Faculty Are Not Anti-AI. They Are Anti-Unbounded AI.

Nearly two-thirds of Harvard professors in the Faculty of Arts and Sciences now say generative AI has had a somewhat negative or very negative effect on their courses. One year ago, that figure was 42 percent. A 25-point swing in twelve months, among more than 460 faculty at an institution with no shortage of access to the best available tools, is not noise. It is a finding.

The reflex in the edtech industry is to read numbers like these as a training problem. Faculty simply need more professional development, better prompts, clearer guidance. That reading is convenient and mostly wrong. These professors are not confused about what the tools do. They watched them operate inside their own courses for two full academic years and reached a considered judgment.

What Actually Changed in Twelve Months

The details of the survey, reported by The Harvard Crimson and summarized by Inside Higher Ed, describe a faculty that is not disengaging from AI so much as fencing it in. Explicit classroom AI policies are now nearly universal at 96.5 percent, up from 90 percent. The share of professors who permit AI use without restriction fell from 8 percent to 4 percent. The share who prohibit it entirely rose from 20 percent to 25 percent.

Meanwhile, nearly nine in ten faculty reported receiving student work they knew or believed was AI-generated, while only 64 percent felt confident they could tell the difference. Referrals to the honor council moved barely at all, from 10 percent to 12 percent. Read those two numbers together and the picture is clear: faculty are seeing far more than they are able to act on.

A Verdict on a Product Category

Almost every AI experience a student brings into a college classroom today was designed for a different job. Consumer assistants are optimized to complete tasks efficiently. That is genuinely useful when the task is the point. In a course, the task is not the point. The struggle is the point. A tool that removes the struggle removes the instruction.

So faculty are responding to the product they have been handed. The mistake would be to conclude that every possible use of AI in a course carries the same cost. The evidence emerging this year points the other way. Research summarized by Brookings and reported by EdWeek over the summer found the strongest results where AI tutoring was bounded to the course material and paired with human oversight, and the weakest results where students were left alone with an open-ended assistant.

Detection Is the Wrong Arms Race

The gap between 90 percent of faculty seeing suspected AI work and 64 percent feeling able to identify it is where enormous institutional energy is currently being spent. It is largely wasted. Detection tools will not outrun generation tools, and the false-positive cost falls hardest on multilingual students and anyone whose prose does not sound like the majority.

The more durable move is to make the process visible instead of interrogating the product. That is an instructional design decision before it is a technology decision, and it is the design principle behind how we build:

Faculty Trust Has to Be Earned Per Course

There is an uncomfortable implication here for anyone selling AI into education. Institutional enthusiasm is not the same as faculty consent, and faculty consent cannot be purchased at the procurement level. A provost can sign a contract in a quarter. A professor decides in week three of the semester whether the tool made the course better, and that decision is the only one that shows up in a survey like this one.

That is why we deploy through the LMS via LTI, in hours rather than months, at the course level. It is also why we are careful about what we claim. We do not have outcome data proving bounded Socratic tutoring reverses the trend in that Harvard survey. We have a design thesis, a pedagogical tradition that predates the technology by a few thousand years, and a research pilot built to test it honestly with institutional partners.

Two-thirds of a Harvard faculty reaching a negative verdict on the current generation of AI tools should be read as useful information rather than as a threat. They are telling the industry, precisely and with data, which product it built wrong. The reasonable response is to build a different one and then prove it works, in their courses, on their terms. Our approach to that is documented on our AI platform page.

Read the Full Article

Read "2 in 3 Harvard Professors Say AI Has Negative Impact" on Inside Higher Ed

About Core Learning Exchange: We provide turnkey Career and Technical Education (CTE) solutions for grades 6-14, offering 450+ courses from 20+ providers aligned to state standards and industry certifications. Our AI platform uses proven Socratic methodology to develop critical thinking skills through personalized, adaptive learning—deployed in hours via LTI integration.

Related Posts