Metraipolis › archive › dc_0c8bc7b43b4f

The Open Measurement Layer: Toward Shared, Auditable Infrastructure for Evaluating Educational AI (2026) · DOI 10.35542/osf.io/becxs_v1

Source
Metraipolis archive
Original
https://doi.org/10.35542/osf.io/becxs_v1
License (as archived)
CC-BY-4.0
Archived
2026-10-07T12:16:34.639Z
SHA-256 of the archived text
adb2fb6b31437b0dc7c6a6e639514029cb9cde9a1b206c40bbaa380d6694054d
Stable URL of this page
https://metraipolis.com/documento/dc_0c8bc7b43b4f

How to cite

The Open Measurement Layer: Toward Shared, Auditable Infrastructure for Evaluating Educational AI (2026) · DOI 10.35542/osf.io/becxs_v1 — Metraipolis archive, https://doi.org/10.35542/osf.io/becxs_v1 (CC-BY-4.0); archived copy: https://metraipolis.com/documento/dc_0c8bc7b43b4f (Metraipolis archive dc_0c8bc7b43b4f, sha256 adb2fb6b31437b0dc7c6a6e639514029cb9cde9a1b206c40bbaa380d6694054d)
BibTeX
@misc{dc0c8bc7b43b4f,
  title = {The Open Measurement Layer: Toward Shared, Auditable Infrastructure for Evaluating Educational AI (2026) · DOI 10.35542/osf.io/becxs_v1},
  howpublished = {Metraipolis archive},
  url = {https://doi.org/10.35542/osf.io/becxs_v1},
  note = {License CC-BY-4.0; archived copy https://metraipolis.com/documento/dc_0c8bc7b43b4f; sha256 adb2fb6b31437b0dc7c6a6e639514029cb9cde9a1b206c40bbaa380d6694054d},
  year = {2026}
}

Archived text

Reproduced under CC-BY-4.0. Attribution: Metraipolis archive, https://doi.org/10.35542/osf.io/becxs_v1. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Articolo recente con licenza aperta: scheda dai metadati Crossref (posted-content).
Titolo: The Open Measurement Layer: Toward Shared, Auditable Infrastructure for Evaluating Educational AI
Autori: Kumar Srivastava
Pubblicato: 2026-08-12 · editore: Center for Open Science
DOI: https://doi.org/10.35542/osf.io/becxs_v1
Licenza dell'articolo: CC-BY-4.0

Abstract (testo degli autori, CC-BY-4.0):
Generative AI has severed the assumption that underwrites most educational assessment: that a strong product indicates a strong learning process. The field's most considered response is to move assessment toward richer, process-based evidence gathered over time. I argue that this response is correct but incomplete, because process-based assessment at scale is AI-mediated assessment, and the measurement of educational AI itself remains overwhelmingly closed, vendor-controlled, and difficult to reproduce. An ecosystem that has opened its content and its tools while leaving judgment proprietary has not yet opened the layer on which the credibility of the others depends. This paper defines an open measurement layer comprising five components: shared construct ontologies, open evaluators, contamination-controlled benchmarks, limitations-first documentation, and interoperable instrumentation. Openly licensed artifacts that instantiate each component show the layer to be buildable today. The paper closes with open problems, including governance of held-out evaluation sets, documentation standards for evaluators, and incentive design, whose resolution will require coordination among institutions rather than action by any single organization.

Fonte: https://api.crossref.org/works/10.35542%2Fosf.io%2Fbecxs_v1 — metadati Crossref, consultati il 2026-10-07. Interrogazione: «open science infrastructure».