The goal of this panel session is to discuss requirements for a governance structure around responsible design, development and deployment of Machine Learning (ML) solutions in National Statistical Offices.

With the wide-scale utilization and adoption of Artificial Intelligence (AI), core principles such as ethics, privacy, equity, procedural fairness, trust, accountability, transparency, explainability, interpretability, reproducibility, robustness, quality and legality are far more important today than ever before. As a result, contextualization or the mapping exercise from model properties to system properties will be necessary as more and more complex Machine Learning or ML-based solutions are used in National Statistical Offices (NSOs) for various purposes, ranging from collection, integration, processing and dissemination.
Moreover, issues related to Responsible ML are complex and broad and encompass not only technical issues but also societal, legal, and ethical ones. One of the key components of Responsible ML systems is explainability, but other issues such as fairness, bias detection and the ability to objectively audit AI systems are also critical for successful applications and adoption of AI in NSOs. The session will consist of 3-4 invited speakers with the aim to discuss these topics, among others, e.g. uncertainty, robustness, … as they pertain to the Official Statistics. The session will be divided in two parts, with a series of brief presentations by the speakers followed by a panel discussion between the speakers, moderated by the chair.

Organiser: Dr Saeid Molladavoudi 

Chair: Dr Saeid Molladavoudi 

Speaker: Dr Barteld Braaksma  

Speaker: Ms Claire Clark

Speaker: Andrea del Monaco

Speaker: Marco Marini