Trustworthy Artificial Intelligence in the Context of Metrology

TL;DR

The paper addresses how to ensure trustworthy artificial intelligence within metrology by integrating uncertainty quantification into ML-based measurement models. It articulates three core research strands at NPL: explainable AI to promote transparency and traceability, a metrology-focused framework for evaluating ML uncertainty, and best-practice guidelines for training data in safety-critical marine navigation. It also discusses AI certification, regulatory implications, and challenges posed by large language and foundation models. The work aims to enhance transparency, traceability, and accountability in AI-enabled metrology with practical impacts across engineering, healthcare, and environmental monitoring.

Abstract

We review research at the National Physical Laboratory (NPL) in the area of trustworthy artificial intelligence (TAI), and more specifically trustworthy machine learning (TML), in the context of metrology, the science of measurement. We describe three broad themes of TAI: technical, socio-technical and social, which play key roles in ensuring that the developed models are trustworthy and can be relied upon to make responsible decisions. From a metrology perspective we emphasise uncertainty quantification (UQ), and its importance within the framework of TAI to enhance transparency and trust in the outputs of AI systems. We then discuss three research areas within TAI that we are working on at NPL, and examine the certification of AI systems in terms of adherence to the characteristics of TAI.

Figures (3)

• Figure 1: Self-validating thermocouple with protective sheath. Image courtesy of CCPI Europe.
• Figure 2: Metrology requirements for machine learning uncertainty evaluation.
• Figure 3: Areas of good practice for training data preparation.