AIAI Seminar-4 March 2024-Talk by Jake Barratt and Jorge Gaete Villegas

Speaker: Jake Barratt

Title: Improving Model Fairness with Time-Augmented Bayesian Knowledge Tracing

Abstract: modelling student performance is an increasingly popular goal in the learning analytics community. A common method for this task is Bayesian Knowledge Tracing (BKT), which predicts student performance and topic mastery using the student’s answer history. We demonstrate an inherent bias in BKT with respect to students’ income support levels and gender, using publicly available data. We find that this bias is likely a result of the model’s ‘slip’ parameter disregarding answer speed when deciding if a student has lost mastery status. We propose a new BKT model variation that directly considers answer speed, resulting in a significant fairness increase without sacrificing model performance. 

This paper will be presented at the LAK24 conference in March.

Speaker: Jorge Gaete Villegas

Title: Developing an Enhanced ICU Patient Mortality Prediction Model via Stochastic Block Modelling and Machine Learning

Abstract: Severity of illness scores offer critical assessments of patient severity by quantifying their deviations from normal clinical ranges and predicting mortality risk. However, their effectiveness can be compromised by various domain-specific issues, including incomplete patient information, complex relationships between clinical variables and medical outcomes, and patient heterogeneity. Despite recent efforts integrating machine learning to improve these models, challenges persist, such as reliance on complete data, handling of data sparsity, and increased model complexity. In this talk, we propose the use of hierarchical Stochastic Block Modeling (SBM), an approach drawn from network science, to address these limitations while leveraging existing medical knowledge. Our approach results in models with significant advantages over conventional ones, including resilience to incomplete data and adaptability to patient heterogeneity while maintaining top performance with reduced information requirements.