Role Definition
| Field | Value |
|---|---|
| Job Title | Linguist — Research |
| Seniority Level | Mid-Level |
| Primary Function | Studies language structure, acquisition, and use through scientific methods. Conducts fieldwork documenting endangered languages, designs and runs psycholinguistic experiments, analyses corpora using statistical and computational tools, develops formal models of grammar and meaning, and publishes research. Works at universities, government agencies (GCHQ, NSA — cryptolinguistics), tech companies (NLP teams), and research institutes. |
| What This Role Is NOT | NOT a translator or interpreter (SOC 27-3091 — language service delivery, scored Red). NOT an NLP engineer (building production ML pipelines). NOT a postsecondary linguistics teacher (SOC 25-1124 — teaching-dominant). NOT a Speech-Language Pathologist (SOC 29-1127 — clinical intervention). This is the academic/research scientist who studies language as a phenomenon. |
| Typical Experience | 5-10 years. PhD typical (Job Zone 5). Specialisation in a subfield: phonology, syntax, semantics, sociolinguistics, psycholinguistics, computational linguistics, or documentary/field linguistics. |
Seniority note: Entry-level research assistants (0-2 years) doing corpus annotation and data cleaning would score Red — heavily automatable execution work. Senior professors and principal investigators directing research programmes, securing multi-year grants, and shaping theoretical paradigms would score upper Yellow or borderline Green due to stronger goal-setting, accountability, and institutional relationships.
Protective Principles + AI Growth Correlation
| Principle | Score (0-3) | Rationale |
|---|---|---|
| Embodied Physicality | 0 | Primarily desk-based. Field linguists travel to remote communities but work in structured interpersonal settings, not unstructured physical environments. |
| Deep Interpersonal Connection | 1 | Field linguists build trust with speaker communities for language documentation — cultural sensitivity and rapport are essential. Psycholinguistic experimenters interact with participants. But most mid-level time is analytical. |
| Goal-Setting & Moral Judgment | 2 | Formulates research questions, selects theoretical frameworks (generative, functional, usage-based), designs experiments, interprets results within linguistic theory, and makes ethical decisions about endangered language documentation and community engagement. |
| Protective Total | 3/9 | |
| AI Growth Correlation | -1 | Weak negative. AI/NLP advances subsume some computational linguistics functions into ML engineering. LLMs perform tasks (parsing, generation, translation) that linguists previously modelled theoretically, reducing demand for purely computational linguistics research while increasing demand for AI-evaluation roles. |
Quick screen result: Moderate protection (3/9) with weak negative AI correlation suggests Yellow Zone — a research-heavy knowledge role with meaningful human judgment in design and interpretation but significant AI exposure in execution tasks.
Task Decomposition (Agentic AI Scoring)
| Task | Time % | Score (1-5) | Weighted | Aug/Disp | Rationale |
|---|---|---|---|---|---|
| Research design & hypothesis formation | 20% | 2 | 0.40 | AUG | Formulating novel research questions about language structure, acquisition, or change. Selecting theoretical frameworks and designing studies. AI cannot originate linguistic hypotheses grounded in fieldwork observation and theoretical debate. |
| Fieldwork & language documentation | 15% | 2 | 0.30 | AUG | Documenting endangered languages in remote communities — elicitation, transcription, cultural context. Requires human presence, speaker trust, and cultural sensitivity. AI transcribes recordings but cannot conduct elicitation sessions or interpret pragmatic nuance. |
| Data analysis — phonological/syntactic/corpus | 15% | 3 | 0.45 | AUG | Statistical analysis of corpora, phonetic measurements (Praat), syntactic parsing, typological comparisons. AI handles routine analysis faster but interpreting results within linguistic theory — understanding why a pattern exists — requires human expertise. |
| Psycholinguistic experiment design & execution | 10% | 2 | 0.20 | AUG | Designing reaction-time, eye-tracking, and ERP experiments to test processing models. Running participants, controlling for confounds. AI assists with stimulus generation and data analysis but cannot design novel experimental paradigms. |
| Computational modelling & NLP development | 10% | 4 | 0.40 | DISP | Building formal grammars, training language models, developing parsing algorithms. LLMs and transformer architectures now perform many tasks computational linguists previously built bespoke models for. Core NLP engineering increasingly done by ML engineers, not linguists. |
| Report writing & publication drafting | 15% | 4 | 0.60 | DISP | Drafting research papers, grant proposals, and technical reports. AI generates first drafts, structures arguments, and formats citations. Academic peer-reviewed writing still requires human voice and argumentation, but the drafting stage is displaced. |
| Literature review & secondary research | 5% | 5 | 0.25 | DISP | Semantic Scholar, Elicit, and Consensus synthesise existing linguistics literature faster and more comprehensively than manual search. |
| Teaching, mentoring & public engagement | 5% | 1 | 0.05 | NOT | Supervising students, public lectures, community outreach on language preservation. Requires human presence and pedagogical judgment. |
| Stakeholder consultation & advisory | 5% | 2 | 0.10 | AUG | Advising tech companies on NLP, consulting with indigenous communities on language revitalisation, providing expert testimony on forensic linguistics. Requires trust and domain expertise. |
| Total | 100% | 2.75 |
Task Resistance Score: 6.00 - 2.75 = 3.25/5.0
Displacement/Augmentation split: 30% displacement, 65% augmentation, 5% not involved.
