Mapping Curricula to Occupational Work Activities Using a Shared DWA Ontology: Evidence from the Course-Skill Atlas and O*NET

Technical Summary: Mapping Curricula to Occupational Work Activities Using a Shared DWA Ontology

Problem Statement
Universities face increasing pressure to demonstrate how academic programs prepare students for data-intensive and AI-driven occupations. However, there is limited evidence regarding how documented curriculum signals can be systematically compared with occupation-specific work activities at the program level. Existing challenges include the difficulty of translating academic language (syllabi, course descriptions) into occupational taxonomies (job tasks) due to differences in vocabulary, granularity, and documentation conventions. Direct text matching between curriculum features and occupational task statements is often fragile and lacks traceability. Furthermore, while competency frameworks exist, they do not empirically show how these expectations are represented in documented curriculum signals across institutions. The core problem is the lack of a traceable, shared-ontology approach to align large-scale curriculum data with task-derived occupational requirements without relying on superficial text matching.

Methodology
The study develops a curriculum analytics framework that maps curriculum-side signals from the Course-Skill Atlas (CSA) and occupation-side task requirements from O*NET into a shared Detailed Work Activity (DWA) space using official DWA identifiers as the common ontology.

• Data Sources:
  • Curriculum Side: The CSA dataset, which infers skill signals from over three million U.S. higher education syllabi. The study uses institution-field-year records and a wide score matrix of activity-level features.
  • Occupation Side: The O*NET 30.3 database, specifically the Task Statements, Tasks to DWAs, and DWA Reference files. The focal occupation is Data Scientists (SOC 15-2051.00).
• Vector Construction:
  • Curriculum Vector: CSA feature labels are mapped to official O*NET DWA IDs. Raw curriculum weights are calculated as the syllabus-count weighted mean of CSA scores, then L2-normalized to represent relative emphasis.
  • Occupation Vector: O*NET task statements are linked to DWA IDs via the Tasks to DWAs file. Counts are aggregated by DWA ID, then L2-normalized to create a task-derived occupation profile.
• Alignment Metrics:
  • Cosine Similarity: Measures the directional similarity of relative emphasis between curriculum and occupation vectors.
  • Coverage@K: A nominal measure indicating whether the top K occupation-side DWAs appear in the curriculum vector (presence/absence).
  • Skill Gap Index (SGI): An occupation-weighted deficit measure that captures the magnitude of under-emphasis for high-priority occupation DWAs.
• Validation Strategy:
  • An 8 × 8 validation matrix compares eight curriculum groups (e.g., Computer Science, Mathematics, Business) against eight occupation profiles (e.g., Software Developers, Statisticians, Accountants).
  • Rank-based positive-control checks verify if theoretically expected curriculum groups (e.g., Accounting for Accountants) rank highest in cosine similarity.
  • Sensitivity analyses test robustness against variations in coverage thresholds and weighting schemes (e.g., binary vs. task-link-count weights).
  • Cohort-level robustness checks aggregate curriculum groups into "Core" (technical/quantitative) and "Applied" (business-oriented) cohorts.

Key Results

• Focal Case (Computer Science vs. Data Scientists):
  • Full Nominal Coverage: All 16 task-derived Data Scientist DWAs were present in the Computer Science curriculum vector (Coverage@10, @20, @30 = 1.000).
  • Weak Relative-Emphasis Alignment: Despite full coverage, the Cosine Similarity was low (0.114), and the Skill Gap Index was high (0.886).
  • Emphasis Gap: The largest gaps occurred in high-weight occupation DWAs, such as "Apply mathematical principles or statistical approaches" (Job Weight: 0.459 vs. Curriculum Weight: 0.031). The study concludes the gap is one of emphasis, not absence; Data Scientist activities are present but receive relatively low weight in the documented curriculum signals compared to their importance in the occupation profile.
• Validation Matrix:
  • The framework successfully recovered theoretically plausible patterns. Computer Science ranked highest for Software Developers and Database Architects; Mathematics ranked highest for Statisticians and Operations Research Analysts; Accounting ranked highest for Accountants.
  • Notably, Mathematics showed the highest cosine similarity with the Data Scientist profile, reinforcing the quantitative nature of the O*NET task-derived profile.
• Robustness: Results remained stable across alternative weighting schemes and coverage thresholds. The Core cohort (CS, Math, Engineering, Physics) consistently showed higher alignment with technical occupations, while the Applied cohort aligned better with business-oriented roles.

Key Contributions

1. Traceable Shared-Ontology Framework: The study introduces a method for curriculum–occupation alignment that uses official DWA IDs as a shared key, avoiding direct text matching and ensuring auditability from raw data to final metrics.
2. Distinction Between Coverage and Emphasis: The analysis reframes curriculum–occupation gaps. It demonstrates that a curriculum can fully cover the types of activities required by an occupation (nominal coverage) while significantly under-emphasizing them relative to the occupation's own priorities (weighted alignment).
3. Diagnostic Tool for Program Assessment: The framework provides a reproducible, data-driven approach for examining how documented curriculum signals correspond to occupational work activities, serving as a starting point for curriculum review rather than a definitive measure of teaching quality or student learning.

Significance and Claims
The paper claims that this DWA-ID alignment approach offers a structured, traceable way to communicate curricular priorities to stakeholders (employers, policymakers, accreditation bodies). By separating nominal coverage from relative emphasis, the framework helps identify specific "emphasis gaps" where curriculum signals may not reflect the intensity or centrality of occupational work activities.

The authors explicitly state that DWA-ID alignment should not be interpreted as a direct measure of teaching quality, student learning, or occupational readiness. Instead, it serves as a diagnostic tool to examine how documented curriculum signals (syllabi) communicate learning activities. The study emphasizes that the observed gaps reflect the representation of activities in course documentation, which may differ from enacted instruction or student mastery. The framework is presented as a reusable approach for evidence-informed communication and curriculum analytics in higher education, particularly for emerging, interdisciplinary fields like data science.