The Great AI Double Standard: Why Higher Ed Demands Tech Fluency While Questioning Its Use
Poets&Quants ·

Business schools want graduates who can work confidently with artificial intelligence. Applicants and researchers, meanwhile, must work out how much AI assistance is acceptable when presenting themselves or their ideas. The resulting ... The post The Great AI Double Standard: Why Higher Ed Demands Tech Fluency While Questioning Its Use appeared first on Poets&Quants .
Business schools want graduates who can work confidently with artificial intelligence. Applicants and researchers, meanwhile, must work out how much AI assistance is acceptable when presenting themselves or their ideas.
The resulting question is awkward but legitimate: when does using a valuable professional tool become an unacceptable shortcut?
Harvard Business School Executive Education describes AI capabilities as essential for business leaders seeking to remain competitive. Caltech’s admissions guidelines permit some AI assistance but prohibit other uses, including generating essay outlines and translating an essay written in another language. Nature Portfolio takes a different approach, explicitly stating that AI use is governed by risk level rather than tool type.
These policies do not prove that higher education is uniformly hostile to AI. They reveal something more interesting: institutions are drawing different boundaries around assistance, authorship, and independent judgment.
The challenge is not to eliminate those boundaries. It is to make them understandable, defensible, and fair.
Consider Caltech’s published guidelines for Fall 2027 applicants. They permit AI-assisted grammar and spelling review of completed essays, brainstorming questions, and research about the application process. They classify AI-generated outlines and drafts, replacement of an applicant’s voice, and translation of an essay written in another language as unethical uses. Failure to comply may result in admission being rescinded.
This is not a blanket prohibition. It is an explicit attempt to distinguish assistance from substitution.
There is a legitimate reason to draw that distinction. An application essay is not merely a document containing accurate information. It may also be intended to reveal how an applicant reflects, communicates, and makes sense of experience. A polished account can be factually correct while still failing to demonstrate the individual capability an institution wants to assess.
But that explanation creates further questions. Why is one form of language assistance acceptable while another is not? What exactly must remain the applicant’s independent contribution? How should a multilingual applicant distinguish acceptable editing from prohibited rewriting?
Applicants should not have to infer these answers from slogans about authenticity. Institutions should explain the capability being assessed and show, through concrete examples, where assistance becomes substitution.
Clear rules are only half the problem. The other half is enforcement.
In August 2023, Vanderbilt announced that it was disabling Turnitin’s AI detector “for the foreseeable future.” Its explanation raised concerns about false positives, transparency, potential bias against non-native English writers, and privacy. The university emphasized clearer expectations, assessment design, and conversations with students instead.
That announcement matters because it came from an institution responsible for protecting academic standards, not from an advocate arguing that standards should disappear.
Research provides a reason for caution. A 2023 study examining seven detectors reported an average false-positive rate of 61.22% on a sample of 91 human-written TOEFL essays. The findings raised serious questions about how the tested systems treated non-native English writing.
The limits matter as much as the headline. This was a particular sample, tested against particular detectors at a particular time. It does not establish the performance of every product available today. Nor does it demonstrate that every international applicant faces the same risk.
It does establish that confident-looking detection results can be wrong—and that errors may not be evenly distributed.
An institutional decision should therefore distinguish a signal that warrants inquiry from evidence that establishes a violation. Those are not the same thing.
Institutions often present disclosure as the straightforward solution: explain how AI was used, accept responsibility, and move on.
That is a sensible starting point. It may not be the end of the problem.
A 2025 study published in Proceedings of the National Academy of Sciences , “Evidence of a social evaluation penalty for using AI,” reported four preregistered experiments involving 4,439 participants. The research found negative evaluations associated with AI use, including perceptions of competence and motivation, in workplace-related settings.
These experiments did not establish that MBA admissions officers or journal editors penalize compliant disclosure. That distinction must remain explicit.
Nevertheless, the findings raise an important policy question: could institutions formally permit assistance while individual evaluators informally discount the people who acknowledge using it?
The appropriate response is not to assume that such a penalty exists everywhere. It is to investigate whether it occurs.
