Connecticut College AI and the Liberal Arts Symposium

This lightning talk introduces Iterative Failure Tracing (IFT), a framework that reframes generative AI errors as catalysts for critical inquiry. While Large Language Models threaten to replace student effort, their propensity for noun-heavy syntax and factual hallucinations creates a productive validation burden (Sanz-Tejeda et al., 2026). By identifying these critical incidents, students transition from the success trap of passive acceptance to genuine rhetorical authority (Colby, 2025). My presentation demonstrates how prompting as writing functions as a rigorous rhetorical act. Drawing on sabbatical findings from 42 student artifacts, I illustrate the metacognitive shift that occurs when students perform linguistic repair on AI output (Graham, 2023). Finally, I will examine how to scaffold assignments that require students to locate and correct AI failures, ensuring the human writer remains the primary agent of meaning-making in the liberal arts tradition.