Political Scientists

19-3094.00
Median wage $142,080/yr5,540 employed (US)Rank #307 of 923 scored · top 33% by substitution

Study the origin, development, and operation of political systems. May study topics, such as public opinion, political decisionmaking, and ideology. May analyze the structure and operation of governments, as well as various political entities. May conduct public opinion surveys, analyze election results, or analyze public documents.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution33
Exposure28
Augmentation72

Substitution — the headline: capability discounted by cost, barriers and adoption.

Exposure — technical capability alone, regardless of whether anyone deploys it.

Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.

Tasks on the substitution scale

14 rated tasks, binned by substitution score.

Position among all scored occupations

Distribution of 923 occupation scores; the marker is this occupation.

Tasks with substitution ≥ 70

7%

Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.

Why this score

The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.

Task automatabilityw 35%27

panel mean rating 2.1/5 → substitution pressure 27/100

Technical feasibility todayw 20%29

panel mean rating 2.2/5 → substitution pressure 29/100

Cost vs. human wagew 15%37

panel mean rating 2.5/5 → substitution pressure 37/100

Adoption barriersw 20%inverted — strong barriers lower the score45

panel mean rating 3.2/5 (barrier strength) → substitution pressure 45/100

Sector adoption velocityw 10%35

panel mean rating 2.4/5 → substitution pressure 35/100

Task breakdown (14 tasks)

Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.

Maintain current knowledge of government policy decisions.

72

CI 46–97 · exposure 67 · augmentation 88 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Policy research organizations, government agencies, and think tanks are rapidly adopting AI-powered policy monitoring and briefing systems; adoption is well underway in the information and professional services sectors where political scientists work.
Sector adoption velocityclaude-sonnet-53/5Political science and policy research sectors show moderate AI adoption for information retrieval and summarization, with pilots more common than fully embedded production systems.
Augmentation potentialclaude-haiku-4-5-202510015/5AI transforms productivity by maintaining comprehensive, real-time policy awareness at scale, allowing human political scientists to focus on analysis and interpretation rather than information gathering, while the human remains in the loop for strategic judgment and validation.
Augmentation potentialclaude-sonnet-54/5AI substantially improves an analyst's ability to monitor, summarize, and cross-reference policy changes, greatly increasing research productivity while the scientist still evaluates significance and implications.
Task automatabilityclaude-haiku-4-5-202510015/5Current AI systems can continuously monitor, aggregate, and summarize government policy decisions from official sources, press releases, legislative databases, and news feeds with high efficiency, easily achieving >50% time savings versus manual tracking. This is a straightforward information retrieval and synthesis task well within the scope of large language models and web-crawling agents.
Task automatabilityclaude-sonnet-52/5Tracking and synthesizing government policy changes requires continuous monitoring and contextual judgment about relevance and significance, which AI can support but not fully replace end-to-end today.
Adoption barriersclaude-haiku-4-5-202510011/5There are no legal, regulatory, or organizational barriers preventing AI systems from monitoring and summarizing publicly available government policy information; it is pure information work with no licensing requirement or human signoff mandate.
Adoption barriersclaude-sonnet-52/5No licensing requirement blocks AI assistance in monitoring policy, though institutional trust in AI-sourced political analysis and risk of misinformation create moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510015/5The cost of continuous AI monitoring plus summarization is orders of magnitude cheaper than the loaded salary of a political scientist or policy analyst spending hours daily on manual research and reading.
Cost vs. human wageclaude-sonnet-53/5AI tools can cheaply scan and summarize large volumes of policy documents, but human verification and interpretive oversight are still needed, keeping costs roughly comparable to a partially-augmented human workflow.
Technical feasibility todayclaude-haiku-4-5-202510015/5Multiple deployed products (policy monitoring platforms, news aggregators, government tracking services, and LLM-based briefing systems) reliably perform policy surveillance and summarization at scale in production for government analysts, think tanks, and research organizations.
Technical feasibility todayclaude-sonnet-53/5News aggregation, summarization tools, and research assistants (e.g., LLM-based briefing tools) exist and are used to track policy developments, but they have material error rates in currency, sourcing accuracy, and nuance.

Collect, analyze, and interpret data, such as election results and public opinion surveys, reporting on findings, recommendations, and conclusions.

