Randomized trial evaluates agentic identity and delegation calculus effectiveness in human-AI systems, suggesting new pathways for multi-agent interactions.
BIT-CSKC-AIDC-001 presents a precanonical typed calculus for agentic identity and bounded delegation in human–AI and multi-agent systems. It separates authentication, context grant, mandate and operational admission; defines effective authority as an exact-use live intersection; and studies delegation attenuation, aggregate multidimensional budget conservation, explicit redelegation, epoch-based revocation, non-splicing, joint control and inert portability. The record preserves the complete research path from the initial P1 checker-disagreement hold through P1D diagnosis and P1R repair. P1R reran 37 frozen cases using two separated evaluator implementations, each executed twice: 12 TRUE, 17 FALSE and 8 UNKNOWN. All 15 locked metrics passed. Cross-evaluator disagreement, false admission, authority emission and real-world effects were all zero. F6 mechanized AIDC-T3 through AIDC-T8 and F5A-L01 through F5A-L15 using exact Lean 4.30.0. Two clean builds completed successfully, and all 21 audited declarations reported no axiom dependencies. AIDC is an evidence-bearing calculus only. It does not mint CVC/CARL authority and does not claim distributed protocol validation, production security, privacy or legal validation, field validation, standardization or formal-canonical status. Final status: PRECANONICAL INSTITUTIONAL CORE · BOUNDED PROTOTYPE PASS · CORE MECHANIZED · EXTERNAL REVIEW OPEN · PRODUCTION/FIELD NOT CLAIMED.
No takes yet. Share an insight, caveat, or question.
Quang Trịnh Bùi (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: