Updates from the HLG-MOS Projects

Generative AI

  1. The final report, spanning over 150 pages, has been completed. It includes six chapters: 1. Using, Implementing, and Developing; 2. Governing and Managing; 3. Mitigating and Monitoring; 4. Building Organizational Capability; 5. Preparing for the Future; and 6. Official Statistics and AI: Quo Vadis?
  2. Preparations are ongoing for the AI-readiness sprint which will be held on September 22nd with BSTN;
  3. Plans for a webinar to accompany the report's release.

Advanced Survey Cost-Effectiveness with Nonresponse Treatment (ASCENT)

  • Timeline: Final draft of the project planned for December 2025.

  • Work Package 1: Still finalizing the number of use cases.

  • Work Package 2: Will include 4 use cases.

  • Work Package 3: Expected to include 4–5 use cases.

  • Upcoming Presentation:

    • Beatrice Baribeau, DG of the Modern Statistical Methods and Data Science Branch (StatCan), will present the project in Geneva, November 2025.

    • The presentation will provide an overview of the final document.

  • Collaboration & Engagement: Working with Anders to leverage the Chief Methodologist network (including key ASCENT players via the dedicated subforum) to gather more information on responsive design.

 

Roadmap for Multi-Source Statistics (RAMSES)

During the reporting period (July–September 2025), the project made tangible progress, building on the outcomes of the first collaborative sprint. Thematic inputs were consolidated, and significant steps were taken in preparing the Handbook’s structure at the Chapter–Section–Subsection level. In parallel, the actual drafting of content has been launched by the thematic sub-teams, marking the transition from preparatory work to substantive writing.

Sprint Meeting

As planned, an online sprint was conducted in June 2025. The sprint was structured over two days, with sub-team focused discussions on Day 1 and a joint plenary session on Day 2 (25th of June).

The objective of the sprint was successfully achieved: to lay the groundwork for drafting the Handbook for Multi-Source Statistics (MSS) by identifying key priorities, defining topics and structure, and setting out a roadmap for drafting activities.

The sprint proved to be a significant milestone for the project. It enabled agreement on the main topics and priorities within each thematic area, while also consolidating initial ideas on the content to be included in the Handbook. Importantly, a number of cross-cutting issues were identified, along with approaches for addressing them in a coherent and harmonized way across themes. Beyond these technical outcomes, the sprint played a crucial role in strengthening collaboration and engagement among all sub-teams and the project management group. This reinforced alignment has created a solid foundation for the next phase of work, ensuring that the drafting process moves forward in a coordinated and effective manner.

Following the sprint, two plenary meetings were organized, and the four thematic sub-teams continued their work with active support from the project management team. These discussions focused on both organizational and technical aspects of the Handbook and have further advanced the harmonized drafting effort.

Use Case Questionnaire

The project also advanced on the use case dimension. Several use cases have already been collected through the dedicated questionnaire circulated among RAMSES Working Group members, with contributions received from Australia, Ireland, New Zealand, Portugal, Sweden, and Greece. Italy did not submit a questionnaire but presented its use case at the September plenary meeting; their formal questionnaire response is still expected.

The collected responses have been analyzed and synthesized into a summary report. This work is essential in ensuring that the Handbook does not remain purely conceptual but is firmly grounded in practical, real-world experiences. By documenting how different countries approach multi-source statistics in practice, the Handbook will reflect solutions, challenges, and lessons learned across contexts, thereby increasing its usability and relevance.

Handbook Drafting

The Handbook is conceived as a practical guide rather than a purely technical manual, with the ambition of serving as a reference document that can be applied in real statistical practice.

Drafting work has already begun within each thematic sub-team and is progressing in line with the agreed timeline. Authors are volunteers from each group, and their texts undergo consultation both within the sub-team and with the project management team. Once a sub-chapter draft is completed, it is reviewed by the responsible sub-team and the Project Manager, who also ensures overall consistency across the four thematic streams. The final review of the Handbook will be carried out by the UNECE, providing an additional layer of quality assurance and coherence.

Next Steps

With the sprint completed, the project will now focus entirely on the writing and review phase. The immediate priorities are to continue and intensify the drafting process, ensuring that progress remains steady and aligned across all thematic areas.

In practical terms, this means:

  • Advancing the work of the thematic sub-teams and the plenary group on drafting the Handbook.
  • Strengthening harmonization across themes to guarantee consistency in both content and style.
  • Systematically incorporating feedback from plenary discussions into the evolving drafts.

This phase will be decisive for the success of the project. By maintaining strong collaboration and a clear focus on coherence, the RAMSES team will ensure that the Handbook emerges as a comprehensive, user-oriented, and authoritative reference.


Electronic

Updates from the Modernization Groups

Group

Work

Status

Description

Blue-skies Thinking

Identifying Topics/Opportunities

BSTN Call discussions in recent calls included the following topics:

1. Trust in NSOs:

Key Themes:

  • Trust and confidence are foundational to the work of National Statistical Offices (NSOs).
  • The information landscape is evolving due to declining survey responses, misinformation, and generative AI.
  • Trust must be actively built and maintained through transparency, data stewardship, and clear communication.
  • Once lost, trust is difficult to regain.

