Hanan Saleh

Hanan is from Edmonton, Alberta, Canada. She holds a Bachelor’s degree in Computer Science form the University of Alberta. Hanan is currently in her last year of study in the MScIs program (thesis route). Her research interests lie primarily in the area of causal discovery in time series data with a focus on deterministic systems which are ubiquitous in nature. The inspiration behind this research stems from the desire to make AI systems more ethically robust.

Title: Causality in Unobtrusive Observations of Learning in Digital Environments

Abstract: The search for causal relations from observational or experimental data is an open and pervasive problem that spans many fields of research. In the area of learning this is especially important. The ability to determine the effect of a new teaching strategy or the cause of an upswing in student performance is persistently desirable. In computer science courses, code editors and integrated development environments (IDE) offer students a plethora of features promising to accelerate the programming process and instil the necessary competency skills for seamless migration to industry. In this work we apply current causal discovery methods to investigate the presence of a causal link between IDE feedback and student competency which is measured by the number of issues at the end of the coding timeline. The coding activities of students and any issues encountered along the way were timestamped over the coding lifetime. Based on expert knowledge, we expressed our causal assumptions in a causal graph of observed and latent variables. Then, due to the nature of the data, we were able to test for causality using Pearl’s d-separation method and Granger’s prediction method. Our findings show that there could be a potential causal link between IDE feedback and student competence growth. However, our results are statistically insignificant. This is partially due to the size of our dataset (N=29) and or the application of an incongruous method. This inspires our next phase of research which aims at exploring new ways of uncovering causal relations in time series data.


Related Sessions

View full schedule