Human-AI Collaboration Conditions for Environmental Product Declaration Reviews

This research explores conditions for trustworthy human-artificial intelligence (AI) collaboration in Environmental Product Declaration (EPD) reviews for construction. EPDs are essential for life-cycle-based sustainability assessment; however, their review is a manual, knowledge-intensive process performed by a few specialists. While AI integration can support review processes, establishing trust and accountability in AI-supported reviews, as well as their explainability to relevant actors, remains unexplored. The research addresses the critical challenge of building and maintaining trust and ensuring continuous explainability in human-AI collaboration within the EPD review process for the construction sector. We propose a systematic approach that identifies and classifies the essential steps and sociotechnical conditions in the EPD review process, and develops metrics and implementation strategies for human-AI collaboration through structured engagement with relevant stakeholders, including verifiers, manufacturers, and program holders. Using multi-level analysis and ethics-based auditing, it develops a comprehensive trust framework and quality metrics for human-AI collaboration in environmental reporting. This interdisciplinary methodology will result in a comprehensive integration program for AI-co-verified EPDs to ensure technical accuracy, compliance with expert knowledge, and social acceptance. It further lays the groundwork for a wider cross-sectoral comparison by exploring the key technical and socio-ethical conditions for responsible AI integration into EPD development. Results support advancing sustainable transitions in the construction sector while contributing to responsible AI implementation in knowledge-intensive processes.