Proposed Actions:

  • Launch a sprint to:
    • Review existing literature on trust.
    • Survey NSOs (e.g. via BSTN) on their trust-building strategies.
    • Identify common challenges and potential solutions.
    • Host an online session to gauge interest in a broader project.

2. Quantum Computing Risks

Key Points:

  • Quantum computing is transitioning from theory to practice.

  • It poses a threat to traditional cryptographic methods and data security.
  • NSOs may not be able to rely solely on vendors for quantum-safe solutions.
  • Quantum safety should be considered as part of broader trust initiatives.

Proposed Actions:

  • Evaluate centralized vs. federated data systems for quantum resilience.
  • Consider quantum computing as a sub-topic in trust-related projects.

3. AI-Readiness:
*Note: These discussions in BSTN let to the AI-Readiness Sprint on Sept 22nd.

Key Points:

  • There is a need for clear benchmarks to evaluate AI interpretation of statistical data. 
  • It is important to engage beyond official statistics (e.g., private sector, academia).
  • NSOs should consider internal and external organisational readiness.
  • AI often reads press releases, not databases – APIs and discoverability are key.
  • NSOs face internal data access challenges due to silos.
  • NSO websites may become less relevant as LLMs dominate information access.
  • Risks of misinterpretation by AI and the exponential spread of errors are serious concerns.

Applying Data Science and Modern Methods

Uncertainty Quantification Task Team

Progress Update

  • Five members have taken the lead in finalizing the report.
  • We aim to complete the final version by late September or early October.
  • We agreed to use Overleaf (LaTeX) to draft the report in PDF format.
  • Each chapter has a designated lead responsible for contributing content directly in LaTeX.
  • We will reach out to all members in late September to share the first draft and request their input.

Advancing Responsible AI Task Team and Responsible AI Training Program

Tasks have been allocated, and progress is expected to be being made - there is a need to follow-up with individuals to check progress. Existing modules are now considered to be transformed into learning materials. 

Capabilities and Communication


Work and Job of the Future - Generational diversity

Work on the document "Generational diversity: fostering an inclusive organizational culture in the NSOs' is ongoing. Team will aim to produce draft document by 10 November, to be submitted for the Workshop on the Modernisation of Official Statistics. Next team call is on 2 October.

Webinar on mental well-being

We have received an unprecedented number of registrations for the webinar—1,587 in total—driven largely by participants from Statistics Indonesia, while also attracting strong representation from UNECE member countries.

The webinar will take place on the morning of 30 September. We are currently finalizing the platform to be used and will send the invitation link to all registered participants next week.

Ethical management (Data and Business)

The Ethics document is currently being prepared for publication, which is expected next year.
The first meeting of the Organizing Committee for the Expert Meeting on Ethics in Official Statistics—scheduled to take place on 12–13 March 2026—has already been held, and invitations to the meeting will be sent to countries next week.

Expert Meeting on the Dissemination and Communication of Statistics

Preparations for the DissComm are on track, and we have begun receiving papers for the meeting, which will be posted on the website shortly. Meeting will take place on 5-7 November in Athens.

AI for official statistics - communication perspective

The Communication and AI series has attracted strong interest. Our first meeting brought together 51 participants and sparked a lively discussion. For the upcoming session on 24 September, we already have over 90 registered participants, showing clear engagement with the topic. The series will continue with four additional meetings in the months ahead.

Supporting Standards


 



 


Interoperability between DDI and SDMX for Data Integration task team

The last call of this group took place in July, and work will resume in September. Having seen a number of case studies/example, the next stage is probably to consider the question of mappings.

Statistical Architecture Framework task team

Work is ongoing. This activity has also branched out into some work looking at steady states.

Webinar on GSIM 2.0 

The webinar on v2.0 of the Generic Statistical Information Model took place on 2 September, and was very successful. It was attended by 134 people, despite not being widely advertised because of constraints on the capacity of MS Teams. The recordings of the presentations were made into youtube videos, as replacements for the training for the previous v1.2) of GSIM.

There have been significant calls for capacity building on GSIM from some member states, and the webinar helps to support some of these activities.

Standards discussion at the ISI Congress in the Netherlands SDMX Global Conference in Rome

Not started

This has been superseded by a proposal submitted to an alternate meeting that more of the group's members are attending. As of the last meeting, we were still waiting for the outcome of abstract selection from the organizers.

Revision of GAMSO

Although work on this was placed on hold for a few months, Chris is working on a proposal to be made to the task team members. If at all possible, this work needs to be concluded by the end of the year.

Survey for stakeholders

There has been a broader survey (done by the CES) of how much our outputs are being used by member states. Many of the Standards Group outputs scored very highly in the results, for example the GSBPM and GAMSO (which is no surprise). Hopefully the recent webinar on GSIM 2.0 will boost it's use in future surveys.

Additionally, the output from one of our recent activities on "Implementation Standards in the Context of GSBPM" also scored very well, which is good to see.