Research Article
“But it’s supposed to be hard work”: rethinking effort and legitimate academic work in the age of generative AI
Synthesis: Benau, a clinical psychology professor, argues that much of the unease about generative AI in academia is not a judgment about learning but a moral judgment about effort. Drawing on the Protestant Work Ethic and the meritocratic ideology that equates hard work with commitment and legitimate achievement, he proposes that AI "unbundles" effort, commitment, competence and value, which had long been assumed to be intertwined. His conclusion is deliberately uncomfortable for both sides: reducing the effort required to do academic work does not by itself make that work less legitimate, and yet struggle is not a virtue either. The boundary he settles on is ownership — legitimate AI assistance depends on whether a person can rightfully claim the output as their own, not on how many hours the tool saved.
Key Findings
- Effort carries moral weight in academic culture. Beliefs that hard work is intrinsically virtuous and earns achievement, inherited from the Protestant Work Ethic and reinforced by meritocratic ideology, shape how AI use is judged.
- That moralization is what makes routine use feel like cheating. Benau reports using AI to proofread manuscripts, troubleshoot statistics, improve slides and draft emails, and describes the discomfort as persistent even while arguing it is unwarranted.
- AI unbundles processes long assumed to be one thing. Effort, commitment, competence and value came bundled in academic work; a tool that reduces the first two while preserving the last two exposes an assumption rather than settling a question.
- The delegation test is inconsistent, not principled. Asking a colleague to troubleshoot statistics or hiring a copyeditor draws no moral objection, while the same assistance from a machine is treated as suspect.
- Expertise earns latitude that learners have not earned. Those with subject and procedural expertise may delegate more of their work to AI; the processes that build that expertise in learners remain essential.
- Legitimacy rests on responsibility, not labor. Responsibility for content begins and ends with the human presenting the output as their own, so the question is whether the human can validly claim it.
- Neither prohibition nor unregulated use follows. The essay rejects policing AI as firmly as it rejects permitting unmonitored use, and asks when, where, how and for what purposes the tool should be used instead.
The argument in brief
The essay's method is reframing rather than evidence. Benau starts from what AI use feels like — the discomfort of disclosure, the sense of cheating at tasks a colleague could have done — and asks what belief makes that feeling intelligible. The answer is an ethic that treats sustained effort as morally significant in itself, not merely instrumental to the product, and that treats visible ease as evidence of insufficient dedication.
His own case is the illustration: a mid-career professor using AI for mundane tasks that free attention for substantive work, who still feels the need to justify the practice. He notes that faculty can certainly use AI dishonestly, and that there comes a point where the tool does the work one was hired to do, but argues that concession does not answer the more interesting question of why the machine feels different from the colleague.
The historical parallel does the heavy lifting. Figures, slides and presentation software were once produced by specialists or delegated to administrators, and each shift was contested before becoming unremarkable. Benau uses that record to argue that the current discomfort is at least partly a response to novelty and to effort norms rather than to a change in the epistemic status of the work.
Where the essay draws its line
Two distinctions carry the argument. The first separates experts from learners: delegation is defensible for someone who already holds the expertise the task would build, and not for someone who does not, which is why an instructor may use AI to draft a syllabus while a student cannot use it in place of the learning the assignment exists to produce. The second separates process from product: evaluation in academia mixes the two, judging output while inferring dedication from the effort presumed behind it, and that mixture is what makes efficient work read as dishonest.
The practical boundary he lands on is ownership. A person must be able to rightfully and validly claim the resulting output, which makes disclosure, verification and genuine understanding conditions of legitimate use rather than bureaucratic requirements. It also implies that the same tool use can be legitimate for one person and not another, depending on what they can defend.
For assessment, this points away from detection and toward designs where students demonstrate what they can claim. For faculty, it points at the inconsistency in how we treat machine help versus colleague help, which is a question of institutional norms rather than individual virtue.
What this means for practice
- Instructors: examine the assignment's purpose before banning or permitting a tool. If the process is the learning, say so and assess the process; if the product is what matters, effort norms are doing work that the criteria should be doing.
- Instructors: teach students to claim their output. Disclosure plus the ability to explain and defend a submission is a defensible boundary, and it is teachable in a way that "don't use AI" is not.
- Faculty developers: surface the effort norm explicitly in AI workshops. The discomfort many faculty report is a moral response to a norm nobody has stated, and naming it makes the policy conversation possible.
- Administrators: apply one standard to machine and human assistance. If hiring a copyeditor or consulting a statistician is legitimate, the justification has to rest on ownership and disclosure rather than on the identity of the helper.
- Researchers: treat moral discomfort around AI use as data. Shame, guilt and imposture feelings are documented responses, and their presence shapes disclosure behavior and therefore what any AI-use study can measure.
Limitations
- The essay is a conceptual analysis with no data; its claims about moralization rest on the author's reasoning and on cited survey work, not on a study of academic norms.
- The evidence it draws on about moral emotions and disclosure is recent and largely self-report, and the historical parallels to slides and graphics are analogies rather than tests.
- The expert–learner distinction is asserted without an account of how expertise is certified or how much delegation a given level permits.
- The ownership criterion is stated at a level of abstraction that leaves hard cases — co-authored work, heavily edited AI drafts, tool-assisted analysis — without a decision procedure.
- The argument is situated in US higher education and in the author's own discipline; norms about effort in other systems and professional fields may differ.
Citation
Benau, E. M. (2026). “But it’s supposed to be hard work”: Rethinking effort and legitimate academic work in the age of generative AI. Frontiers in Education, 11.