Large language models (LLMs) produce language that is often indistinguishable from human-generated language, but little is known about how people process it. Do they attribute human-like mental states and accountability to machine-generated language? Through the lens of pragmatic theory we explore how people process language when it is marked as generated by AI. In Experiment 1, we find that people believe the content of sentences presented as AI generated, even if they deem the source unreliable. They also accommodate presuppositions presented from such a source, implying that they adopt a general cooperative stance. In both respects, participant behavior resembled that with an unreliable human interlocutor (Experiment 2), raising the possibility that people reason over machine generated language the same way they reason over human language.
Bródy et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: