Answers are grounded in this paper and cite their sources.
The analysis and decision-making of scientific data and models have become increasingly important with the progress of data science, artificial intelligence, and computing. Although such analyses draw on available data, limited samples, measurement errors, missing data, and model assumptions all affect the results. In addition, many complex systems are inherently random. Therefore, in addition to prediction of a subject under study by a model, the uncertainty in the prediction must be studied. Uncertainty quantification (UQ) uses probability theory, statistics, and computational methods to study the sources of uncertainty and their magnitudes, and to study the effects of uncertainty in model results. This paper discusses sources of uncertainty in data, models, and randomness; uses confidence intervals, probability distributions, Bayesian methods, and Monte Carlo methods for UQ; and presents applications of these methods in healthcare, climate science, and artificial intelligence. UQ does not aim to make unknowns vanish; it makes model results more credible by describing the uncertainties in the results.
No takes yet. Share an insight, caveat, or question.
Wenjia Zhang (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: