Uncertainty Quantification Overview

Principles

Definition

Uncertainty quantification is an academic field and methodological approach focused on identifying, characterizing, and propagating uncertainties in computational and mathematical models and their associated data. It encompasses a suite of techniques to assess how variability and lack of knowledge in inputs affect model outputs and predictions, thereby providing measures of confidence and enabling informed decision-making under conditions of uncertainty.

Ontological type

Core Methods

Engineering Applications

Error Propagation

Key developments

Key Figures

Robust and Probabilistic UQ era

Eduardo S. Schwartz [1] is associated with the American Finance Association [3] and the University of Washington [4] in this era. His key contribution, exemplified by the 1994 paper Investment Under Uncertainty [6], formalized how investment decisions respond to uncertainty and helped lay the groundwork for uncertainty quantification in capital budgeting. Robert S. Pindyck [2] is associated with the Massachusetts Institute of Technology [5] in this era. His key contribution, as reflected in Investment Under Uncertainty [6], showed how uncertainty affects the timing and magnitude of investments, providing an early probabilistic perspective that complemented other non-probabilistic approaches of the era.

Ensemble and Calibration Methods era

Nicholas Bloom [1] is a leading figure in the Ensemble and Calibration Methods era (2003-2016), with affiliations at Stanford University [3] and the University of Pennsylvania [4]. Bloom [1]'s key contributions include introducing measures of economic policy uncertainty [6] and analyzing the role of uncertain business cycles [7], which advanced how uncertainty could be propagated through models and informed policy-relevant forecasting in this era. Steven J. Davis [2] is a prominent economist with ties to Stanford University [3] and the University of California, Los Angeles [5] during this era. Davis [2] centers his contributions on the paper Measuring Economic Policy Uncertainty [6], whose findings highlighted the macroeconomic impact of policy uncertainty and motivated uncertainty-aware forecasting and calibration practices central to this era.

Probabilistic Deep Surrogates era

Tyler J. VanderWeele [1] is affiliated with Harvard University [3] and the University of California, San Francisco [4] during this era. In the 2017 paper 'Sensitivity Analysis in Observational Research: Introducing the E-Value' [6], he introduced the E-Value [6], a simple sensitivity metric that quantifies how strong an unmeasured confounder would need to be to overturn an observed association, thereby advancing uncertainty quantification in observational causal analysis. Peng Ding [2] is associated with Harvard University [3] and Stanford University [5] during this era. As a co-author of the 2017 paper 'Sensitivity Analysis in Observational Research: Introducing the E-Value' [6], Peng Ding helped advance probabilistic uncertainty quantification through the E-Value [6], a tool for calibrating and interpreting observational findings.