arXiv:2606.05330v1 Announce Type: new Abstract: Large language models can shift human beliefs across high-stakes domains, but most persuasion studies rely on pre/post belief change. These endpoint measures identify whether persuasion occurred, yet miss where and how beliefs moved within a dialogue. We present PERSUASIONTRACE, a framework for studying persuasion in human-LLM interaction. Built on a web-based experimental platform, PERSUASIONTRACE contributes a tool for multi-turn persuasion studies and a process-level evaluation protocol: it records multi-turn belief reports from human or simul
The proliferation of advanced large language models necessitates a deeper understanding of their persuasive capabilities and the mechanisms by which they influence human beliefs beyond simple endpoint measures.
Understanding the process of LLM-driven persuasion is crucial for mitigating risks, developing ethical AI, and leveraging AI for beneficial influence across critical domains.
This framework shifts the focus from 'if' persuasion occurs to 'how' it occurs within human-LLM interactions, enabling more granular analysis and intervention strategies.
- · AI ethics researchers
- · Social scientists
- · AI developers
- · Regulatory bodies
- · Malicious actors using unsophisticated persuasion tactics
- · AI systems lacking transparency in their persuasive mechanisms
More sophisticated tools for analyzing human-AI interaction and belief dynamics will emerge.
Development of 'persuasion-resistant' or 'persuasion-aware' AI agents and user interfaces could become a design priority.
Enhanced understanding of human cognitive vulnerabilities to AI influence could inform new educational strategies and digital literacy programs.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL