Assessing Human-AI Collaboration in Design Education: A Process-Oriented Rubric Grounded in an Extended AI-TPACK Framework

Abstract

Generative artificial intelligence (GenAI) can now produce visually convincing design proposals in seconds, undermining a basic assumption of product-based grading: that the quality of a final artifact is a reasonable proxy for the quality of the thinking behind it. This paper addresses the resulting assessment gap in two steps. First, we extend Mishra and Koehler's (2006) TPACK model with a domain-general fourth competency, AI-Validation Knowledge — the instructor's capacity to subject probabilistic AI output to deterministic, discipline-specific verification before treating it as evidence of a legitimate design decision — arguing that this competency, not merely technological or pedagogical fluency, is the theoretical warrant for any assessment instrument that evaluates human-AI collaborative design work. Second, grounded in this extension, we report the design principles and expert validation of the Merit Report, a process-oriented rubric developed within a Design and Development Research (Type 2) programme and validated by a purposively sampled panel of senior academics (n = 5). Findings show the instrument underwent a traceable, substantive revision in response to expert feedback and was independently judged by a panel member to align with formal architectural accreditation criteria, evidence that its validation was a genuine feedback process rather than a formality. We then examine, criterion by criterion, which components of the rubric are transferable beyond architecture to other AI-mediated design and applied disciplines and which require a new, domain-specific substrate — a more precise transferability claim than the single-sentence assertions common in adjacent literature. We conclude that the instrument is theoretically well-founded and successfully validated by experts (DDR Type 2), completing the conceptual phase of this research programme. The necessary next step—external validation via scored student portfolios—is planned for an upcoming live studio implementation.