Research Article
Generative AI in Humanities Education: A Conceptual Framework for Academic Integrity and Critical Digital Literacy
No primary data: a conceptual article built by theory synthesis, with no sample, instrument or fieldwork of its own.
Synthesis: Stornaiuolo and Elkins (2026) argue that the humanities' integrity problem and its literacy problem are the same problem, and that treating them separately produces the two failure modes now visible in the field: prohibition-and-detection on one side, uncritical adoption on the other. Their framework joins academic integrity and critical digital literacy around evaluative judgment — the student's capacity to judge quality in a discipline where the genres being assessed are exactly what generative AI now produces fluently.
Key Findings
- The genres are the problem. Essays, summaries, translations and commentaries are both the humanities' traditional assessment instruments and the model's strongest outputs, so the discipline meets AI at its core rather than its periphery.
- Integrity and literacy are one educational problem. The framework treats them as two sides of the same capacity rather than as a rule problem plus a skills problem.
- Evaluative judgment is the connective construct. The student's ability to judge the quality of an interpretation — not to detect a machine — is what both integrity and literacy require.
- Prohibition and detection fail on their own terms. A rule that depends on detection spends credibility the discipline does not have, and a ban does not teach the judgment the discipline exists to teach.
- Assessment design is where the framework lands. The practical consequence is redesigned tasks that require demonstrated judgment, not better policing of the old ones.
The framework's argument
The authors build the framework by synthesis across academic integrity research, assessment design, evaluative judgment and digital literacy, and they organize it around what students must be able to do rather than around what they must not do. In a humanities course that means judging an interpretation's warrant, situating a claim in a scholarly conversation, and recognizing the difference between fluent synthesis and argument.
The argument's strength is that it does not require the humanities to abandon its assessment genres. It requires the discipline to assess the judgment the genres were always proxies for, which is a harder task but a coherent one. A seminar paper that is graded on interpretive choices, defended in class, and revised after critique is evidence of judgment in a way a polished essay is not.
Why this matters for the humanities specifically
The generic AI-in-education literature is dominated by STEM examples: problem sets, code, worked solutions. The humanities' difficulty is different, because the output that AI produces is not a wrong answer to be checked but a plausible reading to be argued with. That makes detection useless as a quality signal and makes the discipline's own habits — close reading, sourcing, the seminar — the relevant machinery.
The framework also explains why humanities instructors have been the most likely to respond with prohibition. If your assessment is an essay and the model writes essays, the fastest defense is to remove the tool. The authors' counter is that this removes the practice the discipline needs: judging texts is now a skill exercised against machine output as well as human output, and students who never practice it are less prepared, not more protected.
What this means for practice
- Instructors: assess the judgment, not the artifact. Ask for interpretive choices, warrants and revisions rather than a final polished product.
- Assessment designers: build tasks where the student must evaluate a model-produced reading. That converts the tool from a threat into the object of study.
- Curriculum designers: integrate critical digital literacy into existing humanities methods teaching rather than as a separate module.
- Administrators: do not write integrity policy that depends on detection. The framework's premise is that the evidentiary route is closed.
Limitations
A conceptual framework with no empirical test — no sample, no instrument, no outcome data — so its claims are arguments rather than findings. The synthesis is selective rather than systematic, and the humanities focus means the framework's transfer to STEM assessment is untested. Evaluative judgment is a construct with its own measurement difficulties, and the paper does not operationalize it. The article appears in a new journal, which limits the peer-review record available to judge it against.
Citation
Stornaiuolo, A., & Elkins, K. (2026). Generative AI in Humanities Education: A Conceptual Framework for Academic Integrity and Critical Digital Literacy. International Journal of Pedagogy Across Humanities, 1(1).