Astronomers
19-2011.00Median wage $128,820/yr2,120 employed (US)Rank #481 of 923 scored · top 52% by substitution
Observe, research, and interpret astronomical phenomena to increase basic knowledge or apply such information to practical problems.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution28
Exposure23
Augmentation68
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
17 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
6%
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%23
panel mean rating 1.9/5 → substitution pressure 23/100
Technical feasibility todayw 20%23
panel mean rating 1.9/5 → substitution pressure 23/100
Cost vs. human wagew 15%26
panel mean rating 2.0/5 → substitution pressure 26/100
Adoption barriersw 20%inverted — strong barriers lower the score41
panel mean rating 3.3/5 (barrier strength) → substitution pressure 41/100
Sector adoption velocityw 10%30
panel mean rating 2.2/5 → substitution pressure 30/100
Task breakdown (17 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.
Calculate orbits and determine sizes, shapes, brightness, and motions of different celestial bodies.
85CI 79–91 · exposure 80 · augmentation 100 · importance 3.3/5 · click for rater detail
Calculate orbits and determine sizes, shapes, brightness, and motions of different celestial bodies.
85| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Astronomy is a highly digitized, information-intensive field where large survey projects (Gaia, LSST, SDSS) routinely employ automated pipelines and ML models for orbit and property calculations; adoption is rapid and systemic in major observatories and research institutions. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Astronomy and astrophysics research is a highly digitized field with rapid adoption of automated pipelines and machine learning for data processing, though full research-agent autonomy is still limited. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI-driven tools dramatically augment astronomer productivity by automating routine calculations, detecting anomalies in large datasets, and enabling rapid cross-matching across surveys—enabling astronomers to focus on hypothesis generation, interpretation, and discovery while the system handles computational bulk. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI and computational tools dramatically enhance astronomers' ability to calculate orbits and physical properties, handling massive datasets far beyond manual capability while astronomers interpret and validate results. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Most of the computational work—orbit calculations, photometric measurements, astrometric position determination, and kinematic analysis—can be performed end-to-end by existing AI/ML systems and numerical software with >50% time savings. However, interpretation of ambiguous or novel celestial phenomena still requires human expertise, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 4/5 | Orbit calculation and photometric/astrometric analysis are highly mathematical, well-defined tasks that AI-assisted computational tools and software already handle extensively, though novel research interpretation still requires human oversight.“}, |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers prevent automation; publication norms and institutional practices favor human-reviewed results, but these are soft constraints rather than hard requirements. Domain expertise and interpretation still gate some decisions, creating mild friction but not blocking substitution. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There is no licensing, legal, or human-contact requirement for performing celestial mechanics calculations; it's purely computational work. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | The marginal cost of running existing computational models and software pipelines on modern hardware is orders of magnitude cheaper than the loaded wage of a professional astronomer performing these calculations manually, especially for large survey datasets. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Computational orbit and photometric calculations are extremely cheap relative to a PhD astronomer's time once pipelines are built, offering massive cost savings at scale. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature, production-grade systems exist for these tasks: specialized astronomy software (e.g., SIMBAD, Gaia pipelines), machine learning models trained on large surveys, and standard numerical libraries are deployed routinely in observatories and research institutions worldwide to compute orbits and celestial body properties at scale. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Established astronomy software (orbit determination pipelines, photometry pipelines, N-body simulators) already automates most of these calculations reliably in production observatories and survey pipelines like Gaia and LSST. |
Analyze research data to determine its significance, using computers.
58CI 55–61 · exposure 50 · augmentation 100 · importance 4.7/5 · click for rater detail
Analyze research data to determine its significance, using computers.
58| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Astronomy is among the most digitized and computationally driven sciences; major surveys (Sloan, LSST, JWST pipelines) already employ automated detection and filtering systems. Adoption of ML-assisted analysis pipelines is fast and deep within the field, though human interpretation remains the final gate. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Astronomy research groups increasingly use AI/ML tools for data analysis (e.g., in survey astronomy), but adoption is uneven across subfields and often remains at pilot or specialized-tool stage rather than universal integration. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments astronomer productivity by automating exploratory analysis, hypothesis generation, and statistical testing, freeing human judgment for interpreting results and framing new questions. This represents a high-productivity human-in-the-loop model already widely adopted in observational astronomy. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially augments astronomers by automating pattern recognition, filtering large datasets, and flagging candidates for further study, greatly boosting productivity while humans retain interpretive control. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can automate significant portions of data analysis—statistical testing, pattern detection, anomaly identification, and visualization—but determining 'significance' requires contextual scientific judgment about whether findings address research questions and warrant follow-up. Current systems excel at computational steps but not the interpretive synthesis that constitutes the full task. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can accelerate data processing, pattern detection, and statistical analysis, but determining scientific significance requires domain judgment, hypothesis framing, and contextual interpretation that current systems cannot fully replicate end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few hard legal or licensing barriers prevent automated analysis; however, publication norms, peer review, and institutional expectations that a human scientist must vouch for significance introduce moderate friction. Liability for incorrect conclusions is borne by the institution but does not prevent automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement restricts this task, though scientific credibility, peer review norms, and publication accountability create moderate institutional friction against fully automated conclusions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Cloud-based AI services for data analysis (compute, inference, storage) are substantially cheaper than employing a full-time astronomer for routine analysis tasks, though integration and interpretation overhead narrows the advantage. At scale, infrastructure cost per analysis iteration is likely 5–10× lower than human wage. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Computational tools reduce processing time substantially, but the specialized data curation, model validation, and oversight needed still require costly expert astronomer time comparable to traditional analysis costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Mature tools exist for exploratory data analysis, statistical inference, and automated detection pipelines (e.g., in survey astronomy), but they operate within narrow scopes and require expert oversight to validate results and interpret context. Production systems handle components reliably but not end-to-end significance determination. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Machine learning pipelines are routinely deployed in astronomy for classification, anomaly detection, and signal processing, but final significance assessment still relies on human astronomers reviewing and interpreting outputs. |
Develop and modify astronomy-related programs for public presentation.