Reinstatement check (Acemoglu): AI creates new tasks — evaluating LLM linguistic competence (syntactic generalization, semantic understanding), auditing NLP systems for linguistic bias, designing benchmarks for AI language models, and bridging theoretical linguistics with ML research. These reinstatement tasks are real but absorbed by existing researchers rather than creating net new positions.
Evidence Score
| Dimension | Score (-2 to 2) | Evidence |
|---|---|---|
| Job Posting Trends | 0 | BLS does not track linguists separately — they fall under Social Scientists All Other (19-3099) or are split across postsecondary teachers, computer/information research scientists, and interpreters/translators. Academic linguistics postings stable for 2026 starts (tenure-track in syntax+AI, computational linguistics). No surge, no collapse. Tiny occupation (~3,000-4,000 research linguists estimated). |
| Company Actions | 0 | No AI-driven layoffs of linguists. Tech companies (Google, Meta, Apple, Amazon) restructuring NLP teams — shifting from linguist-led to ML-engineer-led, but retaining linguists for data quality, evaluation, and low-resource language work. Academic departments not closing linguistics programmes but hiring fewer tenure lines. |
| Wage Trends | 0 | Median ~$98,860 for broad linguist category (BLS). Academic salaries tracking inflation. Industry computational linguists competitive ($120K-$180K) but title increasingly replaced by "NLP Scientist" or "Applied Scientist." No real-terms decline or surge. |
| AI Tool Maturity | -1 | LLMs (GPT-4, Claude, Gemini) perform parsing, translation, text generation, and corpus analysis that previously required bespoke computational linguistics models. Praat, ELAN, and corpus tools augmented by AI. Tools augment ~45% of task time and displace ~30%. Not eliminating positions but compressing person-hours per project. |
| Expert Consensus | 0 | Mixed. LSA (Linguistic Society of America) acknowledges AI transformation. Some argue LLMs make theoretical linguistics more relevant (testing predictions about language universals); others worry computational linguistics is being absorbed into ML. Consensus: the field is transforming, not disappearing. |
| Total | -1 |
Barrier Assessment
Reframed question: What prevents AI execution even when programmatically possible?
| Barrier | Score (0-2) | Rationale |
|---|---|---|
| Regulatory/Licensing | 1 | No individual licensing, but IRB oversight mandates human PIs for human subjects research (psycholinguistic experiments, fieldwork with speaker communities). Federal research grants (NSF, NIH, NEH) require named human investigators. |
| Physical Presence | 0 | Primarily desk-based. Fieldwork requires travel but in structured interpersonal settings, not unstructured physical environments. |
| Union/Collective Bargaining | 1 | Academic linguists often covered by faculty unions (AAUP, AFT). Collective bargaining agreements protect positions in universities, slowing AI-driven restructuring. Government linguists (intelligence agencies) have civil service protections. |
| Liability/Accountability | 1 | Research integrity — personal accountability for methodology, data handling, and ethical conduct with speaker communities. IRB violations and research misconduct attach to named individuals. Endangered language documentation carries ethical obligations to communities. |
| Cultural/Ethical | 0 | No strong cultural resistance to AI in linguistics research. Speaker communities may resist AI processing of sacred or sensitive linguistic data, but this is niche. |
| Total | 3/10 |
AI Growth Correlation Check
Confirmed at -1 (weak negative). AI adoption has a paradoxical effect on linguistics: LLMs make some computational linguistics functions redundant (bespoke parsers, rule-based NLP systems), but simultaneously create demand for linguists who can evaluate, benchmark, and interpret AI language capabilities. Net effect is weak negative because the displacement of computational linguistics roles into ML engineering slightly outweighs the creation of AI-evaluation roles. The field is not powered by AI growth (unlike AI Security) nor independent of it (unlike nursing) — it sits in the uncomfortable middle where AI both enables and partially subsumes the work.
