Artificial intelligence is changing many areas of science. Anthropologist Rodrigo Ochigami’s project “Beyond ‘Ground Truth’” explores how researchers are redefining scientific evidence using new AI methods for inference, observation, and prediction. With the Veni grant, Ochigami wants to help scientists think more clearly and critically about changes in their field.
The project explores how researchers are redefining accepted evidence in three key fields: pure mathematics, particle physics, cosmology, and climate science. “This is important not just for anthropologists and science and technology experts like me, but for scientists in all sorts of fields, who need to quickly understand what AI methods are reliable, what they can use, and why,” Uchigami says.
Discount in mathematics
In mathematics, artificial intelligence makes new forms of reasoning possible. Mathematicians are increasingly trying to discover and prove theorems using AI techniques. But what if the chain of conclusions generated by a computer is so long and complex that no one can understand it? Should it then be considered a “proof”? Some mathematicians see computer-aided demonstrations as proof, because computers are supposed to make no mistakes when checking them. Others disagree and say that such a result is meaningless, arguing that research should help us understand mathematical ideas better, not just mechanically produce more proofs.
Observation in physics
In physics, AI methods, such as machine learning, are fueling debates about what counts as observable evidence. Physicists debate the use of AI in their research, for example in discovering new particles or imaging black holes. One of the main points of discussion is that AI-assisted results are often not “direct” observations. They are often generated by algorithms that already contain certain theoretical assumptions.
Forecasting in Climatology
In climate science, Uchigami looks at how researchers make difficult predictions. When creating computer models to predict the future of complex ecosystems like the Amazon, scientists must consider extreme scenarios that have never happened before. So they can’t just rely on existing data. And in terms of AI, there’s a lack of “ground truth” data to build on. This lack of reliable data is a major problem in all of Uchigami’s research.
Since inference, observation, and prediction are fundamental to various scientific fields, Ochigame wants to contribute to a better understanding of how AI is changing the production of scientific knowledge.
