EXPOSED
This is a catch-all bucket for college instructors outside the named disciplines — the modal worker builds syllabi, delivers lectures, grades written work, and advises students, often off the tenure track. Slide decks, lecture outlines, quiz banks, rubric-based grading, and feedback drafts are already competently generated by AI, and no license protects the role; what holds is live classroom presence, mentorship students actually pay tuition for, and accountability for what a cohort learns. The real threat is institutional: enrollment pressure plus AI-assisted course production shrinks section counts and consolidates adjunct lines faster than it eliminates the job outright.
Mixed — a routine tier and a judgment tier. Lecture prep, quiz banks, and rubric-scored essay feedback — the bulk of a 3/3 or 4/4 load — are already drafted competently by a model, but running a live seminar discussion, reading when a room has lost the thread, and calibrating a cohort's midterm results into next week's plan still require a person in the room, which is what keeps this at 10 rather than down in the automatable band.
Some physical or field component. Teaching bodies are in physical classrooms — proctoring exams, running a lab or studio section, holding in-person office hours, moving between campuses on adjunct schedules — but the environment is a controlled, scheduled room with predictable equipment, so this sits at 9 rather than in the field-work range where conditions are uncontrolled.
No licence, no signature requirement. No state license, board exam, or continuing-education requirement gates who teaches an interdisciplinary or general-studies course; the institution's accreditation and the department chair carry the credentialing weight, and adjunct appointments are frequently made on a master's degree plus a semester-by-semester contract, which is why this is a 3 and not a 0 — degree requirements exist, they just protect nobody's job.
The human relationship is the product. Students choose sections by instructor name, come back for letters of recommendation and thesis advising, and the mentorship a first-generation student gets in office hours is a substantial part of what tuition buys — that relational core puts this at 14, though it stops short of the high teens because large gen-ed sections and high adjunct turnover mean many students never form the bond at all.
Meaningful discretion. You decide what counts as evidence in a paper, whether a plagiarism case goes to the academic integrity board, whether a struggling student gets an incomplete or a failing grade that affects their aid eligibility — real discretion with consequences, but bounded by department-approved learning outcomes, appeal processes, and grade-distribution norms, which holds it at 10 rather than the ownership-of-ambiguity band.
Has AI actually changed your work?