People
Yikai (EK) Lu
-
Assistant Professor
My research interests include latent variable modeling for innovative item types and applications of machine learning, deep learning, and LLM for psychological and educational data.
PhD, Joint in Psychology and Computer Science, University of Notre Dame
MS, Computer Science and Engineering, University of Notre Dame
MS, Applied and Computational Mathematics and Statistics, University of Notre Dame
BS, Psychology, University of Minnesota - Twin Cities
BS, Computer Science, University of Minnesota - Twin Cities
- Latent variable modeling for innovative item types
- Data mining
- Deep learning and large language models
- Knowledge tracing
- Measurement models (item response theory and cognitive diagnosis models)
- I am currently accepting doctoral students in the Quantitative Methods in Education (QME) program. Students with research interests that align with mine are encouraged to apply.
I am interested in integrating innovative item types, such as answer‑until‑correct items, within latent variable models, and data mining, predictive modeling, and applications of AI (e.g., deep learning and large language models) for educational big data, with an emphasis on improved interpretability.
One of the innovative item types is answer‑until‑correct (AUC) items, which are multiple‑choice items where students can keep responding until they choose the correct option. I have developed new item response theory models for this testing procedure and explored applications such as computerized adaptive testing and detection of differential item functioning, which is critical for ensuring fairness across demographic groups.
Educational data mining can yield rich insights about students. From platform log data, we can learn their tendencies and interaction patterns and predict their proficiency, for example, the probability of answering items correctly. In one of my projects, I developed a deep learning model that uses students’ interactions with the system, curriculum information, and students’ class memberships to learn from the data and to uncover the effects of each action type and the associations among problems.