51

CI 46–55 · exposure 50 · augmentation 88 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Political science and policy research institutions are moderately digitized and adopt BI/analytical tools, but adoption remains cautious and pilot-heavy rather than production-wide. Academic publishing norms and funding structures slow replacement of human analysis, though data tools and automation are increasingly common in practice.
Sector adoption velocityclaude-sonnet-53/5Academic and policy research sectors are adopting AI tools for data analysis and drafting at a moderate pace, with pilots and partial integration but not yet widespread production-level replacement.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly assists political scientists through automated data cleaning, exploratory analysis, visualization, and statistical reporting, allowing them to focus on interpretation and theory. Large language models and visualization tools also help with report drafting and synthesis, meaningfully raising productivity while the scientist remains the critical decision-maker.
Augmentation potentialclaude-sonnet-55/5AI substantially augments this task by rapidly processing large datasets, generating statistical summaries, and drafting narrative reports, greatly increasing researcher productivity while the analyst retains interpretive control.
Task automatabilityclaude-haiku-4-5-202510013/5Data collection and basic statistical analysis of election results and surveys can be largely automated with current AI and tools (cleaning, descriptive statistics, visualization). However, interpretation requiring contextual understanding of political dynamics, causal inference, and nuanced conclusions still requires substantial human expertise, limiting full task automation to roughly 50% time savings.
Task automatabilityclaude-sonnet-53/5AI can automate data cleaning, statistical analysis, and drafting of interpretive summaries, but framing research questions, contextualizing findings within political theory, and validating conclusions still require significant human judgment.MB
Adoption barriersclaude-haiku-4-5-202510013/5There are moderate barriers: research credibility and institutional reputation depend on human expertise and judgment, peer-review expectations require authorship and accountability, and grant funding often requires named qualified researchers. However, no legal licensing requirement exists, and tools can be adopted incrementally without formal authorization.
Adoption barriersclaude-sonnet-52/5No licensing requirement mandates human-only interpretation of political data, though reputational risk and academic/policy credibility norms create moderate friction against fully automated conclusions.
Cost vs. human wageclaude-haiku-4-5-202510012/5While data processing tools are inexpensive, political scientists command substantial salaries ($60k–$100k+), and meaningful automation still requires data engineering, oversight, and validation costs. The cost of AI systems and integration likely approaches or exceeds the cost of delegating simpler analytical tasks to junior staff, with no clear order-of-magnitude savings.
Cost vs. human wageclaude-sonnet-53/5AI tools substantially cut time on data processing and drafting, but human oversight, validation, and specialized interpretation still add cost, keeping the ratio only moderately favorable rather than order-of-magnitude cheaper.
Technical feasibility todayclaude-haiku-4-5-202510013/5Deployed products exist for data collection (web scraping, APIs), statistical analysis (R/Python libraries, BI tools), and report generation, but they function reliably only within narrow technical scope. Human review of interpretation and conclusions is essential in production; no end-to-end system performs this task autonomously at scale in political science practice.
Technical feasibility todayclaude-sonnet-53/5Products like advanced LLMs and statistical/BI tools reliably handle data analysis and report drafting, but integrated end-to-end interpretation of survey/election data with domain nuance is not yet a mature turnkey product.

Write drafts of legislative proposals, and prepare speeches, correspondence, and policy papers for governmental use.

42

CI 30–54 · exposure 42 · augmentation 88 · importance 2.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption remains limited and cautious in government sectors. While some agencies experiment with AI drafting tools, the pace is slow due to risk aversion, regulatory caution, and the preference for human-authored policy. Production deployment at scale remains rare.
Sector adoption velocityclaude-sonnet-53/5Government and policy sectors are adopting generative AI for drafting tasks, but slower than finance or tech due to bureaucratic caution and procurement barriers.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can substantially assist by generating initial drafts, offering multiple language options, and accelerating ideation, allowing human policy writers to focus on refinement, legal precision, and strategic alignment. This augmentation is already demonstrable in policy organizations using LLMs for draft scaffolding.
Augmentation potentialclaude-sonnet-55/5AI is highly effective as a drafting assistant, generating first drafts of speeches, correspondence, and policy papers that political scientists can then refine and finalize.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can generate draft text quickly, legislative proposals and policy papers require domain expertise, legal precision, and alignment with specific political objectives that current AI cannot reliably handle end-to-end. Human review and revision of AI drafts typically demands substantial rework, falling short of the 50% time-saving threshold at equal quality.
Task automatabilityclaude-sonnet-53/5AI can draft speeches, correspondence, and policy papers competently, but legislative proposals require precise legal drafting, procedural knowledge, and political judgment that still need substantial human revision.'
Adoption barriersclaude-haiku-4-5-202510014/5Legislative and policy writing often carries legal and political accountability requirements; elected officials and agency heads typically must sign off on proposals, and there are implicit or explicit requirements for human authorship and authorization. Error costs in policy language are high and asymmetrically borne by organizations.
Adoption barriersclaude-sonnet-53/5No formal licensing requirement, but political sensitivity, confidentiality, and the need for expert judgment on policy substance create moderate institutional friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI inference costs are low, but integration into legislative workflows, prompt engineering, fact-checking, and mandatory human review add substantial overhead. The total cost per usable output remains closer to or exceeds the loaded cost of a skilled policy writer.
Cost vs. human wageclaude-sonnet-54/5AI drafting tools are far cheaper per document than a political scientist's time for initial drafts, though human review costs remain necessary.
Technical feasibility todayclaude-haiku-4-5-202510013/5LLMs are deployed in some government and policy settings for initial draft generation, but they have material error rates in legal language, factual accuracy, and policy detail. Production use exists but with significant human oversight required, not autonomous reliable performance.
Technical feasibility todayclaude-sonnet-53/5LLM products are already used for drafting government communications and policy memos in practice, but legislative-quality drafting still requires expert legal/political review, limiting full reliability.

Forecast political, economic, and social trends.

36

CI 30–41 · exposure 25 · augmentation 75 · importance 3.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Political science and policy organizations adopt AI gradually for data analytics and literature review, but actual forecasting remains heavily human-led; most AI adoption in the sector is assistive rather than replacement-oriented, reflecting skepticism about AI reliability in high-stakes prediction.
Sector adoption velocityclaude-sonnet-53/5Political science, policy, and consulting sectors are professional services with growing AI pilot adoption for research assistance, though full forecasting automation remains rare in practice.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly assists political scientists by rapidly processing large datasets, identifying historical correlates, stress-testing assumptions, and generating scenario narratives; these tools substantially raise analyst productivity while human judgment remains essential for interpretation and final forecasts.
Augmentation potentialclaude-sonnet-54/5AI can significantly speed up literature review, data synthesis, and scenario drafting, meaningfully boosting analyst productivity while the human retains final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5Current AI can extract and synthesize historical data, identify patterns, and generate trend analyses, but forecasting complex, contingent sociopolitical systems requires human judgment about unprecedented events, policy shifts, and nonlinear dynamics that AI struggles with at sufficient reliability for academic or policy use.
Task automatabilityclaude-sonnet-52/5Forecasting requires synthesizing ambiguous qualitative signals, expert judgment, and contested theoretical frameworks that current AI cannot reliably replicate end-to-end; AI can assist with data aggregation but not the core judgment task at equal quality.
Adoption barriersclaude-haiku-4-5-202510013/5Organizational norms, client preference for human expertise, and institutional liability concerns around incorrect forecasts create moderate friction; however, no legal requirement mandates human forecasters, and adoption barriers are primarily reputational and internal rather than regulatory.
Adoption barriersclaude-sonnet-52/5No formal licensing requirement, but institutional trust, reputational stakes, and the need for defensible expert judgment create moderate friction against pure AI substitution.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI inference and data processing are cheap, but integration, validation, and human oversight required to produce actionable forecasts create comparable or higher all-in costs than hiring experienced analysts for specialized forecasting work.
Cost vs. human wageclaude-sonnet-52/5AI tools can cheaply generate drafts or aggregate data, but the actual valuable forecasting output still requires expensive expert review and validation, keeping overall cost comparable to or only modestly below human-only work.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI tools can support trend analysis and data synthesis, no deployed product reliably performs end-to-end political, economic, and social forecasting with credible accuracy; most production systems handle narrow sub-tasks (market prediction, sentiment analysis) rather than the integrative judgment this task demands.
Technical feasibility todayclaude-sonnet-52/5Some analytics products offer trend detection and scenario generation, but no deployed product reliably performs expert-level political/economic/social forecasting in production without heavy human oversight.

Interpret and analyze policies, public issues, legislation, or the operations of governments, businesses, and organizations.

33

CI 25–41 · exposure 25 · augmentation 63 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI for policy analysis is nascent. Think tanks, government agencies, and policy organizations are experimenting with AI as a research aid, but production replacement of analyst roles is minimal; institutional conservatism and the requirement for expert credibility slow uptake.
Sector adoption velocityclaude-sonnet-53/5Political science and policy research are part of professional/information services where AI adoption is moderate—used for literature review and drafting, but full analytical work is still largely human-driven.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully assist by rapidly summarizing policy documents, extracting key provisions, identifying precedents, and generating initial outlines for analysis. A political scientist using AI tools can work faster on information-gathering tasks, though the interpretive core remains human-driven.
Augmentation potentialclaude-sonnet-54/5AI substantially aids literature review, summarization of legislative text, and drafting of analytical frameworks, meaningfully boosting researcher productivity while human judgment remains central.
Task automatabilityclaude-haiku-4-5-202510012/5Policy analysis requires contextual judgment, weighing competing values, and synthesizing evidence across complex socio-political systems. While AI can summarize documents and identify patterns, genuine interpretation—assessing policy trade-offs, identifying unstated assumptions, and generating novel insights—remains beyond current systems' reliable capability at scale.
Task automatabilityclaude-sonnet-52/5AI can summarize and draft analysis of policy documents and legislation but genuine interpretive analysis requiring domain expertise, contextual judgment, and original synthesis exceeds current reliable automation levels for the full task.
Adoption barriersclaude-haiku-4-5-202510014/5Policy interpretation often informs high-stakes decisions affecting public welfare, budgets, and legal compliance. Organizations typically require policy analysis to be vetted by credentialed staff or experts; liability and reputational risk create strong institutional preference for human accountability and judgment.
Adoption barriersclaude-sonnet-52/5No licensing requirement restricts this task, but organizational and academic norms still favor credentialed experts for authoritative policy analysis, creating moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI systems useful for this task (LLMs, research tools) require significant prompt engineering, fact-checking, and expert validation. The human oversight cost remains comparable to or exceeds the cost of having an analyst perform the work directly.
Cost vs. human wageclaude-sonnet-53/5AI tools are cheap for initial drafting and summarization, but the human expert review, validation, and synthesis needed keep overall costs roughly comparable to a skilled analyst's time.
Technical feasibility todayclaude-haiku-4-5-202510012/5No production system reliably performs end-to-end policy analysis or government operations analysis independently. Tools exist for document summarization and information retrieval, but they cannot substitute for the interpretive and synthetic reasoning that defines political science work.
Technical feasibility todayclaude-sonnet-52/5Products like LLM-based research assistants can summarize legislation and draw comparisons, but no deployed system reliably performs expert-level political analysis without significant human oversight and fact-checking.