45CI 30–60 · exposure 38 · augmentation 75 · importance 3.5/5 · click for rater detail
Develop and modify astronomy-related programs for public presentation.
45| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Educational and cultural institutions (planetariums, museums) adopt new tools slowly and prioritize human expertise in public-facing content. AI-assisted development tools are emerging but remain in early/pilot phases in the astronomy education sector rather than mainstream production deployment. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Academic and scientific institutions show moderate AI tool adoption for coding tasks, with pilots and partial integration but not deep production-scale automation specific to this niche task. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially augments astronomy program development through code scaffolding, visualization suggestions, automated testing, and rapid prototyping of interactive features, allowing human astronomers to focus on scientific accuracy, pedagogical design, and creative storytelling. This is a strong augmentation case where AI handles mechanical work and the astronomer retains judgment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially speeds up code drafting, debugging, and content generation for presentations, meaningfully boosting astronomer productivity while they retain creative and scientific control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with code generation and basic program structure, the task requires deep domain knowledge of astronomy, pedagogical judgment about public engagement, and iterative refinement based on audience feedback—aspects that remain human-dependent. Current AI systems lack the contextual understanding to independently develop programs that balance scientific accuracy with accessibility. |
| Task automatability | claude-sonnet-5 | 3/5 | AI coding assistants can generate and modify presentation software or visualization scripts substantially, but astronomy-specific accuracy, domain-specific data pipelines, and public-facing design polish still require significant human oversight. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Museums, planetariums, and educational institutions prefer human astronomers to design public programs, and many have established internal processes and trust relationships. However, no legal or licensing barrier strictly prevents AI-assisted or AI-generated program development, though organizational friction and quality standards impose moderate friction. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or regulatory requirement mandates human authorship for public science presentation software. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted code generation reduces some labor costs, but the overhead of domain validation, scientific accuracy checking, pedagogical design, and iterative refinement by qualified astronomers remains substantial. The full-cycle cost is likely comparable to or higher than hiring a skilled astronomer-programmer. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI coding assistance is cheap per query, but the overall task requires domain expert review, testing, and integration, keeping all-in cost roughly comparable to a human-augmented workflow rather than an order of magnitude cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No mature production system reliably develops astronomy presentation programs end-to-end. Existing code-generation AI can draft snippets, but developing a coherent, scientifically accurate, and publicly engaging program requires human oversight and domain expertise that deployed products do not reliably provide at scale. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed code-generation tools (e.g., Copilot, ChatGPT) reliably assist with scripting and debugging, but no product autonomously develops complete public astronomy presentation programs without human integration. |
Teach astronomy or astrophysics.
31CI 25–37 · exposure 30 · augmentation 75 · importance 4.1/5 · click for rater detail
Teach astronomy or astrophysics.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Educational institutions, especially those teaching advanced subjects like astrophysics, adopt AI slowly due to accreditation constraints, faculty union protections, and conservative pedagogy. While some universities experiment with AI-assisted tutoring, replacement of primary instruction remains rare and contentious in higher education. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Higher education is adopting AI tools for tutoring, content creation, and grading assistance at a moderate pace, with pilots common but full replacement of instructors rare. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments teaching by drafting lecture notes, generating problem sets, explaining difficult concepts in multiple ways, and providing instant student feedback—freeing instructors to focus on mentorship, assessment, and creative pedagogy. Professors using AI tools can handle larger classes and iterate curricula faster while remaining essential to the learning process. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids astronomers in preparing lecture materials, generating visualizations, answering student questions, and creating problem sets, meaningfully boosting teaching productivity while the instructor remains central. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can generate course materials, explanations, and practice problems at scale, but teaching requires real-time engagement, adapting to student confusion, and responding to unexpected questions—tasks that demand human judgment and presence. While AI can draft lectures or answer standard questions, it cannot fully replicate the pedagogical interaction that characterizes effective teaching. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can generate lecture content, explanations, and practice problems, but live teaching involves interactive instruction, mentoring, and adapting to student needs that current systems cannot fully replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Universities and professional institutions have strong regulatory, accreditation, and reputational incentives to employ credentialed humans for formal instruction. Teaching requires institutional accountability, legal signatures on courses, and cultural expectations that learning is mediated by qualified professionals—not merely information delivery. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Academic institutions typically require credentialed faculty to teach and assess students, and accreditation standards create organizational and credentialing friction, though no strict licensing law mandates a human specifically for all teaching functions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce content creation costs, but a full teaching system (inference, integration, moderation, oversight) remains expensive relative to hiring adjunct faculty at lower wage rates, especially in contexts where teaching is part of a broader research or institutional role. The per-student cost advantage is modest and sector-dependent. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI content generation is cheap, actual teaching still requires human oversight, curriculum design, and interaction, so all-in cost savings versus a human instructor are modest for full course delivery. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI tutoring systems and educational chatbots exist and perform narrow instructional tasks reliably (answering factual questions, explaining concepts), but no deployed system fully handles the breadth of classroom teaching—managing discussions, assessing understanding, adjusting pacing, and motivating diverse learners. Products like tutoring chatbots work in supplementary roles but not as primary instructors at scale. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tutoring products and content generators exist and are used to supplement instruction, but no deployed system independently teaches full astronomy/astrophysics courses reliably at scale in place of instructors. |
Measure radio, infrared, gamma, and x-ray emissions from extraterrestrial sources.
