Skip to main content

From short-term gains to deep learning: Designing AI learning agents through productive failure

PDF

Abstract

This conceptual paper proposes a conceptual framework for designing Productive Failure-informed AI learning agents that function as pedagogical facilitators rather than answer-providing tools. Grounded in Productive Failure (PF), the framework structures AI-supported learning across three interconnected phases: idea generation and exploration, consolidation, and knowledge assembly. Within each phase, it specifies the role of the AI facilitator, the structure of AI–learner interaction, and the cognitive processes intended to support learning. The framework positions AI agents as mediators that activate prior knowledge, encourage multiple solution attempts, support comparison with canonical knowledge, promote recognition of knowledge gaps, and guide the organisation of emerging understanding into generalisable principles. Theoretically, the framework extends PF into AI-mediated learning contexts by translating its underlying mechanisms into interaction design principles. Practically, it provides educators and developers with a pedagogically grounded approach to designing AI-supported environments that prioritise cognitive engagement, conceptual understanding, and knowledge transfer. Future research should empirically examine the framework across disciplinary, learner, and institutional contexts.

Keywords

Generative AI, Productive Failure, AI learning agent, learning design, Conceptual paper.

Author Biography

  • Polly Lai, Western Sydney University

    Senior Lecturer in Digital Learning