Compliance-by-Construction Argument Graphs: Using Generative AI to Produce Evidence-Linked Formal Arguments for Certification-Grade Accountability
Summary: arXiv:2604.04103v1 Announce Type: new
Abstract
High-stakes decision systems increasingly require structured justification, traceability, and auditability to ensure accountability and regulatory compliance. Formal arguments, commonly used in the certification of safety-critical systems, provide a mechanism for structuring claims, reasoning, and evidence in a verifiable manner. At the same time, generative artificial intelligence (GenAI) systems are becoming integral to decision-support workflows, assisting with drafting explanations, summarizing evidence, and generating recommendations.
However, current deployments often rely on language models as loosely constrained assistants, which introduces risks such as hallucinated reasoning, unsupported claims, and weak traceability. This paper proposes a compliance-by-construction architecture that integrates Generative AI with structured formal argument representations.
Proposed Architecture
The proposed approach treats each AI-assisted step as a claim that must be supported by verifiable evidence and validated against explicit reasoning constraints before it becomes part of an official decision record.
The architecture combines four key components:
- Typed Argument Graph Representation: Inspired by assurance-case methods, this representation organizes the claims and evidence in a structured format.
- Retrieval-Augmented Generation (RAG): This component drafts argument fragments grounded in authoritative evidence, ensuring that the information is reliable and relevant.
- Reasoning and Validation Kernel: This kernel enforces completeness and admissibility constraints, which are vital for maintaining the integrity of the argumentation process.
- Provenance Ledger: Aligned with the W3C PROV standard, this ledger supports auditability by tracing the origins and transformations of claims and evidence.
System Design and Evaluation
The paper presents a system design and an evaluation strategy based on enforceable invariants and worked examples. The analysis indicates that deterministic validation rules can effectively prevent unsupported claims from entering the decision record while allowing GenAI to accelerate argument construction.
Conclusion
This innovative approach aims to enhance the accountability and regulatory compliance of high-stakes decision-making processes. By integrating Generative AI with structured formal argumentation, the proposed compliance-by-construction architecture enhances the reliability and traceability of claims and evidence, ensuring a more robust decision-making framework.
The results of this study could pave the way for future advancements in the certification of safety-critical systems, ultimately contributing to improved safety and compliance standards across various industries.