Logic for Interactive Consistency: Paraconsistency, Norms, and Accountable Large Language Models
Proceedings of the AAAI Symposium Series · 06.11.2026 · 10 (1) · 10-17
DOI: 10.1609/aaaiss.v10i1.43385 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 06.11.2026. Verfasst von Thomas Arnold und 1 weiteren Autor:innen.
Abstract (Originalsprache)
The turn toward pluriversal and paraconsistent logic to remedy shortcomings in current AI alignment methods seeks, among other things, to represent and respect conflicting belief systems, including those around ethical and social norms. The constitutional AI reliance on a consensus filter, to take one alignment example, collapses different perspective to reach a single, majority-derived view. While acknowledging that paraconsistent logic shows a weakness in methods of alignment, this paper seeks to push further logical standards for putting LLM's to the test. The real time challenge of human-AI interaction, including the demands of natural language instruction and coordination, suggest the importance of at times retaining classical logic's form of consistency (and use of negation) for successful social collaboration. In this paper we elaborate upon these challenges to form a supplemental, not necessarily conflicting, position for holding generative and agentic AI to stronger logical account.
Zitieren
APA
Arnold, T., & Scheutz, M. (2026). Logic for Interactive Consistency: Paraconsistency, Norms, and Accountable Large Language Models. Proceedings of the AAAI Symposium Series, 10(1), 10-17. https://doi.org/10.1609/aaaiss.v10i1.43385
Vancouver
Arnold T, Scheutz M. Logic for Interactive Consistency: Paraconsistency, Norms, and Accountable Large Language Models. Proceedings of the AAAI Symposium Series. 2026;10(1):10-17. doi:10.1609/aaaiss.v10i1.43385
BibTeX
@article{Arnold2026,
author = {Arnold, Thomas and Scheutz, Matthias},
title = {Logic for Interactive Consistency: Paraconsistency, Norms, and Accountable Large Language Models},
journal = {Proceedings of the AAAI Symposium Series},
year = {2026},
volume = {10},
number = {1},
pages = {10--17},
doi = {10.1609/aaaiss.v10i1.43385}
}
EUVIMED