AMIA KDDM Working Group Collaborative Workshop
Causal AI for Health: Moving from Prediction to Actionable Inference

Introduction

Artificial intelligence (AI) models have achieved remarkable success in predicting health outcomes using electronic health records (EHRs), multimodal biomedical data, and large-scale observational datasets. However, prediction alone is insufficient for clinical and public health decision-making. Healthcare requires understanding why outcomes occur, what would happen under alternative interventions, and which patients benefit most from specific treatments.

Causal inference provides the theoretical and methodological foundation to move from risk prediction to actionable intelligence. Emerging advances in causal AI integrate structural causal models, target trial emulation, counterfactual reasoning, and modern machine learning methods to estimate treatment effects, address confounding, and improve generalizability across diverse populations.

This collaborative workshop, organized by the AMIA Knowledge Discovery and Data Mining (KDDM) Working Group, will bring together researchers, clinicians, epidemiologists, and AI scientists to explore:

• Foundations of causal inference for health data
• Integration of causal modeling with machine learning and deep learning
• Causal discovery from EHR and multimodal data
• Generative AI for healthcare decision making
• Bias, transportability, and fairness in AI systems
• Real-world implementation and regulatory considerations
• Case studies demonstrating causal AI for precision health

The workshop will feature a keynote by Dr. Yi Guo, a leading expert in causal inference and causal AI using real-world data in medicine, invited talks from leading experts, poster presentations, and an interactive panel discussion on the future of causal AI in healthcare. By shifting the focus from black-box prediction to causal reasoning, this workshop aims to advance trustworthy, deployable AI systems that can directly inform interventions, clinical guidelines, and health policy.

Objective

This collaborative workshop is designed for a broad and interdisciplinary audience interested in advancing the methodological rigor, trustworthiness, and real-world impact of AI in healthcare. The intended audience includes biomedical informaticians, clinical researchers, epidemiologists, biostatisticians, machine learning scientists, health data scientists, and clinician–investigators who are developing or deploying AI models using electronic health records (EHRs), claims data, registries, genomics, and multimodal health data. The workshop will also be valuable for health system leaders, implementation scientists, and policymakers who seek to understand how causal inference can strengthen decision-making beyond traditional predictive modeling. Given the rapidly growing interest in trustworthy and actionable AI, we anticipate a diverse range of expertise among participants. Approximately 25% of attendees are expected to be novice learners seeking foundational knowledge in causal inference concepts such as directed acyclic graphs (DAGs), counterfactual reasoning, and confounding. About 50% are expected to have intermediate-level experience with predictive modeling and statistical analysis and are interested in integrating causal approaches—such as target trial emulation, heterogeneous treatment effect estimation, and causal machine learning—into their research. The remaining 25% are expected to be advanced researchers who are actively developing causal AI methods, working on regulatory or translational applications, or leading multidisciplinary research programs.

By the end of the workshop, participants will:

• Understand how causal inference differs from and complements predictive modeling in healthcare AI applications.
• Identify key sources of bias in observational health data, articulate when causal methods are required for clinical or policy decisions, and recognize appropriate study designs for estimating treatment effects using real-world data.
• Gain insight into emerging approaches integrating structural causal models with machine learning and deep learning, including heterogeneous treatment effect estimation and model transportability.
• Develop a deeper appreciation for the role of causal reasoning in improving fairness, generalizability, and interpretability of AI systems.
• Engage in critical dialogue about how causal AI can move healthcare analytics from risk prediction toward actionable, intervention-oriented intelligence supporting precision medicine, learning health systems, and evidence-informed policy.

Submissions

Topics of Interest

Submissions should be relevant to the workshop themes, including but not limited to:

• Foundations of causal inference for health data
• Integration of causal modeling with machine learning and deep learning
• Causal discovery from EHR and multimodal data
• Generative AI for healthcare decision making
• Bias, transportability, and fairness in AI systems
• Real-world implementation and regulatory considerations
• Case studies demonstrating causal AI for precision health

Submission Instructions

Please use the AMIA 2026 Submission Template (2-page maximum).

All accepted submissions will be presented as poster presentations. Please indicate "Poster Presentation" at the top of your submission.

Important Dates

• Submission Deadline: September 18, 2026
• Notification of Acceptance: TBD
• Workshop: 8:30 AM - 12 PM, November 7, 2026

To submit your work, please prepare a PDF and use the Submit Now button below to upload it.

Submit Now