JobZone Composite Score (AIJRI)
| Input | Value |
|---|---|
| Task Resistance Score | 3.25/5.0 |
| Evidence Modifier | 1.0 + (-1 x 0.04) = 0.96 |
| Barrier Modifier | 1.0 + (3 x 0.02) = 1.06 |
| Growth Modifier | 1.0 + (-1 x 0.05) = 0.95 |
Raw: 3.25 x 0.96 x 1.06 x 0.95 = 3.1418
JobZone Score: (3.1418 - 0.54) / 7.93 x 100 = 32.8/100
Zone: YELLOW (Green >=48, Yellow 25-47, Red <25)
Sub-Label Determination
| Metric | Value |
|---|---|
| % of task time scoring 3+ | 45% |
| AI Growth Correlation | -1 |
| Sub-label | Yellow (Urgent) — AIJRI 25-47 AND 45% >= 40% threshold |
Assessor override: None — formula score accepted. At 32.8, the score sits mid-Yellow between Historian (30.7) and Sociologist (36.3). The linguist scores lower than the sociologist due to weaker barriers (3/10 vs 4/10) and the weak negative growth correlation (-1 vs 0) reflecting computational linguistics being partially absorbed into ML engineering. Scores higher than the historian due to less negative evidence (-1 vs -3). Well-calibrated against the social science research cluster.
Assessor Commentary
Score vs Reality Check
The Yellow (Urgent) label at 32.8 is honest. Linguistics research occupies a unique position — it is the science most directly affected by LLMs because language IS the medium AI has learned to process. The 45% of task time at score 3+ reflects genuine compression: corpus analysis, computational modelling, report writing, and literature review are all being accelerated or displaced by the same AI systems that linguists study. The core intellectual work (research design, fieldwork, theoretical framing, experiment design) remains human-led, but the execution tail is shrinking fast.
What the Numbers Don't Capture
- Bimodal subfield divergence: Field/documentary linguists documenting endangered languages in remote communities are functionally Green — their work requires embodied presence, cultural trust, and cannot be replicated by AI. Computational linguists building NLP models are being absorbed into ML engineering. The average score masks this split.
- Title rotation: Many linguists now work under titles like "NLP Scientist," "Applied Scientist," "Language Data Specialist," or "AI Evaluation Specialist" — roles growing under tech company umbrellas even as "Linguist" as an academic title stagnates.
- Micro-profession vulnerability: With ~3,000-4,000 research linguists in the US, even small productivity gains (10-20% fewer person-hours per project) visibly reduce headcount without formal layoffs.
- LLM paradox: LLMs are simultaneously the biggest threat and the biggest opportunity for linguists. They threaten computational linguistics roles but create new demand for evaluating whether AI actually understands language or merely predicts it.
Who Should Worry (and Who Shouldn't)
Field linguists who spend most of their time in remote communities documenting endangered languages — working with last speakers of dying languages, building trust, conducting elicitation sessions — are more secure than the 32.8 label suggests. This work is irreducibly human and culturally vital. Computational linguists whose primary output is NLP models, parsing algorithms, or corpus analysis tools should worry most — LLMs and ML engineers are absorbing this work. Psycholinguists running experiments retain protection through experimental design expertise but face compression in data analysis. The single biggest factor separating the safe version from the at-risk version is whether your value comes from original fieldwork and theoretical insight or from computational tasks that ML engineers can now perform.
What This Means
The role in 2028: The surviving research linguist uses AI to process corpora in hours instead of months, generates draft analyses with AI agents, and leverages LLMs to test theoretical predictions about language universals at unprecedented scale. But the core — formulating novel linguistic theories, documenting endangered languages through fieldwork, designing psycholinguistic experiments, and interpreting what AI language capabilities tell us about human cognition — remains human. The field will be smaller, more productive per capita, and increasingly split between field/theoretical linguists and those who pivot into AI evaluation and NLP science roles.
Survival strategy:
- Lean into fieldwork and theoretical depth — endangered language documentation, psycholinguistic experimentation, and theoretical innovation are the hardest tasks for AI to automate and the most valued in both academia and industry
- Master AI evaluation and benchmarking — become the linguist who can assess whether LLMs actually generalise syntactic rules, understand semantics, or merely memorise patterns; this is growing demand at every major AI lab
- Bridge linguistics and AI — the "linguist who codes" commanding both theoretical depth and ML fluency is the growth profile; pure theory without computational skills narrows options
Where to look next. If you are considering a career shift, these Green Zone roles share transferable skills with linguistics research:
- Computer and Information Research Scientist (Mid-to-Senior) (AIJRI 57.5) — your analytical methods, experimental design, and computational skills transfer directly to AI/ML research; 20% BLS growth
- AI Auditor (Mid) (AIJRI 64.5) — systematic evaluation methodology, bias detection, and linguistic analysis skills transfer to assessing AI systems for fairness and accuracy
- Speech-Language Pathologist (Mid) (AIJRI ~60) — phonetics, phonology, and language acquisition knowledge transfer to clinical application; strong demand and licensing barriers
Browse all scored roles at jobzonerisk.com to find the right fit for your skills and interests.
Timeline: 3-5 years. LLMs are already compressing computational linguistics and corpus analysis workflows. Field linguistics and theoretical research have a longer runway but the overall profession is shrinking through productivity gains rather than mass displacement.
