Uploaded August 2014 | Updated September 2026, 40 minutes ago
You learned about the scientific method in school -- but how do we take it into the 21st century? Join us for a talk and discussion by Prof. Rick Gerkin on SciUnit, taking science to the next level using tricks from the software engineering world!
SciUnit: a collaborative framework for data-driven scientific model validation
Prof. Rick Gerkin
Rigorously validating a scientific model’s explanatory power requires comparing its predictions against empirical data – both data available during model development and data gathered after publication. However, in most fields model validation remains an informal and incomplete process. This makes it difficult to evaluate a model’s strengths and weaknesses, or to adequately compare two models.
This project aims to improve model validation by drawing inspiration from unit tests, a common form of software testing that validates a single component of a computer program against a single correctness criterion. We are developing cyberinfrastructure for scientific model validation around analogous scientific validation tests – executable functions validating models against an empirical observation by producing a score indicating model/data agreement.
The project enables collaborative construction, logical grouping, and continuous execution of tests by scientific communities, and updating of tests and their results continuously as new data and competing models emerge. Visual summaries of aggregate results provide an up-to-date report of progress in a research area. Merits and deficiencies of competing models are clearly visible, benefiting ongoing modeling efforts and informing new theories and experiments.
You learned about the scientific method in school -- but how do we take it into the 21st century? Join us for a talk and discussion by Prof. Rick Gerkin on SciUnit, taking science to the next level using tricks from the software engineering world!
SciUnit: a collaborative framework for data-driven scientific model validation
Prof. Rick Gerkin
Rigorously validating a scientific model’s explanatory power requires comparing its predictions against empirical data – both data available during model development and data gathered after publication. However, in most fields model validation remains an informal and incomplete process. This makes it difficult to evaluate a model’s strengths and weaknesses, or to adequately compare two models.
This project aims to improve model validation by drawing inspiration from unit tests, a common form of software testing that validates a single component of a computer program against a single correctness criterion. We are developing cyberinfrastructure for scientific model validation around analogous scientific validation tests – executable functions validating models against an empirical observation by producing a score indicating model/data agreement.
The project enables collaborative construction, logical grouping, and continuous execution of tests by scientific communities, and updating of tests and their results continuously as new data and competing models emerge. Visual summaries of aggregate results provide an up-to-date report of progress in a research area. Merits and deficiencies of competing models are clearly visible, benefiting ongoing modeling efforts and informing new theories and experiments.
