There’s a concept I’ve seen in many papers investigating generative AI (genAI) in education—the human in the loop. It’s important enough that Anthropic, makers of Claude, include it in their usage policy, stating, “When using our products or services to provide advice, recommendations, or in subjective decision-making directly affecting individuals or consumers, a qualified professional in that field must review the content or decision prior to dissemination or finalization. You or your organization are responsible for the accuracy and appropriateness of that information.”  

The policy includes use cases that are high-risk, and academia is included, but only in some areas: standardized testing, certification exams, and accreditation. Scholars of genAI within teaching and learning, however, have identified other areas where the concept would be useful.   

Human in the Loop (HITL) is an old concept with several different applications, one of which is in creating automated systems. Here are many reasons it’s a good idea. Whether you are a lawyer creating case citations or a newspaper running a suggested summer reading list, careful human review of output can help avoid embarrassing mistakes. But, really, it’s more than that. A hallucination uncaught is also a clue to a job undone or poorly done. If the AI output replaces the human input, the human’s value comes into question. 

Classroom-Specific Risks

In any case, although Anthropic’s policies don’t include classroom instruction as a high-risk domain, scholars have included HITL as a good practice in several areas in higher education. One is using genAI in formative assessment. This assessment could include providing empathetic, nuanced feedback (Stoica 2022) that the learner can engage with without becoming discouraged, and that’s something genAI feedback will struggle to do convincingly. There is also the matter of summative assessment. At a bare minimum, an instructor ought to be familiar enough with the assessment model and outcome to be able to explain a score to a student.  

Another concern is the creation of instructional materials, where genAI materials can have unintended consequences. These can range from the silly to the irresponsible to the harmful. GenAI tools can draft case studies, examples, or even create full lesson plans in seconds, and that can be a real time-saver for a teacher. But in reaching students effectively, content and delivery both matter. As Kehoe (2024) notes, while the lesson structure might be generally applicable, instructors need to use their own insights to make materials that are both correct in context and are appropriate for their students.  

Human Nuance 

A third area of concern is diversity, equity, and inclusion. Because genAI tool output reflects biases in input through the training data, automated feedback could miss the mark, create misunderstandings, or even offend. Authors such as Klimova and Chen (2024) and Jenks (2025) suggest that humans need to attend to the cultural factors before the feedback is communicated to students. People with disabilities may similarly find the feedback difficult to use (Judijanto 2025; Pierrès et al. 2024). For example, some genAI replies include emojis, which could be difficult for people with vision impairment to access.   

In a recent survey released by AAC&U and Elon’s Imagining the Digital Future center, 86 percent of respondents expressed a fear that the fundamental role and structure of the faculty job will change with genAI. However, present scholarship repeatedly suggests the value of keeping the human in the loop.  

References  

Jenks, Christopher J. 2025. “Communicating the cultural other: trust and bias in generative AI and large language models.” Applied Linguistics Review 16 (2): 787–795. https://doi.org/10.1515/applirev-2024-0196. 

Judijanto, Loso. 2025. “Beyond Access: Cultural, Ethical, and Infrastructural Challenges of AI in Marginalised Education Contexts.” European Journal of Contemporary Education and E-Learning 3 (6): 83–98. https://doi.org/10.59324/ejceel.2025.3(6).07. 

Kehoe, Frank. “Leveraging Generative AI Tools for Enhanced Lesson Planning in Initial Teacher Education at Post-Primary.” Irish Journal of Technology Enhanced Learning 7 (2): 173–182. https://doi.org/10.22554/ijtel.v7i2.124. 

Klimova, Blanka, and Jui Hua Chen. 2024. “The Impact of AI on Enhancing Students’ Intercultural Communication Competence at the University Level: A Review Study.” Language Teaching Research Quarterly 43: 102–120. https://doi.org/10.32038/ltrq.2024.43.06. 

Pierrès, Oriane, Markus Christen, Felix Schmitt-Koopmann, and Alireza Darvishy. 2024. “Could the Use of AI in Higher Education Hinder Students With Disabilities? A Scoping Review.” IEEE Access. https://doi.org/10.1109/ACCESS.2024.3365368. 

Stoica, Eva. 2022. “A Student’s Take on Challenges of AI-driven Grading in Higher Education.” Bachelor Thesis, University of Twente. https://purl.utwente.nl/essays/91784. 

About the Author 

Amanda Sturgill, associate professor of journalism, is the 2024-2026 CEL Scholar. Her work focuses on the intersection of artificial intelligence (AI) and engaged learning in higher education. Dr. Sturgill also previously contributed posts on global learning as a seminar leader for the 2015-2017 research seminar on Integrating Global Learning with the University Experience. 

How to Cite This Post 

Sturgill, Amanda. 2026. “The Human in the Loop: Considerations for Generative AI in Academia.” Center for Engaged Learning (blog). Elon University. February 26, 2026. https://www.centerforengagedlearning.org/the-human-in-the-loop-considerations-for-generative-ai-in-academia/.