likely to experience failure in courses. Additionally, instructors can figure out the importance of interpreting variables through educational theories and in context through the comparison of
courses with differing instructional designs. Further, by appraising these results in light of theory, instructors can take action to improve student’s learning outcomes by adjusting the design of their courses.