EXPOSED
The daily output of a sociologist is text and data: literature reviews, survey instrument drafts, regression and coding runs, interview transcript analysis, grant narratives, journal manuscripts — the exact work current models do at usable draft quality. What resists is the front end (choosing which social question matters, designing a sample that survives peer review) and the field end (gaining access to communities, sitting with respondents, reading a setting in person). No licensure protects the title, and much of the employment is grant- and university-funded, so budget pressure hits before AI does.
Mixed — a routine tier and a judgment tier. An 8 reflects the split in a sociologist's week: coding open-ended responses, running Stata/R models, writing literature reviews and journal drafts are now draft-quality machine output, while framing a researchable question about, say, neighborhood effects and defending a sampling design through IRB and reviewer 2 still requires a person who knows the discipline's arguments — that's mixed, not resistant.
Some physical or field component. The 6 comes from fieldwork that some sociologists actually do — ethnographic observation in a school or shelter, in-person interviews, recruiting respondents at a site — but most days are laptop, dataset, and manuscript, so this sits just above desk-only rather than in the field-truck range.
No licence, no signature requirement. No state licenses the title 'sociologist'; a PhD is a hiring credential, not a legal gate, and nothing in your survey report requires your signature the way a PE stamp or clinical license does — hence a 2 rather than zero, since IRB approval and university appointment still put a named human on the protocol.
Some relationship component. A 7 covers the relationships that matter — repeat access to a community that lets you back in, a named PI a foundation program officer trusts with a grant — but the journal article is read and cited without regard to who wrote it, and survey respondents talk to the study, not to you.
Meaningful discretion. An 11 is right because you decide what counts as a case, where to cut a coding scheme, and whether a finding about disparity is robust enough to publish — real discretion with reputational stakes — but peer review, co-authors, and IRB spread the consequences, and no one's liberty or life turns on the call.
Has AI actually changed your work?