Abstract

"Teaching AI and History"

Louis Hyman, Johns Hopkins University (lhyman@gmail.com)

During the past year, we have lived through a revolution in the accessibility and power of large-language models. This fall at Cornell, I decided to teach a class on that required the use of ChatGPT for all students. The course combined a traditional econometrics sequence with readings in critical data practices in history. Students, who had never coded before, used real historical data sets (ranging from the 1850 Census to the 1950 Survey of Consumer Finance) in ever-more complicated historical analyses. AI was used as an accelerant to give students the tools to quickly ask larger historical questions and probe more complicated data sets. By using AI, students could focus on historical methods and approaches while still using data.
I will describe the class and its methods, its successes and failures, as one model of how to incorporate AI into the study of history. I remain very optimistic about its ability to draw students into historical thinking (and the history major!) while giving them tools (statistics, python, AI) that will be useful in the world.