Search Results

ABSTRACT

Artificial intelligence (AI) is increasingly being explored in educational assessment, but its use in high-stakes performance contexts requires careful design to support quality, consistency, and human accountability. This study designed and evaluated an AI-supported workflow to assist assessors in marking teaching performance assessment (TPA) portfolios. TPA assessors first described how they usually evaluate portfolios, and this process was used to develop a meta-level rubric logic, prompts, and workflow for the AI agent. The workflow was tested by comparing human-only marking with AI supported marking in relation to time, accuracy, cognitive load, feedback quality, usability, bias, and fairness. The findings showed that the AI-supported workflow reduced marking time and perceived workload while supporting more structured, evidence-based feedback. It also contributed to greater consistency and fairness in the moderation process. The AI agent is positioned as a “third eye” that supports human judgement. Human assessors remain responsible for making final decisions and may accept, adapt or override AI-generated suggestions.

Keywords: GenAI, human-in-the-loop, cognitive load, time efficiency, judgement support, moderation, teaching performance assessment