Teach political science.

33

CI 25–40 · exposure 25 · augmentation 88 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Higher education has been slow to adopt AI for core instruction; adoption remains primarily in supplementary roles (chatbots for office hours, automated grading) rather than replacing teaching. Professional and institutional conservatism in academic sectors limits rapid displacement.
Sector adoption velocityclaude-sonnet-53/5Higher education is adopting AI tools for course prep, grading assistance, and tutoring at a moderate pace, with pilots common but full automation rare.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist instructors by drafting lecture outlines, generating discussion prompts, providing real-time fact-checking on current events, and automating grading—all of which can substantially raise instructor productivity while keeping the human as the primary educator.
Augmentation potentialclaude-sonnet-55/5AI substantially aids lecture prep, generating readings, quizzes, examples, and explanations, significantly boosting instructor productivity while the professor remains in control.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can generate lecture content, create study materials, and answer factual questions about political systems, teaching inherently requires real-time interaction, adaptive response to student confusion, and the ability to spark critical thinking and debate—tasks that current AI handles poorly at scale. AI cannot reliably manage the full classroom experience or assess genuine understanding with nuance.
Task automatabilityclaude-sonnet-52/5Teaching involves live delivery, student interaction, mentorship, and adaptive pedagogy that AI cannot fully replicate end-to-end today, though content generation portions can be automated.aquí
Adoption barriersclaude-haiku-4-5-202510014/5Educational institutions have strong regulatory frameworks, accreditation requirements, and institutional norms expecting credentialed human instructors. Legal liability for educational quality and student outcomes, combined with professional standards and institutional resistance, creates meaningful friction.
Adoption barriersclaude-sonnet-53/5Accreditation, institutional norms, and student expectations for human instructors create moderate friction, though no strict licensing law bars AI-assisted teaching content.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI-generated lecture notes and practice materials cost orders of magnitude less per student than hiring instructors, though oversight and refinement still require human involvement. Scaling to replace full instruction delivery would be far cheaper than human salaries.
Cost vs. human wageclaude-sonnet-52/5While AI can cheaply generate lecture materials, replicating the full instructional role (office hours, advising, live discussion) still requires substantial human labor, keeping costs comparable.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI tutoring systems and content generators exist but perform narrowly and unreliably for complex subjects like political science that require contextual judgment and integration of current events. No mature product reliably teaches political science end-to-end in a way comparable to a human instructor.
Technical feasibility todayclaude-sonnet-52/5AI tutoring and content tools exist but no deployed product independently teaches a full political science course with grading, discussion facilitation, and mentorship at production reliability.

Provide media commentary or criticism related to public policy and political issues and events.

31

CI 29–34 · exposure 25 · augmentation 63 · importance 2.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Despite rapid AI adoption in tech and finance, news organizations and academic/policy media remain cautious about AI-generated opinion content. Adoption of AI for commentary remains minimal; most outlets augment rather than replace human analysts.
Sector adoption velocityclaude-sonnet-52/5Media and academic commentary sectors have been slow to substitute AI for human pundits, though AI research tools are increasingly used for background prep.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by drafting background summaries, identifying policy documents, or suggesting argument structures, meaningfully accelerating research and writing. However, the creative and interpretive core—developing original claims and political insight—remains human-driven, limiting augmentation impact.
Augmentation potentialclaude-sonnet-54/5AI can help political scientists quickly research background, summarize opposing viewpoints, and draft talking points, meaningfully speeding commentary preparation while the human delivers final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While LLMs can generate text mimicking political commentary, this task fundamentally requires original insight, nuanced judgment about current events, and credible expertise. Current AI cannot reliably meet the 50% time-saving bar when originality, accuracy, and interpretive depth are required; it would need substantial human reworking.
Task automatabilityclaude-sonnet-52/5AI can draft commentary and summarize political events, but authoritative media commentary requires named expertise, credibility, and real-time judgment that current AI cannot fully replace end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Media organizations face reputational, legal, and editorial liability if political commentary lacks accuracy or attribution. Audiences expect human expertise and credibility; substituting unlabeled AI commentary creates legal and trust risks, and many outlets have explicit policies against AI-generated editorial content.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but media outlets and audiences value named, credentialed experts with reputational accountability, creating moderate adoption friction.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI inference is cheap, but integration into editorial workflows and mandatory expert oversight (fact-checking, reputational risk assessment) adds significant cost, making the all-in cost roughly comparable to a political scientist's loaded wage for original commentary.
Cost vs. human wageclaude-sonnet-53/5AI drafting assistance is cheap, but the human expert's credibility, reputation, and live engagement remain necessary, so overall cost savings are moderate rather than dramatic.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs this task at production quality. AI systems can draft commentary, but editorial review for factual accuracy, nuance, and original argument is necessary, and media outlets do not substitute AI for professional political scientists in commentary roles.
Technical feasibility todayclaude-sonnet-52/5AI writing tools can generate draft commentary or talking points, but no deployed product independently provides credible on-air or bylined political commentary at scale.

