Description: Significant progress has been made in machine learning (ML) systems for predictive tasks. However, their utility diminishes considerably when essential conditions are not met, rendering ML not only ineffective but potentially dangerous. One challenge is the presence of unexplained bias within ML predictions. This unexplained bias, often stemming from the complex relationships between variables, can reduce the reliability of predictions, casting doubt on the fairness and equity of the ML system.
Furthermore, the inherent brittleness of ML predictions further exacerbates the problem. ML models require that the training and testing data are drawn from the same distribution, which hinders their capacity for out-of-distribution generalization. This inherent lack of robustness restricts the suitability of conventional ML systems for the critical task of aiding in personalized decision support. To confront these inherent limitations, this project will leverage causal machine learning, which bridges the disciplines of causal inference and ML. Causal inference equips us with the tools to quantify the impact of interventions, thereby providing actionable insights conducive to personalized decision-making. On the other hand, ML excels in harnessing multi-modal, unstructured, and high-dimensional data, which are increasingly prevalent in today’s data landscape. This fusion of methodologies not only overcomes existing limitations but also paves the way for a more comprehensive and effective approach to data-driven decision support.
This project will include applications across various domains, including healthcare, business and marketing, and engineering, all with the goal of enhancing personalized decision-making. A particular focus will be on modeling temporal causal relationships, which have not been widely explored. The overarching goal is to create AI systems that are not only powerful but also accountable and ethical, ultimately benefiting humanity by providing transparent, understanding, and insightful decision-making support.