A Quantitative Analysis of 21st Century Skills: A Case of Semantic and Psychometric Overlap
Citation. Teyssier-Roberge, G., Gagnon, J., Tremblay, S., & Hodgetts, H. (2025). A quantitative analysis of 21st century skills: A case of semantic and psychometric overlap. International Journal of Selection and Assessment, 33(4), e70030. https://doi.org/10.1111/ijsa.70030
Summary. Teyssier-Roberge and colleagues examine whether the large field of “21st century skills” names distinct constructs or proliferates overlapping labels. Using a literature survey, latent semantic analysis of definitions, and hierarchical clustering of assessment instruments, they find substantial semantic and psychometric overlap across 40 identified skills. The paper argues for more parsimonious skill language, higher-order competency framing, and behavior-based assessment rather than isolated skill labels and broad self-report measures.
Key claims
- The literature search identified 40 different 21st century skills, 446 definitions, and 457 unique assessment measures, illustrating substantial construct proliferation in the field.
- Latent semantic analysis showed high semantic overlap among many skill definitions, consistent with a jangle-fallacy risk where different names imply distinct constructs without enough conceptual separation.
- Hierarchical clustering of assessment instruments found several skills commonly measured with the same instruments, including an interpersonal cluster of teamwork, leadership, communication, situation awareness, decision-making, and collaboration; an executive-skills cluster including goal-directed persistence, inhibition, task switching, shifting, and cognitive flexibility; and an intrapersonal cluster including task initiation, organization, self-management, and emotional control.
- The authors stop short of proposing a final taxonomy, but argue that 21st century skills should be treated as higher-order general competencies rather than long lists of context-specific behavioral manifestations.
- For HR and selection practice, the paper recommends standardizing terms, reducing assessment task impurity, shifting from isolated skill labels to competency-based hiring and training frameworks, and using dynamic behavior-based assessments rather than relying primarily on self-report.
Evidence & limitations
- The study is a proof-of-concept narrative review and quantitative overlap analysis, not a final taxonomy or validation study of a new skills framework.
- The review used Google Scholar, English-language sources, and literature published between 1990 and 2020; this broad coverage helps capture academic and gray literature but does not remove selection and indexing limitations.
- Semantic overlap was estimated with latent semantic analysis of selected representative definitions, and psychometric overlap was estimated by clustering shared assessment instruments; both methods reveal redundancy but do not by themselves establish the best higher-order structure for workplace use.
- The paper is directly useful for AI-era workforce measurement because it cautions against building evergreen skills architectures from fashionable labels whose definitions and measures are not construct-distinct.