Disseminate research results through academic publications, written reports, or public presentations.

31

CI 25–36 · exposure 25 · augmentation 75 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Academic sectors are cautious adopters of automation in research dissemination; institutions and journals are only beginning to grapple with AI policy. Most research communication remains human-driven, and early AI adoption is largely assistive (editing drafts) rather than replacement-level.
Sector adoption velocityclaude-sonnet-53/5Academia is adopting AI writing and presentation tools somewhat, but institutional caution, authorship policies, and slower-moving academic publishing norms keep adoption at a middling pace compared to fast-moving corporate sectors.
Augmentation potentialclaude-haiku-4-5-202510014/5AI excels at augmenting the dissemination process—drafting sections, suggesting structure, checking clarity, formatting documents, and generating presentation outlines. These tools meaningfully accelerate the researcher's workflow while the human retains full control over argument, framing, and accountability.
Augmentation potentialclaude-sonnet-54/5AI tools substantially help with drafting text, creating visuals, summarizing findings, and formatting for journals or slides, meaningfully boosting researcher productivity while they remain in control of content and delivery.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can draft text and structure documents, political science dissemination requires substantial human judgment on framing, argumentation rigor, and audience alignment. AI could accelerate writing and formatting, but generating publication-quality research narratives with original scholarly voice and epistemic responsibility still requires significant human direction and revision.
Task automatabilityclaude-sonnet-52/5AI can draft manuscript sections, summaries, and slides, but the full task includes journal submission, peer review navigation, live presentation, and defending work, which remain human-driven and beyond current automation thresholds.
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: academic institutions hold researchers accountable for published claims, journals require author certification of originality and accuracy, and ethical norms demand human intellectual ownership. Liability and reputational risk mean organizations cannot substitute AI for researcher accountability in dissemination.
Adoption barriersclaude-sonnet-53/5No licensing requirement blocks AI use, but academic norms around authorship, integrity, peer review, and personal credibility for public presentations create meaningful institutional friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI inference costs are low, but integration and mandatory human review/revision of research outputs remain substantial. The cost saving is modest because skilled researchers must validate and reshape AI-generated prose to meet publication standards, limiting the per-task economic advantage.
Cost vs. human wageclaude-sonnet-53/5AI can cut drafting and formatting time significantly at low cost, but human oversight, editing, and presentation delivery still require substantial paid time, keeping overall cost roughly comparable rather than order-of-magnitude cheaper.
Technical feasibility todayclaude-haiku-4-5-202510012/5Current AI systems can assist with drafting and editing, but deployed products do not reliably produce complete, publication-ready academic content autonomously. Research dissemination involves context-dependent choices about emphasis, citation, and positioning that exceed what off-the-shelf systems do reliably without expert oversight.
Technical feasibility todayclaude-sonnet-52/5Writing assistants and presentation tools are widely used to support drafting, but no deployed product reliably manages the full dissemination pipeline (submission, revision, public presentation) end-to-end.

Identify issues for research and analysis.

