As impressive new AI capabilities emerge, we need new methods for appraising and combining human and artificial intelligence.
Statistical foundations for human-AI complementarity. Models present us with new inputs to important decisions, but how do we pinpoint and design for the complementary strengths of different agents working together on complex decision problems?
Beliefs and uncertainty quantification. How do we know what a model knows? How do we quantify uncertainty in its outputs in principled and human aligned ways?
Model explanations and interpretability. What does a rigorous science for communicating and making sense of model internals look like?
AI for behavioral science. How can integrating AI in the scientific process stimulate researchers’ imagination while avoiding new threats to validity?
Our lab has received numerous paper awards and PhDs have gone on to tenure track faculty positions at highly ranked CS departments. If you’re obsessed with doing rigorous work on these topics, consider applying to work with the lab.
