EXPOSED
Education professors split their week between lecturing on pedagogy, grading reflective papers and lesson plans, advising teacher candidates, observing student teachers in K-12 classrooms, and writing accreditation reports — and AI already drafts syllabi, rubrics, discussion prompts, feedback on lesson plans, and literature reviews at usable quality. What holds is the embodied part: sitting in the back of a third-grade classroom watching a candidate manage 25 kids, then coaching them on what went wrong, plus the mentorship relationship districts and licensure boards rely on. The modal worker here is contingent or non-tenure-track faculty, so institutional cost pressure — larger sections, more asynchronous delivery — is a bigger near-term threat than the model itself.
Mixed — a routine tier and a judgment tier. The lecture-plus-rubric-plus-literature-review half of the week is already drafted competently by a model, but the other half — clinical placement supervision, live observation cycles with pre-conference and post-conference debriefs, and CAEP/state program-review evidence that must be tied to actual candidate performance — has no digital substitute yet, which is what puts this at 11 rather than down at 5 with lecture-only humanities faculty.
Some physical or field component. Field supervision means driving to elementary and secondary schools several times a semester, sitting in a classroom you don't control, and reading a candidate's proximity, voice, and transition management in real time — real physical presence, but it's a few hours a week in K-12 buildings rather than the daily uncontrolled-site work of a nurse or electrician, hence 10 and not 16.
No licence, no signature requirement. No state licence gates the job of teaching education courses; a doctorate and, in some programs, a lapsed K-12 credential are hiring preferences, and when a program loses accreditation or a graduate fails licensure, the consequence lands on the institution and the dean, not on your personal certificate — the 3 reflects that even the state approval process names the program, not you.
The human relationship is the product. Teacher candidates choose and stay in programs on the strength of a named mentor who watched them teach, wrote the recommendation the district reads, and takes their call in year one of teaching — districts recruit from specific supervisors they trust, and that referral network is the part of the job no one can transfer or automate, which is why this sits at 15.
Meaningful discretion. Deciding whether a struggling candidate is remediated, removed from placement, or recommended for licensure is a genuinely contestable call with a child's future classroom on the other side of it, but you make it inside program rubrics, disposition checklists, and state approval standards with a committee or chair signing off, which caps it at 11 rather than the high-teens of someone who owns the decision alone.
Has AI actually changed your work?