Leveraging Python-Generated Learning Analytics to Support Student Success in General Chemistry

Doctor Jocelyn E “Josie” Nardo is an Assistant Professor in the Department of Chemistry & Biochemistry at The Ohio State University and a scholar in chemistry education research. Her work examines how learning environments, institutional structures, and disciplinary norms shape students’ experiences and persistence in chemistry, with particular attention to belonging, identity, and professional development in undergraduate and graduate education. Drawing on mixed methods and critical theoretical frameworks, she studies learning ecosystems, mentoring, and academic citizenship to understand how participation and recognition are negotiated within STEM communities. Dr. Nardo’s research integrates qualitative approaches such as narrative inquiry and counter-storytelling with quantitative analyses of large-enrollment courses to inform equitable and evidence-based instructional design. She collaborates with interdisciplinary teams and national partners to redesign chemistry learning environments, develop data-informed feedback systems, and support inclusive mentoring practices. Her scholarship contributes to discipline-based education research, advancing theory and practice aimed at strengthening student success and transforming chemistry education.

Abstract: Large introductory chemistry courses generate extensive assessment, clickstream, and participation data, yet students rarely receive timely, actionable feedback that can meaningfully guide their learning. This project introduces a Python-based learning analytics pipeline designed to transform raw course data into individualized, interpretable reports that help students understand their progress, identify conceptual disparities, and make informed decisions about their study strategies. Using automated scripts, the system integrates data from Learning Catalytics (LC), Canvas assessments, participation records, and instructor-defined concept maps. The pipeline performs data cleaning, correctness detection, round-based comparisons, and score normalization; generates statistical indicators such as quartiles, percent correctness by concept cluster, and historical trajectories; and produces fully self-contained HTML reports with visualizations, embedded plots, and personalized narrative feedback. The computational framework employs pandas for data structuring, NumPy for numerical transformations, Matplotlib for image generation, and custom algorithms that map each LC item or exam question to conceptual categories. Reports synthesize item-level performance, longitudinal improvement, and tailored recommendations derived from a rules-based feedback engine. Students receive these reports after each major activity, allowing them to track learning goals across units, view their “What-If” profiles, and compare first- and second-round LC performance to better understand conceptual growth. Preliminary analyses indicate that students who regularly engage with the reports demonstrate more accurate self-assessment, improved metacognitive strategy use, and more targeted help-seeking behaviors. This work demonstrates how instructor-created Python tools can scale personalized feedback in large-enrollment STEM courses, shifting analytics from a grading mechanism to a pedagogically meaningful resource that empowers student learning.

This seminar will take place in a hybrid format, allowing for both in-person and online attendance. The zoom link for virtual attendees will be shared in the "Upcoming Events" announcement for the week of Mar. 2-6.

If you are a student at UTM and UTSC and would be interested in attending this colloquium in person (travel expenses covered) and meeting our invited speaker, please send your request to chem.reception@utoronto.ca. 

Zoom Meeting Link: https://utoronto.zoom.us/j/87644094555?pwd=2buoB1aQHuo08bDOSx4He6lirooNfH.1
Zoom Meeting ID: 876 4409 4555
Passcode: Colloq2526

All are encouraged to attend!