COOKED
The daily work — literature searches and annotated bibliographies, transcribing and coding interviews, cleaning survey datasets, running standard regressions in Stata or R, and drafting methods sections and tables — is exactly the text-and-spreadsheet work current models do at usable quality and near-zero marginal cost. What resists is the physical and interpersonal layer: recruiting and consenting human subjects, running in-person lab sessions or field surveys, and the IRB and data-integrity legwork a named person must actually do. There is no licensure and no signature requirement here, and the principal investigator — not the RA — owns the findings, so no accountability moat exists.
Nearly all of this fall was the 2020 shock. It has been climbing back since.
Median pay $47,510 → $61,990 +4.4% 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
+4.4%
Percentage only. The projection counts a different population from the 30,640 above — it includes self-employed workers, which for this occupation is most of them, so the two headcounts are not comparable.
Exposed, but growing
AI can already do a lot of these tasks, and the BLS still expects +4.4% more of these jobs by 2034. Demand for the output is growing faster than the work is being automated away — the mechanism BLS gives for software developers, and the combination people most often misread as an error.
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.
~5,200 openings a year on average, including replacing people who leave.
Data AnalystResearch AideGraduate AssistantResearch AssistantResearch AssociateResearch ScientistResearch TechnicianLaboratory AssistantResearch InterviewerResearch Data AnalystPolitical Affairs InternSocial Research AssistantClinical Research AssistantEconomic Research AssistantGraduate Research AssistantMarket Research InterviewerEconomist Research AssistantHistorian Research AssistantSociology Research AssistantBilingual Research InterviewerPostdoctoral Research AssociatePsychologist Research AssistantPolitical Science Research AssistantPostdoctoral Fellow (Postdoc Fellow)
Holding it up: task resistance . Weakest point: liability shield .
Core tasks are already automatable Coding open-ended survey responses to a codebook, running descriptives and OLS in Stata, formatting tables to APA, and chasing down citations are the tasks an LLM does in one pass — the 6 rather than 2 reflects the residual that isn't text: intercept surveys in the field, escorting participants through consent and debriefing in a lab session, and calling non-responders for a phone follow-up.
Some physical or field component Most weeks are a laptop and a shared drive, but the 6 accounts for real in-person duties — setting up eye-tracking or physiological equipment in a behavioral lab, running focus groups, door-knocking or mall-intercept recruitment, and handling paper consent forms and locked file cabinets for identifiable data.
No licence, no signature requirement There is no credential to hold: RA postings ask for a BA and Stata familiarity, IRB human-subjects training (CITI) is a two-hour online module anyone can pass, and the protocol is approved in the PI's name — the 1 rather than 0 is only because your CITI certificate is a documented condition of touching subject data.
Executes defined procedures on defined inputs You make choices daily — how to handle a straddling response, whether an outlier is a data-entry error, whether an interview segment fits code 3 or code 7 — but they run back to a codebook, a pre-registration, or the PI's decision by Friday's meeting, which is why this sits at 5 rather than in the discretion band.
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 (6/20 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 (1/20) is whether the law requires a licensed human to sign. Trust premium (4/20) is whether buyers specifically pay for a person. Judgment and accountability (5/20) 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 10 of this occupation's 22 points (45%).
Embodiment (6/20) 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.
Data Scientists EXPOSED
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 38/100 — EXPOSED.
Genuine two-tier structure: if coding, cleaning, and literature search collapse to model output, the residual is adjudicating ambiguous qualitative codes, detecting fraudulent survey respondents and bot panels (a live crisis for Prolific/MTurk data since 2023), and reconstructing what actually happened in the field. Watch for RA postings that specify data-quality auditing and respondent-fraud screening rather than coding.
Federal research-integrity and human-subjects rules (Common Rule 45 CFR 46 revisions, or NIH/NSF data-management policy) requiring a named human study-team member to attest to consent administration and to certify that AI tools were not used to generate or impute human-subject data — the way FDA 21 CFR Part 11 already requires named signers for clinical data. Journals (ICMJE, COPE) already bar AI authorship; extending that to a named human attestor of dataset provenance would attach a person to each file.
Funder or IRB requirements that consent, sensitive-population interviews, and field survey administration be conducted in person by a trained human — already the norm for prisoner, minor, and clinical populations under 45 CFR 46 Subparts B-D — and enforced against remote/synthetic-panel substitutes. If IRBs formally bar AI-mediated consent, the surviving RA job concentrates in the field-contact layer.
Replication-crisis infrastructure putting names on analytic decisions: preregistration platforms (OSF, AsPredicted) and journal policies requiring a named analyst to document each deviation from the preregistered plan, plus retraction-era demands that a specific person can reconstruct the analysis pipeline under challenge.
The limit. No realistic route to a trust premium — the buyer is a PI or funder who never sees the RA, and grant budgets reward cost, not human provenance. Even with every lever, the routine tier is gone and headcount falls; these raise the score of the surviving job, not the number of jobs.
| New York-Newark-Jersey City, NY-NJ | 2,610 | $65,330 +5% |
| Los Angeles-Long Beach-Anaheim, CA | 2,290 | $59,240 -4% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 2,170 | $65,330 +5% |
| Atlanta-Sandy Springs-Roswell, GA | 990 | $67,440 +9% |
| San Francisco-Oakland-Fremont, CA | 710 | $82,020 +32% |
| San Diego-Chula Vista-Carlsbad, CA | 670 | $66,520 +7% |
| Louisville/Jefferson County, KY-IN | 570 | $82,500 +33% |
| Phoenix-Mesa-Chandler, AZ | 570 | $64,290 +4% |
| Flagstaff, AZ | 40 | $86,020 +39% |
| Bowling Green, KY | 100 | $84,680 +37% |
| Louisville/Jefferson County, KY-IN | 570 | $82,500 +33% |
Lawrence Livermore National Laboratory
The Mercury News reports that Lawrence Livermore National Laboratory in California is using autonomous AI systems to run and expand scientific experimentation.
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.