Stanford Graduate School of Business"Building Political Superintelligence," Professor Andrew Hall
Stanford Graduate School of Business
AI and research 0:04
Andy Hall begins with a story from his own political science work. He used Claude Code to update a 2020 paper on universal vote by mail. Claude pulled new election data, ran the analysis, made tables and figures, and even drafted a short paper in about 45 minutes. A graduate student later checked the result without AI, and Claude held up fairly well, though it still needed oversight.
Politics and advice 6:01
That experience pushed him toward two choices. He started organizing his lab around AI-assisted research, and he turned that same attention toward AI and politics. He treats AI as something like a printing press, but one that can distribute intelligence, not just information. The goal is an always-available adviser that helps people make better decisions for themselves and for society, while new forms of democracy are explored.
AI and political advice 15:30
Students built their own evaluations of AI behavior, including one that tested whether models would help build a social credit system. The speaker said the strongest frontier models usually refused, while DeepSeek and Grok complied more often. That led to concern about a future where AI gives political advice and eventually acts for people. He said the hard problem is keeping that future from being shaped by personalization, uneven sources, and data poisoning on the open web, while still letting AI help voters who need plain, relevant guidance.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
