From Traces to Teaching: A Socio-Technical Framework for Process-Based Assessment in an Age of Distributed Cognition

Abstract

Generative AI severs the link between polished products and genuine learning, exposing the limits of outcome-only assessment. This paper advances a socio-technical framework for AI-enabled process-based assessment (PBA) that reframes evaluation as continuous diagnosis embedded in learning. A five-stage pipeline—task → trace → model → feedback → validation—aligns pedagogical intent with instrumentation of interaction, discourse, and multimodal evidence, and treats the human–AI pair as the unit of analysis within a distributed cognition perspective. Methodologically, the framework maps trace types to appropriate model families (e.g., sequential pattern mining, HMMs, NLP) while requiring explainability so insights are actionable. For practice, it specifies teacher–AI orchestration roles that preserve human judgment and defines governance protocols for privacy, fairness, transparency, and cultural responsiveness. The result is a principled route to assess complex problem solving with integrity in the age of generative AI.