Designing Epistemic Friction: Architecture and Protocols for Trickster AI in Leadership Education

Abstract

This article presents a technical architecture and classroom protocols for Trickster AI, an approach to generative AI in leadership education that prioritizes epistemic friction over optimization. Building on a companion paper, it operationalizes trickster pedagogy through a five‐layer system: pedagogical policy, cultural logic, contradiction engine, rupture scheduler, and debrief‐evidence layers. It presents four protocols: the contradiction clinic, refusal studio, delay device, and relational accountability simulation, guiding interaction, debriefing, accessibility, trauma‐informed care, and relational assessment. The central contribution is a bridge between decolonial theory and educational design practice, showing epistemic friction can be specified as a design problem, ethically governed and translated into pedagogical protocols.