EXPOSED
The text-production half of this job — lecture notes, reading guides, discussion prompts, syllabi, essay feedback, and much scholarly summarizing — is exactly what language models do well, and students are already using them to write the papers you grade. What survives is live Socratic argument in a room, reading a student's confusion in real time, and being a trusted adult who talks with 19-year-olds about death, faith, and what a good life is. The bigger near-term threat is not AI but humanities enrollment decline and adjunctification, which has already made the modal worker in this title contingent and low-paid.
Mixed — a routine tier and a judgment tier. Roughly half the week — building a Kant seminar's reading schedule, writing prompts on the trolley problem, summarizing secondary literature, marking 60 papers against a rubric — is text-in/text-out that a model does at speed, while the part that holds is the unscripted eighty minutes where you hear a student misread Nietzsche and reroute the whole discussion, plus dissertation supervision and departmental judgment about what counts as a philosophy major; that mix of clearly-exposed and clearly-durable is what puts it at 11 rather than in the resistant band.
Some physical or field component. The physical demand is standing in a classroom, writing on a board, holding office hours, and moving between campus buildings — nothing that requires manipulating equipment or working in conditions a remote instructor couldn't reproduce, which is why this sits well below the field-and-tool occupations rather than near them.
No licence, no signature requirement. No state licence, no board exam, no certification body stands between you and the lectern — a PhD (often not even a completed one, for adjuncts) is a hiring credential, not a legal monopoly on teaching Aquinas, and nobody sues the instructor for a bad reading of Aristotle; the 3 reflects that even accreditation requirements bind the institution, not you personally.
The human relationship is the product. Students take Intro to Ethics and stay for the person: they come to office hours to talk about a parent's death or leaving their church, they ask you for recommendation letters because you know how they think, and a decade later they remember the professor, not the syllabus — a chatbot can produce the argument but cannot be the adult who witnessed them change their mind, and that irreplaceability is nearly the whole product at 15.
Meaningful discretion. You decide unilaterally whether a paper is plagiarized or AI-written, how hard to press a student defending a repugnant position in front of peers, what to do when a discussion of suicide turns personal, and where a grade lands on the line between B+ and A- with scholarship consequences — real discretion with real cost, but bounded by rubrics, appeals processes, department curricula, and Title IX referral paths, which keeps it at 11 rather than in the owns-the-call band.
Has AI actually changed your work?