31CI 25–37 · exposure 30 · augmentation 75 · importance 4.0/5 · click for rater detail
Measure radio, infrared, gamma, and x-ray emissions from extraterrestrial sources.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Astronomy has invested in automated survey pipelines and machine learning for candidate selection (e.g., transient detection), but core measurement validation and scientific interpretation remain human-driven. Adoption is moderate: some institutes are fast-moving, but the field overall is cautious about automation of frontier science. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Astronomy research is a slow-moving, resource-constrained, and specialized field where AI adoption for automated pipelines is progressing but full task automation of observational measurement remains limited. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems already assist astronomers through automated source detection, light-curve analysis, photometric redshift estimation, and anomaly flagging. These tools meaningfully accelerate the scientist's ability to filter large datasets and prioritize observations, while the astronomer retains control over hypothesis-driven follow-up. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially assists in processing raw emission data, filtering noise, detecting signals, and flagging anomalies, significantly boosting astronomer productivity in interpreting collected data. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While instrument operation and data collection can be partially automated with scheduling systems, the interpretation of multi-wavelength emissions, anomaly detection, and scientific decision-making about follow-up observations require expert human judgment. Current AI cannot reliably handle the full scientific workflow end-to-end with 50% time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | The physical measurement process depends on specialized telescopes, detectors, and instrument calibration that AI cannot perform end-to-end; AI can assist in data reduction but not the observational task itself.https://airiskindex.io |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Scientific credibility and publication standards require human expert verification of measurements and interpretation. Funding agencies, peer review, and institutional accountability create strong friction against fully automated astronomy without expert human oversight and sign-off. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing barrier exists, but access to instruments (telescopes, satellites) is highly restricted via competitive proposals and institutional control, creating strong practical barriers to automation by generic AI systems. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Instrumenting and maintaining multi-wavelength observatories is extremely capital-intensive; the marginal cost of AI automation for data processing is small relative to the human labor cost, but the total observatory operation remains far more expensive than computational overhead. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Telescope time, instrument maintenance, and specialized hardware dominate costs; AI reduces some analysis overhead but doesn't replace the expensive physical infrastructure needed for these measurements. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Telescope control and automated data pipeline systems exist in production (e.g., survey telescopes), but these focus on data acquisition and preliminary reduction rather than the full measurement and analysis task. Quality assessment and source characterization still depend heavily on human astronomers. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Automated pipelines exist for data calibration and signal processing at observatories, but the actual acquisition of emissions data still requires specialized hardware and human-directed observation planning, not autonomous AI systems. |
Develop instrumentation and software for astronomical observation and analysis.
30CI 25–35 · exposure 25 · augmentation 63 · importance 4.0/5 · click for rater detail
Develop instrumentation and software for astronomical observation and analysis.
30| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Astronomy remains a low-digitization, low-volume domain with long instrument development cycles (5–10+ years). Adoption of AI tools is minimal; observatories prioritize proven methods and expert judgment over automated system generation. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academic and research astronomy adopts AI tools slowly for coding assistance, but instrument development remains a specialized, low-digitization, hands-on engineering process with limited AI integration in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist with code generation, data pipeline design, and literature review, helping astronomers iterate faster on software components. However, assistance plateaus at design choices and instrumentation validation, where domain expertise remains essential. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI coding assistants meaningfully speed up writing and debugging analysis software, simulation scripts, and data pipelines, giving astronomers substantial productivity gains while they retain full control of instrument design and scientific judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with software development and data analysis components, the task requires creative instrumentation design, domain expertise integration, and novel problem-solving that remains firmly in the human domain. Current AI cannot autonomously develop end-to-end instrumentation systems meeting observatory specifications. |
| Task automatability | claude-sonnet-5 | 2/5 | This task blends hardware instrumentation design (highly physical, requiring lab work and fabrication) with scientific software development, where AI can assist coding but cannot autonomously design and build custom detectors, optics, or control systems end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: instrumentation development requires peer review and validation before deployment, regulatory compliance for observatory operations, and institutional sign-off on novel designs. The stakes of equipment failure create high liability asymmetry favoring human accountability and oversight. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but institutional review, grant-funded specialized expertise, and safety/reliability requirements for scientific instruments create moderate friction against wholesale automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Instrumentation development requires specialized expertise, custom hardware design, and extensive testing—activities where human astronomers and engineers remain cost-competitive. AI tools reduce some development labor but cannot replace the domain specialists needed for novel instrument design. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Coding assistance offers modest cost savings on software portions, but instrumentation design requires expensive specialized engineering and physical prototyping that AI does not reduce meaningfully. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Code generation and analysis tools exist and see limited production use, but no deployed AI system reliably develops complete astronomical instrumentation or software stacks independently. Most applications remain confined to narrow subtasks like data processing pipelines rather than full system design. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI coding assistants are used to help write analysis pipelines and simulation code, but no deployed product independently develops astronomical instrumentation or full observation software systems reliably. |
Review scientific proposals and research papers.
29CI 25–34 · exposure 33 · augmentation 63 · importance 3.8/5 · click for rater detail
Review scientific proposals and research papers.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption remains slow in astronomy and academia broadly; while some institutions pilot AI-assisted screening, the core review function remains human-centered; cultural resistance to algorithmic gatekeeping and regulatory/policy inertia in grant-making limit displacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academic and scientific review processes are conservative and slow to adopt AI tools compared to sectors like finance or tech, with only pilot-stage AI-assisted review tools in limited use. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist reviewers by summarizing papers, flagging potential issues, extracting metadata, and organizing key findings, which can reduce read time and improve consistency; however, the augmentation is partial and does not transform the reviewer's core analytical task. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI is quite useful for summarizing papers, checking references, identifying plagiarism, running statistical sanity checks, and drafting review comments, meaningfully speeding up the human reviewer's workflow. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist with preliminary screening (plagiarism, format checks, citation extraction) and summarization, but cannot reliably evaluate scientific merit, novelty, methodology soundness, or research significance—the core of peer review—which requires domain expertise and human judgment that falls short of the 50% time-saving threshold for end-to-end automation. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can summarize papers, check statistical methodology, and flag issues, but genuine peer review requires deep domain judgment, novelty assessment, and understanding of research significance that current AI cannot fully replicate end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: funding agencies, journals, and institutional review committees typically require human experts to sign off on proposal and paper evaluations; liability concerns over incorrect rejections/acceptances, peer-review tradition, and funder/publisher policies create legal and organizational friction against automated substitution. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Peer review is an institutional, reputational, and often quasi-regulatory process; funding agencies and journals require qualified human reviewers, and there's strong norm-based resistance to non-human review in scientific communities. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI infrastructure costs for document processing, combined with significant human oversight needed to validate and act on AI recommendations, approach or exceed the cost of direct expert review, particularly for specialized astronomical research where deep subject knowledge is non-negotiable. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply produce first-pass summaries or checks, but the cost of errors (funding bad science, missing flawed methodology) requires expert human review, keeping the effective cost comparable to human review with AI assistance layered on top. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI tools exist for paper classification, summarization, and citation analysis, no deployed product reliably performs the full task of meaningful scientific review; systems lack the capability to assess research quality, detect subtle methodological flaws, or contextualize work within the field at the standard required by funding agencies or journals. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some journals and funding bodies experiment with AI-assisted screening tools, but no production system reliably performs full scientific review of astronomy proposals/papers without heavy human oversight. |
Serve on professional panels and committees.
29CI 0–59 · exposure 36 · augmentation 63 · importance 3.3/5 · click for rater detail
Serve on professional panels and committees.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Astronomy and academic professional services have modest AI adoption rates in production; while some institutions use AI for meeting note-taking and document drafting, systematic adoption of AI-augmented panel work remains rare and primarily in pilots or early stages. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Academic and scientific governance bodies show negligible movement toward AI participation in committee roles; this is a laggard area for such substitution. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists astronomers on panels by rapidly synthesizing large bodies of research, drafting position papers, organizing evidence, and preparing meeting materials, substantially reducing the preparation burden while keeping humans in decision-making roles. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help astronomers prepare materials, summarize literature, draft reports, or analyze data ahead of meetings, offering moderate assistance to the human committee member. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | AI can fully automate the structural and administrative work of panel participation—drafting position statements, synthesizing research, organizing meeting materials, and drafting recommendations—meeting the ≥50% time-saving threshold. However, the core judgment-making and decision-voting requires human deliberation, so practical end-to-end automation would be partial. |
| Task automatability | claude-sonnet-5 | 1/5 | Serving on panels/committees requires human judgment, deliberation, negotiation, and interpersonal representation that AI cannot perform end-to-end; no time-saving substitution is feasible today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Panels and committees typically require a human member with subject-matter expertise, institutional affiliation, and accountability for decisions; professional norms and organizational bylaws mandate human participation and voting, creating a hard barrier to full substitution. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Committee and panel membership typically requires professional standing, credentials, peer recognition, and institutional appointment—hard human-authorization barriers that AI cannot satisfy. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI can handle research synthesis, document preparation, and logistical support at near-zero marginal cost compared to the loaded salary of senior researchers participating in panels, making it economically compelling for the automatable portions. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this role, so cost comparison favors the human by default since AI cannot deliver the output at all. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI can draft documents and summarize research for panel use, no deployed system reliably performs the full task of serving as a panel member (voting, deliberating, representing institutional positions) in production environments. Current products handle only preparation and support roles. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product serves as a panel or committee member in place of a human expert; this remains outside current product scope. |
Present research findings at scientific conferences and in papers written for scientific journals.
26CI 25–28 · exposure 25 · augmentation 75 · importance 4.7/5 · click for rater detail
Present research findings at scientific conferences and in papers written for scientific journals.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Astronomers are early adopters of AI tools for data analysis and visualization, but actual adoption of AI for writing and presenting research findings at conferences remains experimental and limited; most still conduct these activities with only supplemental AI assistance. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Academic research is a knowledge-work sector with moderate AI tool adoption (writing assistants, literature review tools) but slow institutional adoption for actual authorship or presentation due to norms and journal policies. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can significantly augment astronomer productivity by drafting paper sections, generating figure captions, suggesting organization, and creating visualizations from data, allowing humans to focus on scientific framing and discovery rather than writing mechanics. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially helps with drafting text, creating visualizations, checking grammar, summarizing related work, and preparing slides, meaningfully boosting productivity while the astronomer remains the author and presenter. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can draft text and generate visualizations from research data, the task requires synthesizing complex findings, making novel scientific arguments, and presenting them to expert audiences—activities demanding deep subject-matter judgment and originality that current AI cannot fully replicate end-to-end at the quality required for publication or conference presentation. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft manuscripts and slides but cannot autonomously conduct the research synthesis, judgment calls, and live presentation/defense of findings that constitute this task end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: authorship and accountability in peer-reviewed publications require a human researcher to take intellectual and professional responsibility; journals and conferences expect human authors to stand behind findings, and funders/institutions hold humans liable for results, not AI systems. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Scientific journals and conferences require named human authorship, accountability, and often institutional/ethical certification, plus peer scrutiny of findings, creating strong professional and normative barriers to AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools (language models, visualization software) cost far less per task than an astronomer's time, but the astronomer must still do the core intellectual work of framing findings and deciding presentation strategy, so the effective cost displacement is minimal. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI drafting tools are cheap per use, but the overall task still requires substantial expert time for accuracy, verification, and live presentation, keeping all-in cost comparable to or only modestly below human cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI writing assistants and visualization tools exist and can help draft sections and plots, but no deployed system can autonomously produce peer-review-ready astronomical papers or conference presentations without substantial human authorship, fact-checking, and creative direction. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Writing assistants and slide-generation tools are used to help draft papers, but no deployed product independently produces publication-ready scientific papers or delivers conference presentations reliably. |
Study celestial phenomena, using a variety of ground-based and space-borne telescopes and scientific instruments.
24CI 11–38 · exposure 13 · augmentation 75 · importance 4.5/5 · click for rater detail
Study celestial phenomena, using a variety of ground-based and space-borne telescopes and scientific instruments.
24| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Astronomy has adopted automated survey systems and AI-aided data analysis in research pipelines, but adoption remains concentrated in large well-funded institutions and consortia. Most observational astronomy still involves substantial human decision-making and oversight, with limited displacement of astronomer roles. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academic and research astronomy adopts AI tools (e.g., for data reduction, classification) at a moderate pace, but the observational task itself sees slower adoption due to specialized hardware and institutional processes. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI tools demonstrably augment astronomers' productivity through automated data preprocessing, source detection, classification, and hypothesis generation from large surveys. Tools like machine learning for photometric redshift estimation and anomaly detection in time-series data significantly assist human scientists while they remain in control of interpretation and new investigations. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially augments this task through automated scheduling, anomaly detection, image processing, and data triage, greatly increasing astronomers' productivity in analyzing observational data. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Observation and data collection can be partially automated (robotic telescopes, automated surveys), and AI can assist in initial data analysis, but interpreting celestial phenomena, formulating hypotheses, and designing novel experiments require human scientific judgment. Less than 50% time savings at equal quality is achievable end-to-end. |
| Task automatability | claude-sonnet-5 | 1/5 | The physical operation of telescopes and instruments, along with observational planning and hands-on data acquisition, requires human scientific judgment and physical/remote instrument control that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Access to telescopes (ground-based and space-borne) is gated by institutional licensing, peer review, and observation-time allocation committees. Publication and credentialing of results typically require human accountability. These structural barriers significantly protect the role from full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing barrier exists, but access to expensive shared telescope resources, proposal review processes, and institutional science practices create moderate friction to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Specialized ground and space-based instrumentation, telescope access fees, and computational infrastructure are expensive. AI inference costs for analysis are modest, but the dominant cost remains human expertise and instrument time, making total replacement far more expensive than deploying trained astronomers. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Telescope time, instrument operation, and calibration remain costly and human-dependent; AI reduces some analysis costs but doesn't replace the core observational task, so overall cost savings are limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While automated telescopes and AI-assisted data processing are deployed in some observatories, no current system reliably performs the full task of studying celestial phenomena without substantial human involvement. Products handle narrow subtasks (image stacking, anomaly detection) but lack the interpretive and experimental-design capabilities astronomers require. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously conducts astronomical observation campaigns; AI is used only as a tool within human-directed workflows, not as an independent observer. |
Develop theories based on personal observations or on observations and theories of other astronomers.
