COOKED
This catch-all category — demographers, criminologists, program evaluators, policy and behavioral researchers in agencies, think tanks, and consultancies — spends most hours on literature synthesis, survey instrument drafting, cleaning and coding datasets, running regressions in R/Stata/SPSS, and writing up findings in reports and briefs. Those are precisely the text-and-pattern tasks current models do at usable draft quality, and there is no license, stamp, or signature protecting the output. What persists is research design under ambiguity, primary data collection with human subjects, IRB-defensible methodology choices, and standing behind a politically contested finding in front of a board or legislature.
Headcount grew steadily across the period.
Median pay $83,330 → $101,110 -2.9% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
-1.7% 40,800 → 40,100 on the projections basis
Exposed, and shrinking
Both signals point the same way: the tasks are largely automatable and the BLS projects -1.7% by 2034. This is the case where the score and the forecast agree, and it is the one worth taking seriously.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~3,200 openings a year on average, including replacing people who leave.
PlannerLinguistDemographerEthnologistEtymologistPaleologistPhilologistGroup TesterMental TesterMetaphysicistPsychometristJury ConsultantProgram OfficerPsychometricianPsychotherapistTraffic AnalystTransit PlannerSocial ScientistFleet CoordinatorResearch DirectorHealth PsychologistScientific LinguistSocial PsychologistSports Psychologist
The BLS uses Social Scientists and Related Workers, All Other for work that doesn't fit any named occupation, so it covers roles that have little in common with each other. Two consequences worth knowing before you read anything below:
Holding it up: judgment & accountability . Weakest point: liability shield .
Mixed — a routine tier and a judgment tier The recurring deliverables here — coding a 40,000-record ACS or NIBRS extract, drafting Likert batteries, running fixed-effects models and writing the 30-page evaluation report — are draftable by current models, and the 8 rather than 4 reflects the residual that isn't: specifying identification strategy when the treatment wasn't randomized, negotiating data-use agreements with agencies, and defending a sampling frame to an IRB that will ask why you excluded a subpopulation.
Some physical or field component Most weeks are entirely laptop-and-VPN against restricted-use data enclaves, but the 5 covers the fieldwork tail — running focus groups, ride-alongs and site visits for a corrections or housing program evaluation, intercept surveys, and enumerator training that has to happen in the room.
No licence, no signature requirement There is no licensing board for demographers or criminologists; a bad population projection or misspecified recidivism model produces a corrected erratum and a lost contract, not a suspension, and the 2 rather than 0 exists only because IRB approval and FISMA/Census DRB disclosure review put a named human on the certification.
Meaningful discretion The 10 reflects calls that are genuinely yours — whether a null result gets reported as null, whether a subgroup n=61 is too thin to publish, how to describe a racial disparity in a policing dataset — but they land in a peer-reviewed or committee-vetted document with co-authors and a sponsor's review, not as a single unreviewable decision with someone's liberty or money attached.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (8/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (2/20) is whether the law requires a licensed human to sign. Trust premium (6/20) is whether buyers specifically pay for a person. Judgment and accountability (10/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 18 of this occupation's 31 points (58%).
Embodiment (5/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 46/100 — EXPOSED.
Genuine two-tier structure: if literature synthesis, coding, and regression scripting are absorbed, the residual job is research design under contested causal identification, IRB protocol construction, and primary human-subjects collection (cognitive interviews, hard-to-reach sampling frames, field experiments) that cannot be scraped. Also rises if journals/funders adopt AI-disclosure rules (already at NIH, which bars AI-drafted peer review, and at Nature/ICMJE) requiring human-attributable methodology sections
If federal evaluation policy tightens the named-evaluator requirement — e.g. OMB/Evidence Act implementation guidance or agency Learning Agendas requiring a named senior evaluator to personally attest that design, sampling, and limitations statements are their own professional judgment, as GAO has pushed on evidence-quality attestation — the role's remaining hours concentrate in defensible, attributable calls rather than drafting
Narrow routes only: if forensic/criminological expert testimony rules harden — Rule 702's 2023 amendment already puts the burden on the proponent to show reliability, and courts have begun excluding AI-assisted analysis whose method the witness cannot personally explain — the litigation-facing subset gains a de facto signature requirement. Some states also require licensed or credentialed statisticians for official demographic certifications (census challenges, redistricting deposition work)
If major foundations and government contracting vehicles add AI-provenance clauses — as some GSA and state RFPs have begun doing — requiring identified human researchers of record and barring undisclosed model-generated analysis, buyers are purchasing human authorship explicitly rather than incidentally
No plausible route beyond existing field-collection work; site-based interviewing and ethnography do not scale enough to move the score
The limit. Even with every lever, this SOC is capped in the mid-50s. It is a residual category with no licensing body, no credential registry, and no single professional association able to impose a signature requirement — the structural things that lift lawyers and engineers simply do not exist here. Gains accrue unevenly: the litigation-facing criminologist and the named federal evaluator can be protected; the think-tank policy writer and the contract data cleaner largely cannot.
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 6,220 | $138,250 +37% |
| New York-Newark-Jersey City, NY-NJ | 2,700 | $107,170 +6% |
| Baltimore-Columbia-Towson, MD | 1,070 | $124,590 +23% |
| Atlanta-Sandy Springs-Roswell, GA | 890 | $88,610 -12% |
| Minneapolis-St. Paul-Bloomington, MN-WI | 860 | $103,330 +2% |
| Seattle-Tacoma-Bellevue, WA | 800 | $105,140 +4% |
| Boston-Cambridge-Newton, MA-NH | 760 | $105,670 +5% |
| Los Angeles-Long Beach-Anaheim, CA | 710 | $100,550 -1% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 6,220 | $138,250 +37% |
| Harrisburg-Carlisle, PA | 100 | $135,390 +34% |
| New Haven, CT | 70 | $126,610 +25% |
Devex reports that UN agencies are adopting AI tools amid falling budgets and rising workload demands.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.