Large Language Models Persuade Without Planning Theory of Mind
Jared Moore, Rasmus Overmark, Ned Cooper, Beba Cibralic, Nick Haber, Cameron R. Jones · 2026
Read Large Language Models Persuade Without Planning Theory of Mind by Jared Moore, Rasmus Overmark, Ned Cooper, Beba Cibralic, Nick Haber, Cameron R. Jones (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-17045.txt.
Description
Jared Moore led a six-author team testing whether large language models truly read minds when they persuade, or merely sound as if they do. Most theory of mind tests use static question-and-answer items, but the authors argue that live, first-person interaction is the real test. Their task asks a persuader to sway a target toward one of three policy proposals by revealing facts in a smart order, which requires sensing what the target knows and values.
Across three experiments, models beat human persuaders when the target's knowledge and motives were shown, yet fell below chance when those states were hidden and had to be drawn out. Humans did fairly well in both setups. But when targets were real people, models still swayed beliefs more than humans did. The paper warns against crediting models with human-like mind reading, while noting their real power to shape belief.
Catalog record
| CREATOR | Jared Moore, Rasmus Overmark, Ned Cooper, Beba Cibralic, Nick Haber, Cameron R. Jones |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 83 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-17045