Assessing Learning When AI Can Produce the Product

Abstract

Generative artificial intelligence can contribute to the reasoning and communication represented in student work, complicating inferences from product quality to individual capability. This conceptual article develops an integrated framework for assessing cognitive ownership through mutual evaluation and outcome-aligned verification. Cognitive ownership is defined as a learner’s demonstrated capacity to understand, evaluate, adapt, and accept responsibility for consequential reasoning under specified conditions of assistance. Eight indicators are organized into four assessment domains: understanding, epistemic judgment, adaptive performance, and reflective accountability. Mutual evaluation combines independent student and instructor judgments, evidence-based comparison, and documented responses to disagreement; agreement is not treated as proof of learning.             The framework distinguishes the quality of an AI-assisted product, capability demonstrated during assessment, and learning established through change and retention. A worked higher education example illustrates task design, scoring, and responses to conflicting evidence. Five propositions and a validation agenda address incremental predictive value, task sampling, feedback, fairness, and feasibility. The framework is proposed rather than empirically validated. Its contribution is to connect existing validity and evaluative judgment scholarship with a practical account of the evidence needed to support bounded claims about individual and assisted performance.