Bruce Rushing

Hello, I am Bruce Rushing,
a machine learning researcher and philosopher of science.

I work on machine learning interpretability, statistical modeling, and data mining. My publications include work on the application of the No Free Lunch theorems to scientific theory choice, the effect of cold posteriors on adversarial robustness in Bayesian Neural Networks, and whether the Scaling Hypothesis is falsifiable. In addition, I have theoretical and experimental work explaining the superiority of Mixtures of Experts over Bayesian Model Averages, the application of large language models for causal discovery, the use of truth vectors for interpretability, and the role of heuristics to avoid overattribution errors to AI. Lastly, I have taught a rigorous machine learning course at Purdue University that applies theory and experiments toward philosophical problems like induction, bias, and alignment. Please check out my project portfolio for examples of past projects I have worked on. I am also on GitHub.

My philosophical work focuses on the philosophy of machine learning, emphasizing using better philosophy and math to improve our understanding of machine learning algorithms and on using machine learning to develop a better philosophy of science. I also work on F. P. Ramsey, particularly his model of scientific theories in his philosophy of science and decision-making under uncertainty.

I have a Doctor of Philosophy in Philosophy and a Master’s in Mathematical Behavioral Sciences from the Department of Logic and Philosophy of Science at the University of California, Irvine. Currently, I am a Research Scientist (AI Specialist) at the University of Virginia’s Data Analytics Center. Previously I was the Ross-Lynn Postdoctoral AI Researcher at Purdue University’s VRAI Lab.