On the essay in a time of GenAI

Synthesis: Corbin and colleagues argue that the essay's AI crisis is a category error before it is a technology problem: institutions grade a standardized product while the essay has historically been a process of trying out thought. Recovering that tradition shows what a generator cannot supply — the trace of a mind changing as it writes — and points to holistic, calibrated judgment rather than prohibition.

Conceptual argument in the history of ideas, with no empirical study of its own; findings it cites from other work are the authors' characterizations.

Key Findings

1. The crisis is a misalignment. The essay GenAI simulates is not the essay the tradition describes; assessment turned inquiry into a fixed-shape product, which the simulation exploits.
2. The assessed essay resembles a research report — thesis, evidence, restated conclusion — a genre for presenting findings, not for working out thought.
3. "Essay" names an attempt. From Montaigne onward it was unfinished by design, its value in the thinking the writing produced.
4. Three capacities depend on that form: working at the edge of the known, self-directed learning through the writer-text loop, and meditative rather than calculative thinking.
5. Prohibitions leave the mechanism intact. Bans and AI-use declarations change what students are told, not what the task asks.
6. The fix is structural: holistic judgment by disciplinary experts, calibrated across graders, supported by exemplars and dialogic feedback.

What the essay was for before it became a genre

The word descends from essayer — to try, to test — and the etymology describes a method, not an output. Montaigne's Essais of 1580 were attempts to weigh ideas: meandering, tolerant of contradiction, resistant to conclusions. "I do not portray being," he wrote, "I portray passing." Addison drew the line in 1712 between writing "with regularity and method" and writing that "run[s] out into the wildness of those compositions which go by the name of Essays." Systematic prose starts from a conclusion; essayistic prose lets understanding emerge in the writing.

The tradition is not only European: Chinese xiaopinwen and Japanese zuihitsu share what Claire de Obaldia calls the essayistic spirit. Bacon's "the word is late, but the thing is ancient" names the loss — not a format but a practice of thinking in public and in progress.

Adorno's account: the essay "thinks in fragments," refuses completeness, resists reducing particulars to pre-existing categories; its aspects "interweave as in a carpet." Lukács has it mediating private experience and larger questions, and Arendt's "enlarged thought" names the capacity to hold several perspectives. The essay is where a writer discovers a position rather than defends one already held.

What standardization turned it into, and why GenAI found it easy

The assessed essay is now thesis-driven and rubric-shaped, what Andrew calls an "elaborate game" of structured demonstration. Its success conditions can be stated in advance — exactly what a language model satisfies: a criteria sheet written before any student begins tells the writer — and the model — what a good answer looks like. Validity was already thin when a grader could not tell compliance from inquiry; GenAI made that impossible to ignore.

Predetermined criteria also misdescribe what Sadler calls divergent tasks: criteria fixed in advance cannot anticipate which features will matter in a piece that does not resemble its neighbors. Specifying them invites compliance: the task asks for exploration and rewards conformity. Bans, declarations and traffic-light rules leave this mechanism untouched; they regulate speech about the task rather than the task.

Three capacities that only the exploratory form trains

The first is thinking at the border, Merleau-Ponty's phrase for working where established knowledge gives out. A research report documents a journey completed; the essay invites writer and reader into it, practicing ambiguity without premature closure.

The second is self-regulated learning. "I have no more made my book than my book has made me": content and self-awareness are inseparable, unlike a reflective journal where reflection is added afterwards. The writer must monitor a changing argument and revise — metacognition enacted rather than reported.

The third is meditative thinking, Heidegger's contrast with calculative thought. Nearly every other format runs on calculative logic, working toward a predetermined outcome. Boyle and the Hartlib Circle chose essays in the 1650s because a "loose, open-ended, and collaborative form" suited piecemeal discovery.

What the proposal changes

The remedy is structural change to the assessment task, not a stricter rule around it. Holistic assessment, in the sense developed by Sadler and by Walton and Martin, replaces the checklist with a global judgment grounded in disciplinary expertise, justified afterwards by features of the work that warrant attention. Interpretation comes first, reasons follow; no rubric decides in advance what can count.

This is not impressionism. Reliability comes through intersubjective calibration — graders judging real work together, comparing decisions, articulating their reasoning — building standards from evaluative encounters and cultivating what Tai and colleagues call evaluative judgment. The pedagogy must match: exemplars showing the range of valid forms, students generating criteria from them, peer comparison, dialogue about why one piece is better.

Holistic judgment, they concede, addresses one side, cannot secure validity alone, and remains untested. What changes is the question — not whether the essay survives AI, but whether the task still asks for something process-oriented assessment can evidence and a generator cannot supply.

What this means for practice

• Instructors: Ask questions you do not already have an answer to, and require the working — notes, a revision trail, a note on what changed — so the text shows the process.
• Program leaders: Resist replacing essays with examinations by default; the cheaper substitute measures recall and leaves the capacities the essay trained with no home.
• Assessment designers: Pair holistic judgment with calibration sessions and written justifications against features of the work, and retire the criteria sheet that teaches compliance.
• Faculty developers: Run exemplar-based norming where staff judge real work together and argue about quality, then have students do the same with the same exemplars.

Limitations

• Conceptual throughout: the authors concede nobody knows which structural changes work until tried, and offer assessment literature rather than trial evidence.
• Scope is higher education and the humanities-style essay; generalization beyond it is asserted by analogy.
• Holistic assessment assumes expertise, grading time and calibration capacity many programs lack.
• Judgment without criteria can drift toward impressionism; calibration is the answer offered, but there is no procedure for detecting a shared blind spot.

Citation

Thomas Corbin, Jack Walton, Peter Bannister & Jean-Philippe Deranty (2026). On the essay in a time of GenAI. Educational Philosophy and Theory, 58(3), 198–210.