This report summarizes the key outcomes of the Seventh Machine Learning in Geotechnics Dialogue (7MLIGD), themed “Trustworthy Data-Centric Geotechnics,” held during the Third Workshop on the Future of Machine Learning in Geotechnics (3FOMLIG) in Florence, Italy. As artificial intelligence (AI) and machine learning (ML) models are increasingly applied to site characterization, risk assessment, and geotechnical design, the dialogue examined how data-centric geotechnics can be developed within a framework of technical reliability, transparency, and regulatory compliance. Three invited contributions addressed complementary dimensions of trustworthiness. One focused on the evolving role of AI in engineering decision-making and its implications for professional responsibility and human-machine collaboration. Another articulated the foundational principles of trustworthy data-centric geotechnics, emphasizing explainability, robustness, fairness, and privacy as prerequisites for effective human-AI teaming. A third examined privacy-preserving methodologies, discussing their mathematical guarantees, accuracy-privacy trade-offs, and compatibility with legal requirements. The subsequent panel discussion extended these themes to practical challenges, including dataset validation, uncertainty quantification, model interpretability, traceability of data provenance, and the readiness of existing design codes to accommodate AI-assisted workflows. Attention was given to the implications of emerging European regulatory and standardization initiatives, notably the EU Artificial Intelligence Act, for defining risk levels, accountability structures, and acceptable performance criteria. The dialogue concluded that trustworthy data-centric geotechnics requires systematic integration of rigorous validation protocols, transparent modelling practices, privacy safeguards, and governance mechanisms to ensure reliability, public safety, and professional accountability.
Vitale et al. (Fri,) studied this question.