21CI 11–30 · exposure 8 · augmentation 63 · importance 4.1/5 · click for rater detail
Develop theories based on personal observations or on observations and theories of other astronomers.
21| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Astronomy is a research-intensive, specialized field with slow organizational change relative to information technology sectors. While computational aids are adopted, the core task of theory development remains human-driven with limited evidence of AI displacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academic astronomy is adopting AI tools for data analysis and literature synthesis but the core theoretical work remains largely human-driven with slow, cautious integration of AI capabilities. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by rapidly analyzing large observational datasets, identifying patterns, and summarizing related theories from literature, which can help astronomers formulate and test hypotheses more efficiently, though the creative synthesis remains human-led. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids astronomers by rapidly synthesizing literature, identifying patterns in data, running simulations, and suggesting hypotheses, meaningfully accelerating the human-led theory development process. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Developing novel astronomical theories requires creative synthesis of observational data, evaluation of competing explanations, and generation of fundamentally new ideas—domains where current AI lacks demonstrated capability to operate end-to-end at a level that would meet the 50% time-saving threshold for a practicing astronomer. |
| Task automatability | claude-sonnet-5 | 2/5 | Theory development requires deep scientific creativity, novel synthesis, and judgment about physical plausibility that current AI cannot reliably replicate end-to-end, though it can assist with literature review and pattern-finding. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Academic and professional reputation, peer review, and the requirement for human intellectual judgment and creativity form substantial barriers to automation. Publication and career advancement are tied to individual theorists, creating institutional friction against replacement by AI systems. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No formal licensing barrier exists, but scientific credibility, peer review norms, and the requirement for rigorous validation by trained experts create moderate friction against AI-only theory generation being accepted. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for data processing and analysis are relatively cheap, but they do not replace the astronomer's role in theory development, only assist components of it. The full task still requires the expert human, making all-in costs comparable to or exceeding human labor. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI inference is cheap, the actual theoretical output requires extensive human verification and domain expertise, making the effective cost of AI-generated theory comparable to or more expensive than direct human effort given error correction needs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system reliably develops new astronomical theories in production; this remains a human expertise task. While AI can assist in data analysis and literature review, the creative leap of theory construction is not yet a solved problem with existing products. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously generates validated astrophysical theories; this remains a research-stage capability at best, with AI serving only as a tool for human theorists. |
Conduct question-and-answer presentations on astronomy topics with public audiences.
20CI 10–30 · exposure 17 · augmentation 63 · importance 3.2/5 · click for rater detail
Conduct question-and-answer presentations on astronomy topics with public audiences.
20| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Museums, planetariums, and observatories continue to prioritize human astronomers for public presentations because direct human engagement is core to the mission of science education and public trust. Adoption of AI-only presentation remains minimal. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Science communication and outreach in academia/museums is a low-digitization, slow-adopting sector relative to fields like finance or software. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist astronomers by drafting answers to anticipated questions, generating visuals, suggesting explanations, and providing real-time fact-checking during preparation, meaningfully boosting presenter productivity while the human remains the credible face of the presentation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can help astronomers prepare talking points, anticipate audience questions, and generate accessible explanations, meaningfully boosting preparation productivity. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires real-time audience interaction, improvisation, and nuanced explanation of complex topics in response to unpredictable questions. Current AI cannot reliably perform extended live Q&A with public audiences without significant human supervision and oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can answer astronomy questions in text form but cannot yet fully replicate the live, interactive, credibility-dependent public presentation format including real-time audience engagement and improvisation.5 |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Public audiences expect human expertise and credibility for science communication; institutions typically require human scientists to represent research findings and engage directly with the public, creating both cultural and organizational friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but public trust, institutional reputation, and audience preference for human experts create moderate friction against full substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Deploying a reliable AI presentation system with necessary oversight, moderation, and fallback human support would be more costly than hiring an astronomer for public engagement, especially given the reputational risk of AI failures in a science communication role. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI query costs are cheap, replicating a full live presentation experience with a credible human presenter and setup still requires significant integration, so cost savings are moderate rather than an order of magnitude. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI can generate astronomy facts and drafted responses to common questions, no deployed product reliably performs dynamic public question-and-answer presentations with the adaptive engagement and credibility required. Demos exist but do not substitute for human presenters in production settings. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Chatbots and voice assistants can answer astronomy Q&A but no deployed product substitutes for a live expert presenter fielding unpredictable public audience questions in real settings like planetariums or schools. |
Raise funds for scientific research.
