Abstract

Generative AI offers transformative solutions for public health and clinical practice. However, applying these models in high-stakes domains requires navigating multi-modal time series data, complex causality, and spurious correlations. This tutorial provides a comprehensive, data-to-deployment overview of generative AI in healthcare. We first examine causal analysis to ensure models are robust and grounded in true relationships rather than observational noise. We explore generative models for population health forecasting, tackling time-dependent data and uncertainty quantification. We then introduce advanced techniques—such as utilizing diffusion models for reinforcement learning and optimization—to build flexible clinical decision support systems. Finally, a real-world case study on maternal and child healthcare resource allocation provides a blueprint for responsibly deploying AI in critical care settings.

Tutorial Description

Generative AI offers powerful new tools for healthcare and public health, from forecasting disease trajectories to supporting clinical decisions and allocating scarce resources. However, healthcare deployment introduces challenges that go far beyond standard model development: temporal and multimodal data, noisy measurements, causal ambiguity, spurious correlations, uncertainty, fairness concerns, and the need for reliable decision-making in high-stakes environments.

This tutorial provides a comprehensive data-to-deployment overview of generative AI in healthcare. We will cover four connected themes:

1. Addressing uncertainty in healthcare data (8:00am - 8:50am): We will discuss limitations of current medical AI benchmarks, uncertainty estimation in large language models, robustness in clinical reasoning, and the risks of relying on spurious or poorly validated patterns.

2. Decision-making in healthcare (8:50am - 9:40am): We will introduce methods that turn predictions into actionable interventions, including decision-focused learning, restless multi-armed bandits, reinforcement learning, diffusion policies, and flow-based decision models.

3. Predictive modeling for health (10:00am - 10:50am): We will survey generative approaches for healthcare forecasting and scenario prediction, including health time-series models, multimodal foundation models, mechanistic model integration, and generative agents for public health simulation.

4. Deployment and real-world case study (10:50am - 11:40am): We will conclude with a case study on maternal and child healthcare resource allocation, showing how AI can help practitioners translate high-level policy goals into dynamic, data-driven intervention strategies.

By the end of the tutorial, participants will understand how generative AI methods can be adapted to the realities of healthcare data, connected to downstream decisions, and deployed responsibly in public health and clinical settings.

Target Audience

This tutorial is intended for KDD researchers, graduate students, and industry practitioners interested in generative AI, healthcare analytics, temporal data mining, causal inference, uncertainty quantification, reinforcement learning, and sequential decision-making. No prior healthcare expertise is required, although a basic background in machine learning or data mining will be helpful.

Organizers