30

CI 25–35 · exposure 25 · augmentation 75 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Academic sectors show slow, cautious adoption of AI for research planning; most use remains in supplementary roles (literature discovery) rather than driving primary research decisions.
Sector adoption velocityclaude-sonnet-52/5Academic research is a slower-adopting sector; AI tools for literature discovery are used experimentally but not yet deeply embedded in political science research workflows.
Augmentation potentialclaude-haiku-4-5-202510014/5AI systems can meaningfully assist by mining literature, identifying citation trends, suggesting underexplored intersections, and surfacing emerging themes, substantially improving a political scientist's efficiency in scanning the research landscape while maintaining human judgment on priority.
Augmentation potentialclaude-sonnet-54/5AI can efficiently scan literature, identify trends, gaps, and emerging debates, and suggest potential research questions, substantially aiding a researcher's idea-generation process while they retain final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist in scanning literature and identifying trending topics or gaps, but selecting research issues requires domain expertise, judgment about significance, feasibility, and alignment with funding/institutional priorities that current AI systems cannot reliably perform end-to-end.
Task automatabilityclaude-sonnet-52/5Identifying novel, worthwhile research issues requires original judgment, field expertise, and awareness of scholarly gaps that current AI cannot reliably replicate end-to-end, though it can surface trending topics or gaps in literature as input.dev
Adoption barriersclaude-haiku-4-5-202510014/5Research priority-setting is a high-judgment task embedded in peer review, grant committees, and institutional governance; there are no legal barriers to automation, but strong professional and organizational norms expect human expertise to drive research agenda-setting.
Adoption barriersclaude-sonnet-52/5No licensing requirement blocks AI assistance here, but disciplinary norms, peer review, and the need for domain credibility create moderate organizational friction to relying on AI-identified topics without human vetting.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI tools for issue scanning remain relatively expensive to maintain and require substantial human oversight and refinement, making them costlier than having an experienced political scientist identify issues directly.
Cost vs. human wageclaude-sonnet-52/5AI tools can cheaply scan literature and data, but the human synthesis and judgment needed to validate an issue's significance still requires substantial expert time, keeping overall cost savings modest.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably identifies high-quality political science research issues at the level a professional would require; AI literature analysis tools exist but have material limitations in understanding research context, novelty, and impact.
Technical feasibility todayclaude-sonnet-52/5AI literature-review and trend-analysis tools exist and are used to surface potential topics, but no deployed product independently identifies research-worthy political science issues with the judgment quality expected of a professional.

Develop and test theories, using information from interviews, newspapers, periodicals, case law, historical papers, polls, or statistical sources.

29

CI 21–38 · exposure 25 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Political science and academic research lag in AI adoption relative to tech and finance sectors; most theory development remains human-driven with only emerging use of AI for literature support. Production-level AI-driven theory generation is not yet observed in the field.
Sector adoption velocityclaude-sonnet-53/5Academia and research fields are adopting AI tools for literature review, coding qualitative data, and statistical analysis at a moderate pace, though core theorizing remains human-driven.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by rapidly synthesizing source material, identifying patterns in large datasets, and suggesting connections across papers, which raises researcher productivity in the exploratory phase. However, the core intellectual work of theory formulation and validation remains human-centric.
Augmentation potentialclaude-sonnet-54/5AI substantially accelerates literature review, data extraction from historical/legal texts, and statistical analysis, meaningfully boosting researcher productivity while humans retain interpretive and theoretical control.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with literature review, data aggregation, and statistical analysis of sources, developing and testing theories requires novel intellectual synthesis, causal reasoning, and research design judgment that current systems cannot reliably perform end-to-end. The task inherently demands originality and disciplinary expertise that AI cannot yet match at quality parity with 50%+ time savings.
Task automatabilityclaude-sonnet-52/5AI can help gather and summarize sources and identify patterns, but developing and testing original political theory requires creative synthesis, judgment, and domain expertise that current systems cannot reliably perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Institutional norms in academia strongly favor human authorship and accountability for novel theories; peer review, publication gatekeeping, and disciplinary credit systems legally and professionally require a human researcher to own and defend theoretical claims. Liability and reputational risk for AI-generated theory is high.
Adoption barriersclaude-sonnet-52/5No licensing requirement restricts this task, but academic norms, peer review, and expectations of original scholarly contribution create moderate friction against pure AI authorship.
Cost vs. human wageclaude-haiku-4-5-202510011/5End-to-end automation remains infeasible, so cost comparison is moot; the infrastructure required (domain-specific fine-tuning, human oversight, validation) would exceed the loaded wage of a mid-career political scientist for meaningful research output.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply assist with literature review and data processing, but the theory-development core still requires expensive expert human time, keeping overall cost comparable to or only modestly less than a human researcher.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs theory development and testing; AI tools exist for parts of the pipeline (document parsing, citation analysis, basic summarization) but none integrate them into coherent theory formulation and validation. Products in this space remain research-stage or narrow-scope prototypes.
Technical feasibility todayclaude-sonnet-52/5Deployed products (chatbots, research assistants) can summarize documents and run statistical analyses, but no product autonomously develops and tests novel political theories in production.

Evaluate programs and policies, and make related recommendations to institutions and organizations.

