Who Acts, Who Knows, Who Answers? A Corpus-Assisted Discourse Analysis of Agency, Epistemic Responsibility, and Accountability in Generative AI Higher Education Research

AI is linguistically framed as the primary actor in higher education research, appearing 2,050 times across the corpus with 447 different predicate associations—yet this active framing masks a critical accountability gap: "no sentence ever made AI explicitly responsible" for outcomes, despite AI being described as performing actions.

Responsibility is systematically displaced away from AI through linguistic choices such as passive voice and nominalization (e.g., "responsible AI" rather than "X is responsible for Y"), meaning that even when scholars use responsibility-focused language, they obscure who is actually accountable for what specific actions or harms.

Educators and institutions bear the stated burden of accountability, while students are positioned primarily as knowers and verifiers rather than agents—a distribution that may not reflect the actual power dynamics or decision-making authority in AI-mediated educational settings.

The language used to describe AI (tool, collaborator, evaluator, proxy, infrastructure) is not neutral and actively shapes perceptions of who should be held accountable; these framings influence policy and practice decisions about AI governance in higher education without explicit acknowledgment of their normative implications.

There is a persistent mismatch between functional agency and moral accountability: AI is described as actively performing actions, yet the research discourse systematically avoids assigning it explicit responsibility, creating ambiguity about who answers when problems arise.