Updates from the HLG-MOS Projects

Generative AI

During May and June 2024, the Generative AI Project held two plenaries as well as its regular task team meetings. More specifically, the plenary in May focused on the ongoing work by OECD.AI on the development of generative AI across the wide spectrum of policy activities. The meeting emphasized the importance of robust AI governance and policy frameworks and the OECD AI Policy Observatory's role in monitoring AI developments. It also addressed the importance of trustworthy AI. In this context, there were several discussions among the members on the role of NSOs to the OECD’s AI catalog, a possible AI Risk management framework tailored to official statistics, as well as further collaboration opportunities with the OECD in this space. The Swiss Statistical Office presented the SwissBot to the project in the second plenary (June). Finally, as regards the task teams, work is ongoing to develop the outline of the final report as well as to ensure comprehensive coverage for the drafting.

Open-Source Software Project

The Open-Source Software Project (OSSP) has been conducting regular sub-team and plenary meetings. The two sub-teams are two sub-groups are "Governance and Maintenance" (chair Kate Burnett-Isaacs) and "Repositories and Discoverability". Presentations and discussions have occurred concerning the Awesome List for Official Statistics, as well as Open-Source Licenses. 

Drafting work has begun on the Project Report.

Plans are in place to organize a sprint in September (date TBD).

Updates from the Modernization Groups

Blue-skies Thinking




Identifying Topics/Opportunities

At the April call (occurring after the last EB update) Eric Anvar gave a presentation on the concept of "Data Mesh", which involves the domain-orientated decentralization for analytical data i.e., instead of having analysis of data occurring at a central point with one team, the analysis processes are pushed upstream to local teams with domain expertise. The group discussed options for exploration of the Data Mesh concept under the work of the HLG-MOS.

The May meeting discussed the process of pitching ideas at the dedicated mid-year BSTN session for such. The mid-year session is a time dedicated to hearing ideas that pitches believe may be worth turning into a project or activity.
The group agreed to continue with existing practices (now documented on the wiki) i.e., members of modernisation teams are given the chance to present ideas at the BSTN mid-year pitching session, and feedback is provided by the group.


The June BSTN call will hear pitches for projects/activites at the June call.

Pitches so far include:

  • Data Integration
  • AI Governance at NSOs
Future of NSIs

Three out of five of the chapter workshops have been completed on the Strategic Papers sections. So far Vision, History and Role, and External Challenges have been discussed. The working team is aiming for a deadline of 15 August for the chapter groups to produce their first drafts.

Applying Data Science and Modern Methods





General Overview

Overall, the two established Task Teams are progressing well, and we have been looking at developing a couple of Applied Data Science Collaboration proposals.

 In terms of the Applied Data Science Collaboration proposals, we have a number of countries interested in developing proposals for three areas – PETs, Nowcasting and EO / mobile data / Geospatial. We are currently considering how to best put forward these proposals for consideration for inclusion in the ADSaMM Group’s programme (either 2024 or 2025).

We are also exploring a proposal from Statistics Bureau of Japan to host an in person meeting in Tokyo in October. The aim is to begin work on these as part of the ADSaMM programme in September. If any other HLG MOS members are interested in taking part please contact alison.baily@ons.gov.uk and jo.green@statistics.gov.uk for more information

Task Team #1: Uncertainty Quantification

The Uncertainty Quantification (UQ) Task Team have held 3 meetings since the last update. Here are some of the highlights:

  1. Task Team Meetings and Presentations
    • 2024-05-07: Mohammed completed part 2 of his presentation on Conformal Prediction.
    • 2024-05-21: Siu-Ming presented on “Uncertainty quantification for machine-generated (MLg) statistics using the bootstrap.”
    • 2024-06-04: We started a Google document. We discussed work scope, timelines, deliverables, the creation of a communication channel (Teams or other) and assigned subgroup tasks.
  1. Subgroups Meetings (past and upcoming)
    • 2024-06-03: The "Traditional Methods" subgroup met to plan work, discuss scope, timelines, and deliverables.
    • 2024-06-21: The "Conformal Prediction" subgroup will meet to discuss the literature review and next steps.
    • 2024-06-24: The "Traditional Methods" subgroup will meet to review completed tasks, provide updates, and discuss next steps.


Task Team #2: Advancing Responsible AI.

In the Responsible AI Task Team, the following progress has been made:

  1. Responsible persons were and will be assigned for each module, with deadlines set to ensure timely completion. The modules will be structured into specific, practical blocks tailored to executives, statisticians and data scientists and will include a compliance checklist and a standard template for use cases. Both predictive and generative AI will be covered, addressing their distinct implications and challenges.
  2. Legal expertise will be needed to understand the implications of the EU AI Act and other international frameworks, involving lawyers from NSIs.
  3. Various training methods identified: webinars, workshops and online platforms. It was also agreed to prepare webinars as training modules, with recordings available on the UN training platform. Additionally, there are plans for an in-person workshop during the HLG MOS annual meeting in November in Geneva.
  4. Responsibilities for modules are confirmed (unfortunately still some names missing), with drafts to be completed by August, reviews finalized by September and preparations made for a November workshop, alongside a series of webinars and workshops, including that already mentioned in-person session at the HLG MOS annual meeting. Deadlines for each module should be set in June (there is an online table where each person responsible for a module should enter their deadlines).


