Pedagogical Unforgetting Through Red-Teaming: How Adversarial AI Engagement Transforms Teacher Identity and Practice

• In Event: Educators and AI: Perspective, Practice, and Professional Growth

Abstract

This study examines how red-teaming transforms teachers’ pedagogical identity and practice. Red-teaming involves deliberately provoking AI systems to generate harmful outputs. Twenty-five educators engaged in structured adversarial AI exercises, then wrote reflections analyzed through Rodgers’ Deweyan framework. Analysis revealed three trajectories: Guardians developed safety protocols, Transformers redesigned curricula to expose AI limitations, and Resistors preserved AI-free spaces. All participants experienced “pedagogical unforgetting,” reconnecting with educational values that technosolutionism obscures. Red-teaming catalyzed movement from passive trust to active critique, repositioning teachers as critical agents rather than passive adopters of educational AI.

Authors

• Devan Walton, University at Albany - SUNY

• Jaesung Park, University at Albany - SUNY

• Yingru Zhao, University at Albany - SUNY

• Haesol Bae, University at Albany - SUNY