Research
The Challenge
Every model is wrong — so why trust them?
Climate models are the backbone of our understanding of Earth’s future. They guide decisions about infrastructure, energy, and adaptation that involve billions of dollars and countless lives. Yet every model is imperfect. This isn’t a dig at climate science; it’s a fact about the very nature of scientific models. Each one simplifies reality, leaving scientists, policymakers, and the public unsure whether to act on their results. The stakes could not be higher: how do we plan responsibly when the very tools we rely on are flawed?
My Work
Turning error and disagreement into engines of trust.
I investigate how climate scientists learn from imperfect tools — and how philosophy can make those practices clearer and more credible. My approach is naturalistic: I study real scientific practice through case studies, interviews, and collaborations with modeling teams. I highlight not only how models succeed, but also how scientists diagnose mistakes, refine methods, and sometimes even learn more from failure than from success.
Philosophy, in this sense, is not abstract wordplay: it is a critical companion to science, clarifying what scientists already do and probing its limits. For a deeper dive, check out the game below.
Step into the shoes of a city planner, policymaker, student, or journalist. Explore how error, disagreement, and uncertainty shape real-world decisions.
Play the game
The Insight
Agreement builds confidence. Disagreement sparks discovery. Error drives progress.
My work calls for a change in mindset: trustworthiness does not require truth. Since all models are false in some respects, credibility must come from elsewhere — from patterns of agreement, from productive disagreement, and from the ability to learn through error.
All models are wrong. Some are useful — Stastician George Box (1976)
This shifts us away from a binary view — “perfect or useless” — and toward a more practical stance: models can be wrong and yet useful, flawed but still trustworthy.
Research Themes
Robustness and Evaluation
Why (and when) does convergence across models matter for trust?
Read more →
Error and Failure
How do scientists make progress by diagnosing mistakes?
Read more →
Imperfect Models
Why are weaker models nonetheless useful?
Read more →
AI for Climate Science
How can philosophy guide trustworthy AI for science?
Read more →
Current Projects
Does Model Weighting Reduce Uncertainty?
Bayesian clarification: the risks of narrowing spread without increasing epistemic confidence.
Failure in Science
How scientists learn from failure, from climate modeling, AI, experiments, and beyond.
Geoengineering and Pursuitworthiness
Determing when small-scale intervention experiments are worth pursuing.
Selected Publications
O’Loughlin, R. (2024). Why We Need Lower-Performance Climate Models. Climatic Change. https://doi.org/10.1007/s10584-023-03661-7
O’Loughlin, R. (2023). Diagnosing Errors in Climate Model Intercomparisons. European Journal for Philosophy of Science. https://doi.org/10.1007/s13194-023-00522-z
O’Loughlin, R. (2021). Robustness Reasoning in Climate Model Comparisons. Studies in History and Philosophy of Science. https://doi.org/10.1016/j.shpsa.2020.12.005
Full list in CV →
What’s Next
New projects on failure, uncertainty, and the future of modeling.
- Expanding my account of failure as a source of scientific progress.
- Applying lessons from climate modeling to the design of trustworthy AI systems.
For scientists → Collaboration on uncertainty, robustness, and decision support.
For philosophers → Dialogue on error & scientific discovery, pluralism, and the philosophy of modeling.
For students → Courses and resources on how philosophy clarifies science.
For media → Expert commentary on climate models, uncertainty, and trust in science.
Updated: September 11, 2025