The paper "A Thousand Faces" introduces a persuasion benchmark that grounds persuadees in experience, drawing on psychology and pragmatics. Its starting observation is that two people can hold identical beliefs and desires yet respond to very different arguments: what convinces one may leave the other unmoved.
The authors argue that the relevance of an argument depends on how a person's views were formed. Experiences shape what a person finds credible and what matters to them, so persuasion cannot be reduced to matching a generic belief profile. The benchmark appears designed to capture this variability by modeling persuadees with distinct experiential backgrounds.
This points toward a more nuanced approach to automated persuasion, one that treats individual history as central rather than incidental. The full paper likely details how the benchmark is constructed and evaluated, but the abstract already makes clear that effective persuasion must be experience-grounded.