EXPOSED
The document-production half of this job — syllabi, lecture slides, reading summaries, exam banks, rubric-based grading of essays, literature reviews — is already within reach of current models, and enrollment in sociology majors is shrinking, which matters more than automation for headcount. What persists is live seminar facilitation on contested topics (race, class, deviance) where students specifically want a human in the room, mentoring and thesis supervision, and original empirical research with IRB accountability. The modal worker here is increasingly a contingent/adjunct instructor with no licensure shield and weak institutional protection, which is the real vulnerability.
Mixed — a routine tier and a judgment tier. Roughly half the week — building syllabi, slide decks, discussion prompts, multiple-choice item banks, marking 120 intro-level short answers against a rubric — is text-in/text-out work a model does credibly now, which is what pins this at 9 rather than mid-teens; the part that holds is running a live seminar where a student says something racist or wounded and you have twenty seconds to decide how to use it, plus supervising honors theses and IRB-approved data collection.
Some physical or field component. The physical component is being bodily present in a room at a fixed hour — proctoring, office hours, the field-methods course that sends students out to observe a courthouse or a bus depot — but none of it involves uncontrolled environments or manual skill.
No licence, no signature requirement. No state licence, no board, no certification exam gates who teaches Intro to Sociology; the PhD is a hiring convention, and for the adjunct teaching three sections on a semester contract even that is negotiable, so nothing legally requires a specific credentialed human to sign off on the course.
The human relationship is the product. Students choose sections by instructor and disclose things in your office — family incarceration, immigration status, sexual assault, the reason they missed three weeks — that they will not type into a chatbot, and the letters of recommendation and grad-school advising that follow are worth nothing unless a named person who knows them writes them.
Meaningful discretion. You decide when a discussion of policing or gender has crossed from productive discomfort into harm, whether a plagiarism case goes to the academic integrity office, how to grade an argument you find politically wrong but methodologically sound, and whether an undergraduate research design touching vulnerable populations is defensible to the IRB — real calls with consequences, but bounded by department curricula, appeals processes, and Title IX reporting rules that take the hardest decisions out of your hands.
Has AI actually changed your work?