AI for Computational Research

3.0 credits · Graduate

AI progress has, and will continue to, revolutionize scientific research. On tasks ranging from (some) theorem proving to coding experiments, they are superhuman. However, for much computational research, human researchers must remain in the loop to evaluate, guide, and steer AI, and we must understand its outputs in order to synthesize and effectively communicate with others. In this course, we will work together to develop skills for how to manage AI to help produce high-quality, trustworthy research. Specific modules – depending on individual student interest – will include verifying empirical analyses, theorem proving and autoformalization of math in Lean, coding architecture, qualitative research coding, and paper reviewing. A class project may include building tools and agent scaffolding to help others use AI in the research process. A core theme of the class will be how to develop and retain individual skills while using AI to complement one’s work, along with discussions of the values inherent in the scientific enterprise.