AI augments instructor feedback in economics education but systemic inequities persist without structural reform

Indigenous pedagogies center relational feedback, where critique is tied to community well-being rather than individual performance. The trial’s focus on AI as a tool for efficiency ignores these traditions, which view education as a reciprocal process of knowledge exchange rather than a transactional service. Systems like the Māori whakawhanaungatanga or the African Ubuntu philosophy emphasize holistic development over metrics, a stark contrast to the reductionist approach of AI-mediated grading.

The overreliance on AI in education mirrors historical patterns of technological solutionism, from the introduction of standardized testing to the rise of MOOCs, all of which promised efficiency but deepened inequalities. The adjunctification of faculty labor, a trend accelerating since the 1970s, creates the conditions where AI is even considered as a solution to overburdened instructors. The trial’s focus on economics—a discipline that has long been complicit in neoliberal education reforms—further underscores this historical continuity.

In Finland, where education is highly decentralized and teacher autonomy is prioritized, AI tools are used to support—not replace—human judgment, reflecting a cultural emphasis on trust in educators. South Korea’s ‘hagwons’ (cram schools) demonstrate how AI can be integrated into tutoring systems, but also how it risks exacerbating inequality between students who can afford private tutoring and those who cannot. These examples highlight the need for culturally adaptive approaches to AI in education, rather than one-size-fits-all solutions.

The study’s randomized trial design provides robust evidence that AI can improve feedback quality when used as a supplementary tool, but it does not address the reproducibility of these results across disciplines or cultural contexts. The methodology assumes that feedback quality is universally measurable through revision rates, ignoring qualitative dimensions of learning such as critical thinking and creativity. Additionally, the study does not account for the long-term effects of AI-mediated feedback on student mental health or the erosion of human connection in education.

Artistic and spiritual traditions often view feedback as an act of creation, where critique is inseparable from the process of growth. The trial’s framing of AI as a tool for ‘better feedback’ reduces this complexity to a mechanistic output, stripping away the emotional and intuitive dimensions of teaching. In contrast, practices like the Japanese ‘kaizen’ or the Sufi concept of ‘murshid’ (spiritual guide) emphasize iterative, relational learning that AI cannot replicate.

If AI becomes the primary tool for feedback in higher education, it risks creating a feedback loop where instructors are deskilled, students become dependent on algorithmic validation, and inequalities are entrenched. Future scenarios must consider how AI could be used to redistribute grading labor equitably, such as through peer-led feedback systems or community-based assessment models. The trial’s narrow focus on instructor control does not account for these possibilities, instead reinforcing a top-down model of education.

Marginalized students, particularly those from low-income backgrounds or non-Western educational traditions, are often the most negatively impacted by AI-mediated systems that fail to account for their lived experiences. The trial does not include data on how AI feedback affects students with disabilities, non-native English speakers, or those from historically marginalized racial groups. Additionally, the voices of adjunct instructors—who bear the brunt of grading labor—are entirely absent from the narrative, despite their central role in the system.