Publications by authors named "James Enouen"
Article Synopsis
- Recent advancements in large healthcare datasets have enhanced the research of deep learning models, but raise concerns about interpretability, fairness, and biases, especially when human lives are involved.
- The study focuses on the MIMIC-IV dataset to analyze in-hospital mortality prediction models, revealing issues with model interpretability and biases, particularly regarding demographic features affecting fairness in predictions.
- Key findings indicate that while certain interpretability methods can highlight important features for predictions, they also show reliance on demographic factors, leading to disparate treatment and unfair predictions across different patient groups based on ethnicity, gender, and age.
