Uncertainty quantification, or UQ, is the set of tools that quantitative scientists use to answer a critical question about any mathematical model of a real-world system: “How confident are we in what the model predicts?” In mathematical biology especially, uncertainty enters at every stage of the modeling process. Data are often sparse, noisy, and heterogeneous. Key biological mechanisms may only be partially observed, even in the best case. Model inputs such as parameters, initial conditions, and even modeling assumptions can rarely be measured directly and must be inferred instead. As a result, two models can fit the same data equally well while implying very different uncertainties in the inferred parameters and mechanisms. This undermines the credibility of future predictions. UQ provides a rigorous way to express confidence in predictions, identify where uncertainty comes from, and communicate which conclusions are robust versus which depend on poorly constrained assumptions. This is especially imperative in health-related applications, where model-based conclusions can influence experimental priorities, clinical decision-making, and public policy. Consequently, guidance from the U.S. Office of Management and Budget has emphasized the importance of sensitivity analysis and uncertainty analysis for sound decision-making.

 

This workshop will emphasize a unified view of UQ applied to mathematical medicine and health that spans the full modeling cycle, from data to inference to prediction, and back again. While UQ has a longer history in fields like engineering and the physical sciences, its systematic use in mathematical biology is relatively new and still uneven across subfields. Part of the reason is that early models of biological processes often relied on qualitative agreement or single bestfit calibrations, which is true even today of computationally complex models such as agent-based models (ABMs). The bigger reason is practical. Biological data are far removed from the data assumed in clean theoretical settings. Measurements are sparse, noisy, and indirect, collected at mismatched scales, and shaped by selection effects, batch effects, and unobserved covariates. This creates a gap between what a method guarantees on paper and what is feasible in practice.

 

To bridge this gap, the workshop will bring together UQ method developers, mathematical biologists, and scientific computing researchers, along with early-career researchers and students who are adopting these tools in real modeling applications. The program will be designed to support an exchange of ideas and expertise between communities, pairing perspective talks with applied sessions that walk through UQ processes in representative biological case studies. In addition to clarifying what is feasible under realistic data and computational constraints, we will emphasize practical guidance, including how to choose and justify UQ approaches, how to recognize when data do not support the intended inference, and how to report results in a transparent and reproducible way. A central goal is that participants leave with concrete templates that they can use immediately in their own work, including hands-on experience with modern software tools and reproducible examples that translate principles into practice.

 

To highlight the applied focus of this workshop, and to build a practical guide or reference work for those in the field, workshop participants will be invited to plan a perspective paper that will be formulated during open discussion times and finalized after the workshop.