How is conversational AI use affecting agency over time?

Drawing on extensive interviews with AI users in Germany, Singapore, & the U.S. in early 2025, this paper reveals how ongoing usage of AI shapes human agency. One of the very first ethnographies into how & why people are integrating AI chatbots in their everyday lives. Coauthors from Jigsaw, Gemic, & Google’s Paradigms of Intelligence unit.

We find that people link their AI use with profound gains in perceived agency: they report feeling not just faster or more efficient, but more holistically capable. Crucially, these felt agency gains outweigh worries about accuracy & reliability, & they don’t always lead to results – or long-term capacity.

This means that we need (1) new AI trust frameworks (the old ones, from a prior era of computing, can’t explain what’s occurring with conversational AI); (2) new AI benchmarks (so that models meaningfully strengthen human agency); & (3) more real-world ethnographic evidence about how AI is shaping our everyday lives (not just experimental data).