29

CI 25–32 · exposure 25 · augmentation 75 · importance 2.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Policy research institutions, think tanks, and government agencies are using AI for data gathering and draft support, but adoption of AI-driven autonomous recommendations remains limited. The sector values human expertise and judgment, and organizational inertia around policy-making processes slows deeper automation adoption.
Sector adoption velocityclaude-sonnet-53/5Political science, policy research, and think tank sectors are adopting AI tools for research assistance and drafting at a moderate pace, with pilots more common than full production deployment for actual recommendation-making.
Augmentation potentialclaude-haiku-4-5-202510014/5AI systems excel at augmenting political scientists by rapidly synthesizing large bodies of research, identifying data patterns, and generating draft analyses. These tools meaningfully accelerate the early stages of policy evaluation and can free experts to focus on interpretation, stakeholder engagement, and strategic recommendation formulation.
Augmentation potentialclaude-sonnet-54/5AI substantially aids literature review, data synthesis, comparative policy analysis, and drafting of evaluation reports, meaningfully increasing researcher productivity while humans retain responsibility for final judgments and recommendations.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can gather data, summarize existing research, and draft preliminary frameworks for policy evaluation, the core task requires integrative judgment about institutional context, political feasibility, stakeholder interests, and value trade-offs that human political scientists must ultimately weigh. Current AI cannot end-to-end replace the normative reasoning and contextual understanding needed for credible policy recommendations.
Task automatabilityclaude-sonnet-52/5AI can synthesize data and draft policy evaluation frameworks, but forming credible, context-sensitive recommendations that account for political, institutional, and normative nuance still requires substantial human judgment and expertise, so full end-to-end automation with equal quality is not yet feasible.
Adoption barriersclaude-haiku-4-5-202510014/5Organizations and institutions typically expect policy recommendations to be authored or signed off by credentialed experts who bear reputational and legal responsibility. Client trust, regulatory expectations in some sectors (e.g., government bodies), and liability concerns create strong organizational and reputational barriers to full automation.
Adoption barriersclaude-sonnet-53/5No formal licensing requirement mandates a human political scientist, but institutional trust, accountability for policy consequences, and stakeholder expectations of human judgment create moderate friction against pure AI automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for research assistance and analysis are relatively inexpensive per task, but the task itself requires high-value expert time for interpretation, stakeholder consultation, and final recommendation formulation. The loaded cost of a political scientist's judgment still exceeds the savings from AI-assisted components.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply generate drafts, but the human oversight, domain validation, and stakeholder-sensitive judgment needed to make defensible policy recommendations keep effective all-in costs closer to comparable levels than an order-of-magnitude savings.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs end-to-end policy evaluation and recommendation-making for institutions. AI can assist with literature review, data analysis, and draft reports, but production systems do not autonomously synthesize findings into institutional recommendations at scale or with sufficient fidelity to replace expert judgment.
Technical feasibility todayclaude-sonnet-52/5General-purpose LLMs can produce draft policy analyses and literature summaries, but no deployed product reliably performs credible political science program evaluation and recommendation-making at professional standards in production settings.

Advise political science students.

24

CI 11–37 · exposure 13 · augmentation 63 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Higher education has been slow to adopt AI for core advising roles despite interest in administrative automation. Most institutions use AI only as a supplementary tool for information lookup and scheduling, not replacement advising, reflecting cultural preference for human mentorship and institutional caution.
Sector adoption velocityclaude-sonnet-52/5Higher education has been slow to adopt AI for personalized advising roles, with most AI use limited to administrative support tools rather than substantive mentorship.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist advisors by retrieving policy information, suggesting career pathways based on student profiles, drafting correspondence, and handling routine scheduling—allowing faculty to focus on deeper mentoring conversations and relationship-building.
Augmentation potentialclaude-sonnet-53/5AI can help advisors by drafting resources, summarizing degree requirements, or suggesting readings, but the core advising interaction remains human-led.
Task automatabilityclaude-haiku-4-5-202510012/5Advising students requires understanding individual circumstances, career goals, and nuanced research guidance. While AI can assist with information provision and general recommendations, the core function—personalized mentoring and professional judgment—requires human judgment and relationship-building that current systems cannot reliably replace end-to-end.
Task automatabilityclaude-sonnet-51/5Advising involves personalized mentorship, career guidance, and relationship-building that requires human judgment and rapport, which current AI cannot substitute for end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Significant barriers exist: universities employ faculty advisors partly to fulfill accreditation and duty-of-care requirements; students and parents typically expect human contact for academic and career advising; and institutions face liability if algorithmic advice causes academic harm. Regulatory frameworks around educational quality further protect the human role.
Adoption barriersclaude-sonnet-53/5No licensing requirement mandates a human advisor, but institutional norms, student expectations of personal mentorship, and liability for poor academic/career guidance create moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI-assisted advisory tools (chatbots, scheduling automation) cost far less than hiring advising staff, even accounting for human oversight. A deployed system could handle basic inquiries at negligible cost compared to faculty time at institutional wage scales.
Cost vs. human wageclaude-sonnet-52/5While AI could cheaply handle simple informational queries, the substantive advising component still requires costly human oversight and interaction, keeping overall cost comparable to human-delivered advising.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs sustained academic advising at the standard expected in university settings. AI chatbots can answer procedural questions, but they lack the credibility, legal standing, and contextual understanding of departmental policies and student circumstances that institutional advising demands.
Technical feasibility todayclaude-sonnet-51/5No deployed product provides comprehensive academic advising for political science students; existing chatbots handle only narrow FAQ-style queries, not substantive advising.

