Paper 2 — Normative Memory Under Interpretive Degradation
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This project develops a comparative framework for studying normative memory under interpretive degradation in multi-agent language-model systems. It asks whether Constitutional Founding Texts (CFTs), treated as long-horizon governance artifacts, preserve behavioral constraints more robustly than matched procedural baselines when explanatory continuity degrades across relay, paraphrase, institutional handoff, and substrate change. The project addresses a central governance problem in contemporary multi-agent AI: as systems persist across deployment contexts, model versions, and organizational transitions, the rationales grounding their constraints may become unavailable, unstable, or too costly to reconstruct. In response, the project distinguishes recoverable rationale, local intelligibility, behavioral constraint retention, behavioral drift, and substrate opacity, and uses this vocabulary to formulate a baseline-gated comparative hypothesis over normalized retention profiles. The theoretical contribution is a formal account of degradation, retention, and persistence, supported by external mechanism hypotheses drawn from cultural memory, symbolic boundary maintenance, and nuclear semiotics. The empirical contribution is a falsifiable research design comparing narrative constitutions with matched procedural and role-card baselines under relay, ablation, substrate shift, and partial documentation loss, with particular emphasis on system-level drift in multi-agent settings. The expected outcome is a more precise account of when and how governance artifacts remain behaviorally operative after explanatory continuity weakens, and a foundation for subsequent work on constitutional regime drift, reconstitution, and long-horizon governance under opaque computational conditions.
Fields of Study & Subjects
Ethics and Social Impacts of AI
Cybernetics and Technology in Society
Multi-Agent Systems and Negotiation
Publication details
- Journal / source
- Open Science Framework
- Publication date
- 1 January 2026
- Language
- EN