Beyond “Human in the Loop”: Human Agency Before and After Generative Artificial Intelligence in Education, Assessment, and Measurement

Abstract

Existing “human-in-the-loop” and similar “loop” metaphors are helpful but do not sufficiently position humans as the agents, learners, and responsible actors that educational purposes require. In education, human agency is necessary both before and after GenAI interaction. Beneficial uses of GenAI in education must be human, then AI, then human again, a principle I abbreviate Human-AI-Human (H-AI-H). To develop this argument, I distinguish among three concentric domains: education, assessment (within education), and measurement (within assessment), and identify the actors who must exercise judgment and the expertise they must develop and possess. I use vignettes to illustrate H-AI-H across these educational domains, showing that responsible and effective GenAI use requires human judgment before interaction to define purposes and criteria, and human judgment after interaction to interpret outputs and accept accountability.