Capabilities and Communication


Work and Job of the Future - Extended work on Generic Growth Model

All Task Teams have regular monthly meetings.

Message to the countries to request wider distribution/use of the Generic Growth Model was sent to the countries at the end of May together with the surveys prepared by it's sub-teams (employer branding and evaluation of blended working). Deadline for responses is 30 June. 

Data Analytics

Team had a couple of successful meetings, and prepared structure of the paper on "Enhancing National Statistical Offices through HR Analytics. Key Considerations and Benefits".

 The plan is to prepare first draft of the paper for the Expert Meeting on HRMT on 14-16 October.

Evaluation of blended (hybrid working)

Survey on the evaluation of blended working was sent to the countries at the end of May. Deadline for responses is 30 June.

Employer branding

Survey on employer branding was sent to the countries at the end of May. Deadline for responses is 30 June.

Ethical management (Data and Business)

Work on the Reference Book is progressing, and the main structure has been agreed upon. Many new team members joined this team after the Workshop on Ethics.

Team is planning to have a draft of the Reference Book ready by the end of this year, and will report on the progress during the HRMT meeting in October.

Ethics Workshop 2024 (26-28 March 2024)

Workshop report was published on the website: https://unece.org/statistics/events/Ethics2024

Workshop on HRMT OC

Invitation was sent to the countries on 15 May, with the deadline for submitting abstracts at the end of June. Deadline for registrations is 16 September: https://unece.org/statistics/events/HRMT2024

The programme of the meeting will cover the following topics:
• ‘Employer of Choice’ brand development
• Training/learning and development
• Integration, inclusion and ethics
• Evaluation of blended/hybrid working and data analytics

Team will meet at the beginning of July to review submitted contributions and to decide on the meeting timetable.

AI for official statistics - communication perspective

Continuing to gather and showcase country experiences, working on populating the "AI for communicating official statistics" guidelines


Supporting Standards

 

CES endorsement of GSIM 2.0

We received endorsement of version 2.0 of GSIM from the Conference of European Statisticians

ModernStats World Workshop

  • The abstracts for the ModernStats Word Workshop in October are published on our webpage, and we are seeking registrations: https://unece.org/statistics/events/MWW2024
  • This year’s workshop may be an opportunity to take a strategic look at where we are with our work on standards, and how we wish to focus our attentions in the future, corresponding to the session on what statistical production should look like in 2025 and beyond, which might also have a groupwork element, in addition to presentations/panel discussion (we are still formulating the timetable).
  • While this workshop can help determine the longer-term direction of our work, there is only a short period of time between this workshop and the HLG workshop, so it cannot be the place where 2025 activity proposals are conceived.
  • For this reason, the process for discussing 2025 activity proposals has already started, though previous workshop participants were also invited to suggest ideas.

Consideration of the interplay between AI and standards

We group has had several discussions on this topic, and is seeking collaboration with AI developers to find concrete examples how AI and standards may be relevant to each other, whether standards (especially relating to semantics) can help AI to interpret data and make generated results more explainable, and also whether AI could be helpful in the development of standards and statistical production.

Revision of GSBPM and GAMSO activity

  • Work on the revision of GSBPM is continuing as planned, currently examining feedback received on the Design phase of GSBPM.
  • We have found that our initial suggestion about renaming phase 4 of GSBPM from “collect” to “acquire” has been confusing to speakers of latin-influenced languages, for whom the word acquire has connotations of purchasing/procurement of data.

SDMX-DDI-GSBPM activity

Work on finalizing the SDMX-DDI-GSBPM report is close to completion, but recently slowed down, so renewed efforts have been made to finalise the document.

Common Statistical Data Architecture

We had previously concluded that the CSDA activity would need to be pushed back to next year due to the lack of a leader for this activity. However, we may have found a volunteer to lead the work on CSDA, and will reach out to discuss their plans and timelines for this work. This should enable the activity to commence this year.

Core Ontology for Official Statistics


It has been decided not to start this work until the GSBPM revision is finalised (next year).

65th ISI World Statistics Congress 2025

The proposal we submitted to the 65th ISI World Statistics Congress 2025 in The Hague has been accepted. The proposed session will discuss how implementation standards can be used together with conceptual ModernStats models to improve interoperability at technical, semantic and organizational levels, and how they can be leveraged to build statistical production pipelines that are metadata-driven, semantically consistent and reusable.