19CI 7–30 · exposure 13 · augmentation 63 · importance 3.8/5 · click for rater detail
Raise funds for scientific research.
19| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for fundraising is nascent even in commercial sectors; in academic and scientific contexts, where personal reputation and institutional credibility drive funding, AI adoption remains experimental and limited to minor administrative tasks. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academic and scientific research funding processes are notoriously slow to digitize; AI adoption in grant writing assistance is emerging but institutional and funder-side processes remain largely traditional. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by drafting proposal language, analyzing donor databases, summarizing funding opportunities, and organizing communication workflows, providing meaningful but incremental productivity gains while humans remain the essential relationship and strategy drivers. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools substantially help with drafting proposal narratives, budget justifications, literature summaries, and identifying potential funding sources, meaningfully boosting researcher productivity while humans retain ownership of strategy and relationships. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Fundraising requires relationship-building, persuasive communication tailored to donor psychology and institutional priorities, and strategic negotiation—all human-centric activities that current AI cannot perform end-to-end with equal quality and 50% time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | Fundraising for scientific research (grant writing, proposal development, donor cultivation, networking) requires strategic judgment, relationship-building, and tailored persuasion that current AI cannot fully replicate end-to-end, though drafting portions can be assisted. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Fundraising for research carries significant barriers: funders expect human scientists to represent the work, institutional leadership and board oversight typically require human judgment, and legal/fiduciary responsibility for grant management rests with licensed professionals. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but funding agencies and reviewers expect PI-driven scientific vision, credibility, and personal accountability, creating organizational and trust-based friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI tools might draft emails or organize donor databases at low cost, but the core task of securing funds through persuasion and relationship management must be performed by human astronomers and development professionals, whose loaded cost far exceeds any AI assistance value. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Since AI can only handle a fraction of the task (drafting assistance), the human researcher's time for strategy, networking, and negotiation remains the dominant cost, keeping AI's cost advantage limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product reliably performs independent fundraising for scientific research in production today; this task requires human judgment about donor fit, institutional strategy, and authentic relationship cultivation that AI systems cannot replicate. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI writing tools can help draft grant proposal text, but no deployed product manages the full fundraising process—identifying funders, building relationships, negotiating terms—reliably in production. |
Collaborate with other astronomers to carry out research projects.
9CI 5–13 · exposure 0 · augmentation 63 · importance 4.4/5 · click for rater detail
Collaborate with other astronomers to carry out research projects.
9| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Astronomy remains a traditional, human-centric field where research teams are organized by human expertise and institutional relationships; adoption of AI collaboration tools remains minimal and supplementary rather than substitutive. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academic astronomy is adopting AI tools for data analysis and literature review, but the core collaborative research process among scientists remains largely unautomated and adoption here is slow. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI tools can assist with data analysis, literature review, and preliminary modeling within collaborative projects, but the core collaboration—proposal design, strategic decisions, and research direction—remains fundamentally human-driven. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist collaborating astronomers by summarizing papers, drafting shared documents, analyzing data, and facilitating communication, though the human collaborative process itself remains central. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Collaboration on research projects requires joint decision-making, creative problem-solving, interpersonal negotiation, and shared intellectual leadership—tasks that demand human judgment and cannot be meaningfully automated end-to-end by current AI systems. |
| Task automatability | claude-sonnet-5 | 1/5 | Collaboration on research projects is fundamentally a social, interpersonal, and intellectual coordination activity that current AI cannot perform end-to-end; it requires negotiating research directions, dividing labor, and building trust among human colleagues. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Research collaboration in academia is protected by institutional structures, peer-review authority, publication rights, and the implicit requirement that research direction and validation stem from credentialed human scientists who bear professional accountability for findings. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement for collaboration itself, but strong organizational and professional norms, trust-building, authorship credit, and grant/institutional structures create real friction against AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Replacing human astronomers' collaborative work would require either duplicating the entire research team (prohibitively expensive) or using AI to coordinate them (a marginal cost reduction that does not offset the cost of the humans themselves). |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI system performing this task as a substitute for human collaboration, so cost comparison favors human researchers by default. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product reliably coordinates multi-party research collaboration, manages conflicting scientific perspectives, or conducts the iterative human-driven research process that defines astronomical research partnerships. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product substitutes for scientific collaboration between astronomers; AI tools may support individual sub-tasks (data analysis, literature search) but no system manages collaborative research relationships. |
Direct the operations of a planetarium.
6CI 5–7 · exposure 0 · augmentation 38 · importance 2.5/5 · click for rater detail
Direct the operations of a planetarium.
6| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Planetariums are niche cultural institutions with low digital maturity and strong institutional preference for human direction; adoption of AI automation in this sector remains negligible. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and planetariums are small, under-resourced institutions with generally low AI adoption for operational management tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with scheduling, data monitoring, or technical diagnostics, but the core supervisory, decision-making, and visitor-interaction aspects of directing operations leave limited scope for productivity transformation. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with scheduling, content generation for shows, marketing, and administrative reporting, aiding the director without replacing the oversight role. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Directing planetarium operations requires real-time decision-making, visitor engagement, equipment troubleshooting, and adaptive responses to emergencies or technical failures that current AI cannot reliably manage autonomously without constant human oversight. |
| Task automatability | claude-sonnet-5 | 1/5 | Directing planetarium operations involves managing staff, scheduling shows, facility upkeep, and live audience engagement, none of which can be executed end-to-end by current AI systems.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Planetariums operate as public facilities with safety and liability requirements; a human director is typically legally and organizationally required to maintain control and accountability over operations, especially during live programming and emergencies. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Operational leadership involves staff supervision, safety, budget authority, and public-facing accountability that organizations require a responsible human to hold. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI inference and integration costs would still be dwarfed by the liability and oversight overhead required, plus the loaded wage of the skilled staff member needed to remain responsible for facility operations and visitor safety. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this managerial/operational role, so cost comparison favors the human entirely. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system currently performs end-to-end planetarium operations direction in production; this task demands situated judgment, safety responsibility, and coordination across staff and facilities that exceed today's autonomous capabilities. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product manages planetarium operations; this remains a human management and physical facility role. |
Mentor graduate students and junior colleagues.
5CI 5–5 · exposure 0 · augmentation 38 · importance 4.4/5 · click for rater detail
Mentor graduate students and junior colleagues.
5| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Mentorship remains a core, non-negotiable human function in academic science; despite digitization elsewhere, no sector has moved toward AI-led mentorship in production, reflecting deep cultural and institutional resistance. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Academic mentoring culture is slow to change and highly resistant to AI substitution, as it relies on personal rapport and reputation within research communities. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with administrative mentorship tasks (scheduling, document feedback, resource curation) but offers minimal augmentation to the core relational and pedagogical substance of mentoring itself. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help mentors draft feedback, suggest resources, or assist with career planning documents, but the core relational mentoring activity remains largely unaffected by AI tools. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Mentorship requires sustained relational development, emotional intelligence, personalized guidance calibrated to individual growth trajectories, and modeling of professional judgment—none of which current AI can deliver end-to-end with quality parity to human mentors. |
| Task automatability | claude-sonnet-5 | 1/5 | Mentoring requires personal relationships, career guidance, and tacit judgment based on trust and long-term interaction that current AI cannot replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Academic institutions and funding bodies (NSF, NIH) expect human mentorship as part of PI responsibility; career advancement and grant success criteria reward direct mentoring, creating strong organizational and contractual barriers to substitution. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Academic mentoring is deeply embedded in institutional norms, advisor-student relationships, and career sponsorship that require a human authority figure, though not formally licensed. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Mentorship is inherently labor-intensive and relationship-based; AI systems cannot replicate its value per unit cost, especially since supervisors would still need to oversee AI outputs and provide human mentorship anyway. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI product delivering this task, so cost comparison favors the human mentor by default. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product reliably performs comprehensive mentorship; AI chatbots lack the adaptive responsiveness, earned trust, and accountability relationships that define effective mentoring in academic contexts. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs mentoring of graduate students; this remains a human-led relational activity with no production substitute. |
Supervise students' research on celestial and astronomical phenomena.
4CI 0–7 · exposure 0 · augmentation 63 · importance 4.2/5 · click for rater detail
Supervise students' research on celestial and astronomical phenomena.
4| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Universities and research institutions remain fundamentally human-centered in pedagogy and governance; adoption of automation for core supervisory and mentoring functions is minimal and unlikely in the near term due to the mission-critical nature of research mentorship. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academia adopts AI tools for research assistance but has been slow to change formal mentorship and supervisory structures. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI tools could assist astronomers in preparing feedback on student code, suggesting data-analysis approaches, or summarizing research progress, thereby partially augmenting the supervisory workload, but the core judgment and mentorship remains human. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist in literature review, data analysis, coding, and drafting for the student's research, augmenting the supervisory process even though the human oversight role remains central. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Supervising student research requires mentorship, evaluation of judgment, course correction, and pedagogical decision-making tailored to individual learners—tasks that demand human insight, relationship-building, and contextual wisdom that current AI systems cannot reliably provide end-to-end. |
| Task automatability | claude-sonnet-5 | 1/5 | Supervising research requires mentorship, judgment on scientific direction, and personalized guidance that current AI cannot autonomously perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Academic institutions legally require faculty credentials and professional licenses for student supervision and research oversight; institutional governance, accreditation standards, and the fiduciary duty to students create hard barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Academic advising and mentorship typically require credentialed faculty status, institutional accreditation, and accountability for student outcomes, creating strong structural barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The cost of deploying AI to meaningfully supervise student research would exceed the loaded salary of an experienced astronomer, since the task demands accountability and institutional trust that cannot be offloaded to inference alone. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute for this supervisory role, so cost comparison favors the human entirely. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product reliably supervises research students in an academic setting; this requires continuous adaptive feedback, accountability for student development, and institutional authority that remains the exclusive domain of human faculty. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product acts as an autonomous research supervisor for students; this remains a human academic role. |
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.