INFORMS DAS Webinar: Toward More Objective Causal Modeling For Decision Support

In recent years, disputed risk management and policy decisions in public health and other complex domains have highlighted the need for more reliable and transparent causal models. Traditional approaches to causal inference often rely on untestable modeling assumptions, which can undermine the credibility and robustness of the resulting models. This webinar will introduce an alternative, data-driven framework for causal modeling that emphasizes objectivity by focusing on empirically testable assumptions. We will explore how Causal Bayesian Networks (CBNs) and influence diagrams can in principle (i.e., if relevant data are collected) be learned and validated from observational data, without the need for speculative counterfactuals or subjective judgments. Using a principle of Invariant Causal Prediction (ICP) and advanced machine learning algorithms, this approach provides a more robust foundation for decision optimization, ensuring that the supporting causal models can be independently verified – and, if necessary, refined and corrected – based on statistical mismatches between predictions and observations. Practical applications of this objective causal modeling approach are illustrated with examples from health risk analysis and decision support. The webinar will also address practical challenges such as dealing with residual confounding, missing data, and measurement errors, offering strategies to improve the reliability and validity of causal inferences and models in complex decision-making.

Presenter: Louis Anthony (Tony) Cox Jr., Ph.D., Associate Professor of Business Analytics at the University of Colorado Denver

https://connect.informs.org/das/event...

⭐️Join INFORMS: https://bit.ly/3BkR66X
⭐️YouTube channel: https://bit.ly/3zeAsEl

Social Media:
➡️Instagram:   / informs_orms  
➡️Twitter:   / informs  
➡️LinkedIn:   / info.  .
➡️Facebook:   / informspage