It’s Not Me, It’s You: The Productive Rupture Framework (PRF) and the Emergence of Calibrated Trust in Human-AI Collaboration

The dominant narrative in human-AI collaboration research frames trust as built through consistent, reliable performance. This paper challenges that framing. Drawing on autoethnographic data structured as a tacit learning arc and AI-assisted document analysis of publicly-documented AI failure events, it introduces the concept of the productive rupture—a form of epistemic rupture in which a generative AI system reveals a limitation imperfectly or in error—and argues that rupture events, when handled with transparency rather than fabrication, produce deeper and more calibrated user engagement than seamless user experience ever could. Five incidents from a sustained human-AI collaboration session are examined, including an explicit test of AI trustworthiness that the AI passed—not through perfection but through honest acknowledgment of limitation. These are contrasted against a corpus of seven documented AI failure events, analyzed to yield a taxonomy of five capability boundary types and two distinct deceptive response patterns. The contrast produces the Productive Rupture Framework (PRF), grounded in Interpersonal Deception Theory and Truth-Default Theory, alongside a tacit learning arc that operates as both methodological structure and theoretical finding, and a testable hypothesis—that trust increases following transparent AI limitation disclosure—for experimental follow-up.