When Responsibility Enables Ethics Washing: Responsible Realism as a Critical Lens for Probing Institutional Recommendations for the Use of AI in Higher Education

This work introduces the notion of Responsible Realism, a theoretical resource we propose to characterize the stance in which Higher Education institutions acknowledge the harms and ethical problems of commercial AI, yet still promote its use, shifting the burden of mitigation onto individual end users by calling to “responsible use.” To clarify and exemplify Responsible Realism as an analytical lens, we follow the logic of critical case selection through an interdisciplinary analysis of several institutional guidelines for the adoption and use of (generative) AI in higher education. Through exemplary fragments we illustrate rhetorical moves and assumptions about commercial AI and its use, and probe the analytical reach of the concept of Responsible Realism against actual documentary material. Rhetorics of inevitability, technological determinism, and anthropomorphism offer the premises that underpin and serve as the backdrop against which Responsible Realism operates. Responsible Realism also necessitates prior discursive moves—ethics washing and critical washing—which decouple the acknowledgment of risks and harms from any obligation to act upon that acknowledgment. Altogether, by naturalizing fatalism and bypassing serious critical scrutiny in favor of pro forma engagement, genuine responsibility becomes unfeasible. Paradoxically, by transferring the lion's share of responsibility to the end users, institutions, simultaneously, and in an asymmetric exercise of power, insulate themselves from accountability while retaining the right to claim they acted responsibly by recommending responsible use. Furthermore, tech companies’ accountability is also erased from the equation in the process. We conclude by calling the academic community to critically engage with guidelines and policies for the use of AI, and to reject narratives designed to dilute responsibility.