When faculty distrust students but trust the machine

In a recent AMD promotional video, Oregon State University professor Jonathan Kalodimos explains why he developed an AI platform to help evaluate student presentations. “I can’t really trust their work anymore,” is his opening point, referring to student-written submissions. Dr. Kalodimos puts forward a good point, and one that I have seen in the classroom and my research as well: students are increasingly relying on AI to substitute or augment their work. Because students may use generative AI, many of us no longer see writing as reliable evidence of student learning. Kalodimos explains that he moved toward oral presentations, but assessing hundreds of them takes too much time. His solution was to use AI-assisted evaluation…

I was taken aback when I heard this in the video. 

Let me be clear up front in this piece: I do not think Kalodimos broke a rule. I also understand the problem he is trying to solve related to giving real student feedback. Written assignments are tricky to evaluate as evidence of individual student learning. Faculty workloads make individualized oral assessment difficult. Still, the ethical contradiction is hard to ignore and I kept thinking about the central contradiction: When students use AI, their work becomes untrustworthy, but when faculty use AI, it becomes innovation.

Without giving in to the false equivalency of comparing students and faculty as if they are the same (they are not), there are serious issues in using different moral standards when we are talking about the contentious topic of AI. Quite simply, faculty members have greater authority, and we are accountable for assessment decisions, but that authority should lead to greater scrutiny, not less.

Kalodimos indicates distrust of student writing because he cannot be certain how AI contributed to it (an exceptionally common claim here at OSU). He then relies on another AI system to interpret student speech, apply assessment criteria, and reduce the time he spends evaluating each student. The concern is not unique to our institution. Research on AI text detectors shows how easily automated judgments can be treated as more certain than they are. These systems do not determine authorship. They identify statistical patterns and produce probabilities.

Their performance changes depending on the writing, the model, the editing process, and the population being assessed. Weber-Wulff and colleagues found that detector accuracy declined sharply when AI-generated writing was edited or paraphrased. Liang and colleagues found that non-native English writers were falsely identified as using AI at substantially higher rates. Explain It is not an AI text detector. But the lesson still applies: automated interpretation is not neutral simply because a faculty member reviews the final result. The system decides what information to surface. It interprets the presentation through a rubric. It shapes what the instructor sees first. When the stated goal is to reduce grading time from roughly 25 minutes to eight or ten, the time savings suggest that the system is doing substantial work to narrow and organize the evidence the faculty member reviews. How is this form of AI assistance ethically different from a student using generative AI to organize, refine, or augment a paper? 

As a researcher in the field of AI and a dedicated educator in the modern world, I am confronted by the tension of student AI use every week. I also use AI to help smooth my audio-transcribed feedback of student work. I am transparent about my use, which is fairly high. AI collaborative work is something that will be part of our future whether we like it or not. AI-supported work is not automatically empty or fraudulent. Its value depends on how the tool is used, what human judgment remains, and whether that use is transparent. Kalodimos is taking a clear position within an ethical tension I confront constantly, which is why his framing stood out to me. 

This is also not just a professor describing a teaching strategy.

The video is marketing for AMD, a company that benefits from the expansion of AI infrastructure. The processor is named repeatedly. The platform is presented as a successful use case for AMD technology in teaching and research. That does not prove that AMD personally paid Kalodimos. I have not seen evidence supporting that claim, but financial payment is not the only benefit involved.

Kalodimos gains professional visibility and a research platform. OSU gains attention as an institution developing educational AI. AMD gains a persuasive example of its technology being used in higher education. Other technology partners may gain product development, institutional access, or future customers. Student presentations are not simply assignments anymore. They may also become material through which a system is trained, refined, researched, promoted, and potentially expanded.

Were students clearly told about that larger purpose? Could they opt out? Are their presentations or evaluations contributing to research or product development? Who owns the resulting data? Who benefits if the system is adopted more broadly? 

I’m not interested in “gotcha” questions, and I’m not accusing Kalodimos of acting outside university policy. However, I am also interested in ensuring that we are holding ourselves to a high bar of integrity when it comes to students. OSU should be able to explain what students were told, what data were retained, whether participation was voluntary, how the system was validated, and how students can challenge an inaccurate evaluation. As faculty, all of us should also be able to explain why our use of AI in assessing students is ethically different from the uses we restrict or discourage in student work. That difference may exist, but it needs to be articulated and supported rather than assumed.

I do not think the answer is to stop experimenting with AI. I use it in my own teaching, and I expect its role in education to grow. The answer is to apply the same scrutiny, transparency, and accountability to institutional AI use that we expect from students. In effect, we must become even more rigorous in our ethical inquiries. When student use of AI is treated as suspicious while faculty and institutional use is celebrated, we should pause. Ethical consistency requires us to ask who has power, who receives the benefit, who carries the risk, and whether everyone has a meaningful voice in this process.

Disclosure: AI tools were used in the editing process of this work to correct spelling and grammar errors as well as to improve the clarity of the written work. The thumbnail image is AI gerneated. All ideas were human-generated.