Automated Grading as Faculty Judgment at Scale

The circadian rhythm of academic life has predictable periods of joy and drudgery, the latter most often typified by the virtual pile of projects and papers at the end of an academic term. The progressive “do more with less” mantra and time demands on academic faculty may tempt and even necessitate greater efficiencies. Finding ways to automate grading may be one part survival and another part propagation of faculty judgment at scale. Policies related to the use of artificial intelligence (AI) for both faculty and students have not kept up with the rapid changes in technological capabilities and the norms for use (McDonald et al., 2025). Faculty already have many options for using AI to create presentations and case studies, brainstorm, and refine their academic writing. The use of AI for partial or even completely automated grading is a logical, albeit ethically fraught, additional use. Ahead of policy-driven guidance, use for this purpose requires careful consideration of context, extent, and perceptions as is typical with the fringes of innovation, efficiency, and academic integrity. The use of automated grading has its roots in optical scanning technologies, leading to the course-management-system-based test systems used today. In terms of subjective and qualitative evaluation, though, automation has been more difficult. At a basic level, faculty have the ability to cut and paste comments they write from student to student, resulting arguably in a form of automation and “self-plagiarism.” Taking automated grading one step further in terms of having an AI review an assignment and give a variety of outputs would be the latest frontier. The line of where automated grading might completely alienate students from the expertise and human touch they expect from faculty remains blurry. As faculty, we can easily create a custom AI agent with most large language models that can emulate our written tone and follow a basic analysis script to provide at least initial review and feedback. But the guidance and cautions given to students related to the use of AI for assignments must also resonate for us in terms of transparency, context, and extent of use. While auto-grading as an emerging faculty practice cannot be ignored in the resource-constrained space of higher education, where efficiency in the face of greater teaching loads may tempt “administrative shortcutting,” the potential for privacy concerns looms large, despite institutional licensing that protects content. Even more concerning is how we may be perceived for this practice, even if there is transparency about the quantity and extent of use. The potential for errors, bias, and even harmful outputs if they are not carefully monitored persists, just as it does for student assignment and clinical use cases. Disruptive technologies have regularly pressed higher education to reconsider and evolve approaches to education, ranging from new technologies like calculators to integrating online learning. Sufficient “human in the loop” approaches that ensure that faculty have the final look and say when it comes to grading may still be ethically tenuous, even if there is transparency around the practice. There is also the paradox of faculty use when student use may be restricted or largely prohibited. Expanded use of automated grading also accelerates concerns about “robots grading the work of robots” (McMurtrie, 2024). Namely, students using AI to produce assignments that faculty would grade with their own AI agents would result in a frightening and illusory world of learning, or lack thereof. Ongoing dialogue at the national, institutional, and especially the classroom levels will need to drive how faculty and students perceive the dual value and intrusion of technology into academic environments. To ignore the impacts and potential benefits to the assessment and feedback cycle is just as perilous as ignoring the impacts on student learning and engagement.