Research Areas

1. Learning Analytics:

Studies in this area focus on using log data to explore students’ learning behaviors and patterns in digital learning platforms, such as Learning Management Systems (LMS) and Intelligent Tutoring Systems (ITS).

Project 1: Examining Changes in College Students’ Self-regulated Learning Pattern

• Status: Completed (PI: Surina He)
• Description: This project investigated how college students’ self-regulated learning patterns changed following a formative assessment, with particular attention to differences between low-performing and high-performing students. The analysis utilized log data from the University’s Learning Management System, employing advanced process mining techniques, including Fuzzy Miner and pMiner, to uncover insights into behavioral changes.
• Outcome: Published in Computers in Human Behavior (Download Paper).

Project 2: Role of Preschool Children’s Planning Time on Problem-Solving Performance

• Status: Ongoing (PI: Surina He)
• Funding: Alberta Graduate Excellence Scholarship
• Description: This project explored the impact of preschool children’s initial planning time on their problem-solving performance across tasks of varying difficulty levels. The analysis leveraged mobile log data collected from an Intelligent Tutoring System developed by a Chinese educational technology company. Moderated Generalized Linear and Logistic Regression models were employed as the primary statistical methods to examine the data.
• Outcome: Findings from this project are under review.

2. Computational Psychometric:

Studies in this area focus on applying cutting-edge machine-learning techniques and extensive response process data in the educational assessment field.

Project 1: Detecting Aberrant Responses Behaviors in PIAAC Problem-Solving Tasks

• Status: Completed (PI: Dr. Okan Bulut and Dr. Guher Gorgun)
• Description: This study examined aberrant response behaviors in the PIAAC problem-solving tasks by analyzing test-takers sequential clickstream data. The analysis utilized the BERT model to process clickstream patterns, while aberrant responses were detected using Isolation Forest algorithms, offering an advanced approach to identifying irregular behaviors in computer-based assessment.
• Outcome: Findings from this project are published in Zeitschrift für Psychologie (Download Paper).

Project 2: Systematic Review of the Use of Log-based Process Data in Computer-based Assessments

• Status: Ongoing (PI: Surina He)
• Funding: ATLAS Doctoral Research Fellowship
• Description: This study systematically reviewed publications utilizing log-based process data in educational assessments up to December 31, 2023. Approximately 330 papers were identified and analyzed to address four key research questions: (1) What are the trends in using log-based process data? (2) Which process indicators have been constructed? (3) What latent constructs have been inferred from process indicators, and at what inferential levels? (4) What are the benefits, challenges, and future recommendations for using log-based process data?
• Outcome: Findings from this project are under review.

Project 3: Assessing Collaborative Problem-solving Skills Using Communication and Clickstream Data

• Status: Ongoing (PI: Surina He)
• Funding: Alberta Innovates Graduate Scholarship and Andrew Stewart Memorial Graduate Prize Award
• Description: This study assessed collaborative problem-solving skills using multi-player communication and clickstream data. We mainly used word embeddings, transformers, and dynamic Bayesian approaches to explore all response process data.

3. Educational Data Mining:

Studies in this area focus on mining large-scale educational datasets using advanced statistical and machine-learning techniques.

Project 1: Mining the High School Longitudinal Study of 2009 (HSLS:09) Data

• Status: Completed
• Study 1 (PI: Surina He): This study identified malleable factors grounded in Ecological System Theory to predict college enrollment outcomes for low socioeconomic status high school students. Machine learning techniques, including random forest, support vector machine, and logistic regression, were applied. Findings were presented at the 2023 AERA meeting (Download Slides).
• Study 2 (PI: Dr. Okan Bulut): his study explored high school dropout prediction from a human-machine collaboration perspective. Advanced methods, such as random forest, deep learning, and Explainable AI (XAI), were utilized to uncover predictors of dropout risk. Findings from this study have been published in the 2023 NCME meeting (Download Slides) and Discover Education (Download Paper).
• Study 3 (PI: Tarid Wongvorachan): This study addressed class imbalance issues in classification tasks by comparing various sampling techniques, including random oversampling (ROS), random undersampling (RUS), and a hybrid method combining synthetic minority oversampling for nominal and continuous data (SMOTE-NC) with RUS. Findings from this study have been published in Information (Download Paper).

Project 2: Mining the Programme for International Student Assessment (PISA) 2022 Data

• Status: Ongoing (PI: Surina He)
• Description: This project analyzed the Programme for International Student Assessment (PISA) 2022 data, focusing on the impact of discrepancies between self-reported and response-time-based questionnaire-taking motivation on Canadian students’ test performance. Response Behavior Effort (RBE) was employed as a measure of response-time-based motivation, and the data were examined using the Response Surface Analysis (RSA) method to uncover nuanced relationships between motivation types and academic achievement.
• Outcome: Findings from this project are under review.

4. Academic Motivations:

Studies in this area focus on investigating K-12 students’ academic motivation and academic achievements using both cross-sectional and longitudinal survey data.

Project 1: Educational Expectations

• Status: Completed
• Study 1 (PI: Surina He): This study explored the developmental trajectory of Chinese primary school students’ educational expectations using six waves of longitudinal time-series survey data. Additionally, potential predictors were also explored. We used the latent growth curve model and mixture growth curve models to analyze the data. Findings from this study have been published in Contemporary Educational Psychology (Download Paper).
• Study 2 (PI: Dr. Xiaolin Guo): This study examined the gender differences in the inter-generational transmission process of educational aspirations in late childhood using five waves of longitudinal time-series survey data. We structural equation modeling approach. Findings from this study have been published in Sex Roles (Download Paper).
• Study 3 (PI: Dr. Xiaolin Guo): This study examined how mother-child discrepancy in perceived parental expectations mediate the role of two kinds of filial piety (reciprocal and authoritarian) on children’s academic achievement. Findings from this study have been published in Educational Psychology (Download Paper).