If a school permits grammar assistance, evaluators should receive clear guidance on how to interpret that disclosure. If a journal permits specified writing support, authors should understand what disclosure means—and what it does not mean—for the assessment of their contribution.
Transparency works best when its consequences are predictable.
Nature Portfolio’s central AI policy provides a useful reference point:
“Use of AI is governed according to risk level, not by tool type.”
Its framework permits lower-risk assistance such as language polishing and formatting. More consequential assistance requires human oversight, verification, and disclosure. Uses that replace accountable scholarly judgment, fabricate material, or compromise confidentiality are prohibited. The framework addresses authors, reviewers, and editors, with responsibilities appropriate to their roles.
That approach moves the debate beyond the unhelpful question, “Was AI involved?”
The more useful questions are: What did it do? What could go wrong? Who verified the result? Who remains accountable?
Publishers also use AI-assisted integrity tools themselves. Springer Nature documents tools designed to identify problems including AI-generated nonsense text, image concerns, and irrelevant references.
This is not automatically hypocrisy. Screening a manuscript for fabricated material is different from generating its scientific conclusions. Institutional use and author use must be compared by purpose, risk, and responsibility—not merely by whether both involve AI.
The same principle should apply to business education. Using AI to improve readability is not equivalent to outsourcing the reasoning an assignment was designed to assess. Equally, producing fluent prose is not evidence that the underlying reasoning is sound.
Business schools and journals do not need identical AI policies. They do need policies that people can understand and institutions can enforce responsibly.
Every assessed task should identify the contribution being evaluated: personal reflection, independent reasoning, research design, interpretation, or communication.
Restrictions become easier to defend when their educational or scholarly purpose is explicit.
“Use AI responsibly” is not an operational rule.
Policies should distinguish proofreading, translation, brainstorming, outlining, drafting, analysis, and decision-making. Where a category is prohibited, explain why. Caltech’s itemized admissions guidance illustrates the value of specificity, even where readers may question particular boundaries.
A flag should not become a verdict. Where concerns arise, institutions should examine the relevant evidence, allow a response, and apply human judgment.
Vanderbilt’s published reasoning offers a practical example of why detection-led enforcement deserves scrutiny.
Disclosure requirements should reflect the significance of the assistance.
Institutions should state what must be reported, how it should be reported, and how permitted assistance will be evaluated. A useful disclosure describes the tool’s role and the human verification performed, not simply the product name.
AI policies should address what information may be entered into external systems, who is responsible for checking outputs, and which decisions cannot be delegated.
Nature Portfolio’s risk-based framework demonstrates why authorship, confidentiality, and scholarly judgment belong in the same conversation.
Institutions should examine whether their procedures generate avoidable errors or unequal outcomes. Differences across groups warrant investigation, not automatic conclusions about either misconduct or discrimination.
An appeals process should give people a meaningful opportunity to respond to an allegation of improper AI use.
The strongest argument for reform is not that AI assistance should always be acceptable. Some tasks genuinely require independent performance. Some uses create unacceptable risks. Fabricated evidence does not become legitimate because a machine produced it.
The argument is that restrictions should be connected to the purpose of the task, supported by reliable evidence, and enforced proportionately.
Business schools teach managers to distinguish useful signals from misleading proxies. Their own AI policies should meet the same standard.
The question is no longer simply whether a person used AI. It is whether that person made the contribution being assessed, followed the stated rules, verified the work, and accepted responsibility for it.
AI fluency and academic integrity are not opposing goals. Good policy should make it possible to demonstrate both.
Dr. Hemachandran K is Vice Dean of the School of Business and Director of the AI Research Centre at Woxsen University in Hyderabad, India. Dr. Raul Villamarin Rodriguez is Vice President of Woxsen. The views expressed are those of the authors and do not necessarily reflect the official policy or position of Woxsen University or its partners. The views expressed are those of the authors and do not necessarily reflect the official positions of Woxsen University or any institution discussed. This commentary distinguishes published policies and research findings from the authors’ interpretations and recommendations. Examples are included to examine policy design, not to allege misconduct by any individual or institution. Institutional policies may change; readers should consult the applicable official guidance.
The post The Great AI Double Standard: Why Higher Ed Demands Tech Fluency While Questioning Its Use appeared first on Poets&Quants .