Introduction to AI for Official Statistics

Course Leader

Chris Lam 

Statistical production units and methodologists of NSIs.

Managers and decision-makers who need to understand AI opportunities and risks for official statistics.

• Participants should have a sound command of English, enabling them to actively participate in discussions and make short interventions during the sessions.

• The course is intended for statistical production units and methodologists from National Statistical Institutes (NSIs), as well as managers and decision-makers who need to understand the opportunities and risks of artificial intelligence in the context of official statistics.

• Course participants are expected to have a firm working knowledge of statistical methods (inference, estimation, probability)

• Course participants are expected to have a working knowledge of (statistical) data analyses with Python

This course shall democratize the ethical development and   use of AI

• Understand the fundamental concepts of AI and machine learning and their relevance for official statistics.

• Identify potential applications of AI across the statistical production process. 

• Recognize the data requirements and methodological challenges of applying AI in official statistics. 

• Apply basic AI techniques to official statistics datasets in guided exercises. 

• Assess governance, ethical, and legal issues linked to AI adoption in NSIs (including compliance with GDPR and the EU AI Act). 

• Evaluate opportunities and risks of AI for modernization of official statistics. 

• Formulate a roadmap or project idea for responsible AI use in their institution 

Foundations of AI and Machine Learning

• AI in context: definitions, history, main paradigms.

• Types of AI methods: supervised, unsupervised, reinforcement learning.

• Relevance for official statistics: examples from NSIs and international organizations.

• Tools and environments: overview of Python/R ecosystems for AI.

Data, Models, and Workflows

• AI data requirements: quality, bias, missing values, representativeness.

• Common models: regression, classification, clustering, dimensionality reduction.

• Neural networks and deep learning: basic concepts and applications.

• Practical examples with official statistics datasets

Applications in Official Statistics 

• Use cases. 

• Interactive examples. 

Governance, Ethics, and Regulation 

• Governance frameworks. 

• Ethical aspects: fairness, accountability, transparency, explainability. 

• Risk assessment: bias, discrimination, misuse of AI models. 

• Regulatory frameworks: 

• Institutional policies: integrating AI responsibly into statistical production. 

Practical Integration and Future Outlook 

• AI in production pipelines: integration with existing IT/statistical systems. 

• Capacity building: skills and training needed. 

• Emerging technologies. 

By the end of the course, participants will have a clear understanding of what artificial intelligence (AI) is, how AI models are trained, and how to balance their potential benefits with associated risks.

They will feel confident in identifying custom use-cases for AI within the context of official statistics and in developing simple prototypes.

Participants will gain insight into how AI is reshaping statistical production through automation, predictive analytics, and the integration of non-traditional data sources such as big data, administrative records, and satellite imagery.

For National Statistical Institutes (NSIs), adopting AI responsibly goes beyond the technical aspects. This course will emphasize the importance of aligning AI applications with ethical principles, strong governance frameworks, and relevant legal regulations — including GDPR, the EU AI Act, and the UN Fundamental Principles of Official Statistics.

Mornings are dedicated to executive briefings and deep dives, afternoons to guided hands-on lab sessions.

• Presentations and lectures (40%)

• Exchange of views/experiences (10%)

• Guided Hands-on Lab sessions (50%)

Required Reading 

None

• Register for free to the Microsoft Power Platform
  https://make.powerapps.com/ (with work login)
  used during lab afternoons 1 & 2

• Register for free to the Google Colaboratory
  https://colab.research.google.com/ (with personal login)
  used during lab afternoons 4 & 5

Chris Lam (CBS Netherlands)

Marco Puts (CBS Netherlands)

Yvonne Gootzen (CBS Netherlands)

Florian Dumpert (Statistisches Bundesamt - Destatis)

Herbert Kruitbosch (Rijksuniversiteit Groningen)