EUVIMEDEUVIMED entdecken
Crossref

From Metacognitive Monitoring to Control: Resource Allocation in Neuro-Symbolic AI

Michael Clopton, Sven Thijssen

Proceedings of the AAAI Symposium Series · 06.11.2026 · 10 (1) · 84-96

DOI: 10.1609/aaaiss.v10i1.43394 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 06.11.2026. Verfasst von Michael Clopton und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Metacognitive monitoring assesses the reliability of an intelligent system’s decisions, while metacognitive control uses that information to regulate subsequent action. We introduce locked verification, a controlled paradigm that holds object-level perceptual digit predictions and downstream review conditions fixed while varying only the monitoring signal. Control is operationalized as fixed-budget resource allocation. On semi-synthetic radiation-dosimetry records with CNN and Qwen3-VL front-ends, symbolic conditioning improves monitoring and allocation under a common, gain-aligned policy, while task accuracy remains fixed across confidence conditions. For CNN, M-Ratio increases from 0.666 to 0.976; at a 25% review budget, error capture increases from 0.387 to 0.448 against an oracle ceiling of 0.469, with the same ordering under Qwen3-VL. Across the evaluated settings, the within-setting association between M-Ratio and allocation skill is consistently positive under the primary mapping (median Spearman’s ρ = 0.786; 0.734 without Raw), with complementary monitoring measures showing the same general pattern. Structural controls show that cell-level monitoring and review-unit allocation are distinct, mapping-dependent quantities. The results provide controlled evidence that symbolically informed monitoring can provide downstream value for fixed-policy resource-allocation control.

Zitieren

APA

Clopton, M., & Thijssen, S. (2026). From Metacognitive Monitoring to Control: Resource Allocation in Neuro-Symbolic AI. Proceedings of the AAAI Symposium Series, 10(1), 84-96. https://doi.org/10.1609/aaaiss.v10i1.43394

Vancouver

Clopton M, Thijssen S. From Metacognitive Monitoring to Control: Resource Allocation in Neuro-Symbolic AI. Proceedings of the AAAI Symposium Series. 2026;10(1):84-96. doi:10.1609/aaaiss.v10i1.43394

BibTeX
@article{Clopton2026,
  author = {Clopton, Michael and Thijssen, Sven},
  title = {From Metacognitive Monitoring to Control: Resource Allocation in Neuro-Symbolic AI},
  journal = {Proceedings of the AAAI Symposium Series},
  year = {2026},
  volume = {10},
  number = {1},
  pages = {84--96},
  doi = {10.1609/aaaiss.v10i1.43394}
}
Kostenlos starten →