How People Use AI, and What It Does to Learning and Judgment

AI systems are now used by millions of people daily, but we know surprisingly little about how they are actually used or what effect they have on the people using them. I present three lines of work. First, Clio, a privacy-preserving platform for studying patterns across millions of real AI conversations, and what it reveals about how people use AI for work and learning. Second, a randomized controlled trial with professional software engineers showing that AI assistance can impair skill formation, but that how people interact with AI matters as much as whether they use it: people who asked for explanations learned more than those who delegated their work. Third, ongoing work on how experts make judgment calls when working alongside AI agents, and what this means for human oversight. I will close with open questions about how people can preserve their own agency as they rely on AI for more of their thinking and work, and would welcome the group’s input.

Alex Tamkin is a researcher at Anthropic.