Consult with and advise government officials, civic bodies, research agencies, the media, political parties, and others concerned with political issues.

14

CI 11–16 · exposure 5 · augmentation 63 · importance 2.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Political consulting and government advisory work remain heavily dependent on human relationships, reputation, and institutional trust. Adoption of AI for background research is emerging, but production displacement of the advisory role itself is minimal across public sector and political organizations.
Sector adoption velocityclaude-sonnet-52/5Political science and policy advising are slow-adopting, relationship-driven fields with limited large-scale AI agent deployment for direct advisory roles.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully augment political scientists by synthesizing research, drafting briefing materials, and identifying policy precedents, which raises productivity on preparatory work. However, the core advisory interaction—listening, interpreting stakeholder needs, and offering judgment—remains human-centered.
Augmentation potentialclaude-sonnet-54/5AI can strongly assist by synthesizing research, summarizing polling and policy data, and drafting briefing materials that improve the scientist's advisory output.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires deep contextual judgment, relationship-building, and the ability to synthesize complex political information for diverse stakeholders with competing interests. Current AI cannot replicate the trusted advisory role or navigate the nuanced political dynamics that underpin effective consultation.
Task automatabilityclaude-sonnet-51/5This task requires building trust, situated judgment, and real-time advisory relationships with named stakeholders, which current AI cannot originate or conduct independently.
Adoption barriersclaude-haiku-4-5-202510014/5Strong organizational and reputational barriers exist: government officials and political parties require trusted human advisors who can be held accountable; media and civic bodies prefer human expertise they can attribute and defend; and liability for bad political counsel falls on the human consultant, not the AI tool.
Adoption barriersclaude-sonnet-54/5Advisory credibility, reputational accountability, and institutional trust strongly favor a credentialed human expert being the visible advisor, even if not formally licensed.
Cost vs. human wageclaude-haiku-4-5-202510012/5A political scientist's consultation carries reputational risk and decision-consequence weight that AI cannot assume; oversight and human sign-off are mandatory, so total cost (AI + human verification + liability) exceeds the loaded wage for a human doing the task directly.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply generate background analysis, but the actual consulting engagement still requires paid expert time, so overall cost savings are limited.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can generate political analysis and research summaries, no deployed product reliably performs the core consultation function—advising officials and parties on politically sensitive matters with accountability. Some tools assist with background research, but the advisory relationship itself remains human-dependent.
Technical feasibility todayclaude-sonnet-51/5No deployed product substitutes for a human expert advisor consulting with officials or media; AI is at most a research/drafting aid behind the scenes.

Serve on committees.

0

CI 0–0 · exposure 0 · augmentation 38 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Committee service is an inherently human governance function with no plausible AI substitution pathway in practice. Adoption of AI tools for research support may occur, but not for the committee role itself.
Sector adoption velocityclaude-sonnet-51/5This is a governance/representational function with essentially zero AI adoption in practice; academia and policy bodies have not begun substituting AI for committee roles.
Augmentation potentialclaude-haiku-4-5-202510012/5AI can assist with background research, literature synthesis, and brief drafting before committee meetings, but it cannot assist with the core deliberative, judgment, and voting functions that define the task.
Augmentation potentialclaude-sonnet-53/5AI can help prepare materials, summarize documents, draft agendas, or analyze data discussed in committee, providing moderate support to the human member.
Task automatabilityclaude-haiku-4-5-202510011/5Committee service requires judgment, deliberation, consensus-building, and voting on substantive matters. These are fundamentally human roles requiring contextual wisdom, accountability, and interpersonal negotiation that current AI cannot perform end-to-end.
Task automatabilityclaude-sonnet-51/5Serving on a committee requires embodied presence, relationship-building, real-time deliberation, and personal accountability that AI cannot substitute for today.
Adoption barriersclaude-haiku-4-5-202510015/5Hard legal and institutional barriers require that committee members be designated humans with legal standing, fiduciary duty, and accountability. Formal appointment, voting rights, and liability are tied to natural persons or organizations.
Adoption barriersclaude-sonnet-55/5Committee membership typically requires formal appointment, institutional trust, accountability, and often specific credentials or elected/appointed status that only a human can hold.
Cost vs. human wageclaude-haiku-4-5-202510011/5Committee service is a governance and accountability role, not a cost-saving input task. The entire value lies in human judgment and institutional legitimacy, making cost comparison to AI irrelevant.
Cost vs. human wageclaude-sonnet-51/5There is no AI product that performs this role at all, so no cost comparison favors AI; human presence is the only viable option.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product performs committee service as a substitute for humans. AI cannot legally serve as a voting member, bear responsibility for decisions, or engage in the deliberative processes that define committee work.
Technical feasibility todayclaude-sonnet-51/5No deployed product allows AI to actually serve as a committee member representing an institution or professional body.

Related occupations — Life, Physical & Social Science

How to read this

A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.

What would change this score

New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.