Biography
Prof Ethayarajh works on behavioral machine learning, an emerging field at the intersection of AI, economics, and behavioral science. His research contends that machine learning is not a sterile industrial process; much in the way that it is hardware-bound and software-bound, it is also bound by the behavior of real-world actors such as consumers, firms, and states. By borrowing from fields like economics, his work formalizes this behavior and creates algorithms, tools, and platforms that are compatible with actual actors, not just idealized ones.
Ethayarajh is best known for creating Stanford Human Preferences (SHP), one of the largest datasets of human preferences over text, and Kahneman-Tversky Optimization (KTO), a widely used algorithm for aligning language models with feedback. His work has been recognized with a Meta Fellowship (2021) and an ICML Outstanding Paper Award (2022), and has been incorporated into models with hundreds of millions of downloads.
Prior to joining Booth, Ethayarajh was a postdoctoral fellow at Princeton Language and Intelligence (PLI).聽He holds a Ph.D. in computer science from Stanford University聽and a B.Sc. in computer science from the University of Toronto.
Academic Areas
- Applied AI