ET-ACC: A Conceptual Framework for Explainable and Trustworthy Automated Compliance Checking in the AEC Industry
Abstract
Automated Compliance Checking (ACC) is now a central activity in AEC. It supports design assurance, regulatory compliance, and digital delivery from early design through to review. Yet many current ACC systems remain opaque. Complex rule sets and AI-assisted reasoning often produce results that are difficult to inspect or explain, so transparency, interpretability, and user trust remain limited. This paper introduces ET-ACC, a multi-layer conceptual framework for Explainable and Trustworthy Automated Compliance Checking. The framework evaluates BIM and IFC models against pre-formalised, machine-readable rules. It is organised into five connected layers—Data, Rule, Reasoning, Provenance, and Explanation—supported by a cross-cutting Human-in-the-Loop and Trust Evaluation component. The Rule Layer manages computable rules and logic structures. The Reasoning Layer combines symbolic reasoning with AI-based support. The Provenance Layer records how data, rules, and reasoning steps relate, which enables complete traceability. The Explanation Layer then turns these traces into role-aware, human-readable outputs for designers, regulators, and other stakeholders. A Human-in-the-Loop and Trust Evaluation component strengthens accountability through expert review and structured transparency assessment. The framework is built on design-science methodology and literature across ACC, explainable AI, and compliance modelling. It offers a structured foundation for next-generation systems that place interpretability and auditability at their core. Full implementation is planned as future work. The framework sets a clear blueprint for a practical compliance-checking platform, referred to as ACE-X, that will operationalise the ET-ACC architecture and support more transparent, safe, and accountable digital building regulation processes and automated design assessment workflows across diverse project types and scales.