O*NET Plus packs decades of O*NET-based measures into a single, curated toolbox — improving consistency and psychometric backing for task-level research; it exposes gaps in capturing digital and AI-related tasks and calls for standardized, AI-specific extensions and linkages to dynamic data sources.

Main Finding

The authors introduce "ONET Plus," an extended content model for occupational description that systematically organizes and aggregates validated measures derived from ONET. They synthesize the literature using ONET-based measures, identify widespread reuse and redundancy, evaluate psychometric evidence, and compile a curated, centralized database of ONET Plus measures that have substantial validation support. The work both clarifies measurement choices for researchers and proposes directions for expanding occupational data to better capture contemporary and emergent job characteristics.

Key Points

• Motivation
  • ONET has become a dominant source for occupational characteristics, but researchers frequently adapt or combine ONET descriptors to measure constructs O*NET does not directly capture (e.g., job complexity, emotional labor).
  • This practice enabled much research but generated many overlapping, inconsistently validated measures.
• Contributions
  • Introduces ONET Plus: an extended organizing framework that expands the original ONET content model to include commonly constructed constructs and task/skill dimensions used in applied research.
  • Systematically reviews studies that use O*NET-derived measures and evaluates the psychometric evidence supporting those measures.
  • Curates a "one-stop-shop" database of O*NET Plus measures that have considerable empirical validation, reducing duplication and improving reproducibility.
  • Provides guidance and recommendations for future expansions and standardization efforts.
• Problems identified
  • Proliferation of redundant measures with varied levels of validation.
  • Gaps in O*NET’s original model for emerging job features (e.g., digital/AI-related tasks).
  • Need for clearer conventions on measure construction, reporting, and validation.

Data & Methods

• Literature synthesis and curation
  • The authors surveyed the body of studies that adapt, combine, or derive measures from O*NET descriptors.
  • They collected and catalogued measures used across studies, focusing on constructs built from ONET content that extend beyond ONET’s default labels.
• Evidence evaluation
  • For each curated measure, the authors reviewed available psychometric evidence (e.g., internal consistency/reliability, construct validity, convergent/discriminant validity, criterion-related validity where available).
  • They highlighted measures with "considerable evidence" and flagged those with limited or mixed support.
• Output
  • Assembled a centralized database (the O*NET Plus repository) of validated measures and documentation to facilitate reuse and standardization.
• Notes on scope/limitations
  • The paper operates as a review and curation effort rather than presenting new primary field data; methods emphasize systematic review, cross-study synthesis, and psychometric appraisal.
  • The review is tied to O*NET content and therefore is U.S.-centric in occupation coverage and descriptors.

Implications for AI Economics

• Better measurement of task content and skills
  • O*NET Plus provides more granular, validated constructs that improve task-level measurement—critical for task-based models of automation and AI exposure.
  • Standardized, validated measures reduce measurement error and make cross-study comparisons and meta-analyses more reliable.
• Improved estimates of AI exposure and complementarity
  • Researchers can map O*NET Plus measures (e.g., cognitive complexity, social interaction, manual dexterity, digital/analytical task components) to AI capability profiles to estimate which occupations/tasks are automatable, augmentable, or complementary to AI.
  • The curated measures facilitate more nuanced predictions of displacement risk, wage effects, and labor demand shifts by distinguishing task types that interact differently with AI.
• Policy and retraining implications
  • With clearer measures of task bundles and skill deficits, policymakers can better target reskilling programs and labor-market interventions to tasks most affected by AI (e.g., data annotation, machine supervision, creative synthesis).
• Research agenda recommendations
  • Add AI-specific and digital-skill dimensions to O*NET Plus (e.g., programming/data work, model supervision, prompt engineering, digital literacy, human-in-the-loop coordination).
  • Develop and validate measures of meta-cognitive, creative, and socio-emotional skills that likely confer resilience or complementarity with advanced AI.
  • Link O*NET Plus to alternative, frequently updated data sources (online job ads, crowdsourced task ratings, platform task logs) to track rapid technological change.
  • Standardize crosswalks between O*NET Plus constructs and AI capability taxonomies (e.g., LLM/vision/robotics capability matrices) to enable consistent mapping from AI benchmarks to labor impacts.
  • Promote longitudinal linking of O*NET Plus measures to occupational outcomes (employment, wages, job transitions) to better estimate causal impacts of AI adoption.
• Limitations and cautions for AI-economics applications
  • ONET and ONET Plus are occupation-based; heterogeneity within occupations and firm-level differences mean task exposure can vary—researchers should, where possible, combine occupation-level measures with worker- or firm-level data.
  • The U.S.-centric coverage may limit direct applicability in other labor markets without careful adaptation or recalibration.

Overall, O*NET Plus strengthens the measurement toolkit for AI economics by consolidating validated occupational measures and pointing to concrete extensions and data linkages needed to study AI’s labor-market effects with greater precision.