Abstract:
Introductory programming courses are a well-known source of early
attrition, yet signs of difficulty often reach instructors only after
major assessments, when little time remains to intervene. This talk
examines how early and reliably students who need support can be
identified, and what it takes to act responsibly on such predictions,
drawing on a project at the University of Doha for Science and
Technology (UDST), funded by the Qatar Research, Development and Innovation Council (QRDI), on AI-driven personalized learning and gamification in programming education. Using data from 650 students in 36 class sections across two academic terms, we developed six regression models to forecast final-exam performance from information available one month before the exam. Under nested cross-validation within a term, the best model achieved over 91% accuracy; when we applied models trained on one term to the next, the best accuracy fell to about 84%, and the model rankings changed. A comparison of the two terms’ correlation structures suggests why: assessment grades became stronger predictors of
the final exam, while completion-time features shifted inconsistently,
with some weakening and others reversing in sign. The talk argues that, in educational settings, distribution shift is a structural property
rather than an exceptional event: each term introduces a new cohort,
revised assessments, and new tools, and an early-warning model that is not actively maintained degrades precisely where its errors are most consequential. The talk also addresses student modeling for personalization, including inferring learning styles from uneven MOOC interaction logs without questionnaires and characterizing students through player typology, personality, learning style, and achievement emotion, along with the limits of such inferred labels. It concludes with the requirements that educational AI places on trustworthy machine learning: drift-aware evaluation and maintenance, predictions instructors can interrogate, safeguards against unfair student labeling, and privacy-preserving deployment.

Bio:
Bekir Sait Çiftler is an Assistant Professor and Head of the Department of Data Science and Artificial Intelligence in the College of Computing and Information Technology at the University of Doha for Science and Technology (UDST), Qatar. His research focuses on trustworthy and efficient artificial intelligence, with an emphasis on how machine learning systems behave after deployment, including distribution shift and model maintenance, small and efficient language models, and runtime assurance for LLM-based agents. His earlier work spans federated and privacy-preserving learning, deep reinforcement learning for resource management in wireless networks, and indoor localization for the Internet of Things. His publications have received more than 1,500 citations, and he is a principal investigator on research funded by the Qatar Research, Development and Innovation Council. At UDST, he led the
redesign of the B.Sc. in Data Science and Artificial Intelligence and
its ABET accreditation, designed two master’s programs, and chairs the university-wide AI Committee, under which the institutional AI policy and a Center of Excellence in AI were established. Before joining UDST in 2022, he was a postdoctoral research fellow at Hamad Bin Khalifa University and a machine learning solution architect in industry. He holds a Ph.D. in Electrical and Computer Engineering from Florida International University, and M.Sc. and B.Sc. degrees in Electrical and Electronics Engineering from TOBB University of Economics and Technology and Middle East Technical University, respectively.

https://us06web.zoom.us/j/84636366540
MEETING ID: 84636366540
PASSCODE: 25862