Updates from the HLG-MOS Projects

Generative AI

  • Kick-off meeting to be held on 21st Feb. 
  • Participants from Australia, Azerbaijan, Canada, France (Bank), Ireland, Italy, Italy (Bank), Sweden, Switzerland, UK, BIS and OECD
  • Relevant actions for coordination 
    • Discussion with One-Stop-Shop LLM lead (Sweden) 
    • OECD-BIS SDMX-AI workshop
    • Statistics Centre - Abu Dhabi (SCAD)
    • CES Plenary seminar on AI to be held in June (co-lead: Germany and Canada)

Statistical Open-Source

  • Kick-off meeting to be held on 4th March
  • Organizations signed up: Australia, Canada (Infrastructure), Italy, New Zealand, UK and OECD

Updates from the Modernization Groups

Blue-skies Thinking


Identifying Topics/Opportunities

CONTINIOUS

  • We reviewed the methods of working for the group, and agreed to keep the regular brainstorming sessions, given their value in generating ideas and getting a sense of interest in such.
  • The possible existence of an AI bubble was raised at the last call, which could be explored in some capacity under HLG-MOS. 
  • ONS is interested in doing an activity on the "Future of NSIs", which would involve writing a white paper on predictions, challenges, and opportunities upcoming for NSIs.
    A meeting will be held on the 26th of February to determine the structure and possible contents of the paper.

  • Several pitches were made last year regarding data collection (e.g. Smart Surveys from CBS, Removing Barriers from ISTAT). Given the up coming Expert Meeting on Statistical Data Collection, there could be opportunities for collaboration with the steering committee and the pitchers.
Digital Twins

IN PROGRESS

No significant update since that given in September 2023.

The team has been setting up a similar proof of concept for the approach of using simulation to model field activity and inform design decisions, however this doesn't involve any real time feedback loops (i.e. it is not a digital twin). 

Applying Data Science and Modern Methods


  • 2023 cycle: The Data Editing Task Team's deliverable paper is published. The Modeling and Responsible AI Task Teams are in their final stages with their deliverables and are soon to be published as well.
  • 2024 cycle: We have 23 unique participants signed up, mostly for 2 task teams: a) Advancing Responsible AI in Statistical Offices: Bridging Knowledge and Practice and b) Beyond Point Predictions: Ensuring Reliability in Official Statistics through Uncertainty Quantification.

The third task team, "Business Case for Graph Modeling and Graph Databases Support across the Granular Data Lifecycle," has not gathered interest as there are only a few people signed up, and it is considered not to proceed. We are scheduling kickoff meetings for the 2 task teams and will perform scoping to identify concrete deliverables. 

Capabilities and Communication


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

IN PROGRESS

Three sub-teams met separately, and had productive discussion in their respective areas. Work on the Generic Growth Model for complex organizational themes proposal will be carried on by the main Future of work task team. As agreed earlier, together with other surveys of this team, we will send request to the countries to share with us experiences with growth model and examples. 

We will work on preparing letter that should be sent to the countries to provide examples/experiences of using growth model. 

It was also proposed that each of the sub-teams will prepare examples of how their work fits in the generic growth model.

Data Analytics

IN PROGRESS

We had good first meeting of this sub-team with colleagues from Infrastructure Canada and Statistics Canada leading the discussion. Colleagues from Statistics Canada promised to share their internal documents related to data analytics, and to organize show and tell presentation in the coming months. They will also invite colleagues from ONS UK, ABS and Statistics New Zealand to join this team, as they are collaborating with them on the same topic. Team will be working on the paper exploring the advantages of using data analytics and how NSOs can embark on this journey.

It was suggested that we should try to plot data analytics journey in the Generic Growth Model.

Evaluation of blended (hybrid working)

IN PROGRESS

This sub-team agreed that they will start sharing materials relevant to this topic, Cathy already shared some useful ABS documents. Afterwards sub team will work on the questions for the survey. It was agreed that survey should go out to the countries at the end of May/beginning of June, so there is enough time to collect responses and analyze results to be able to present them at the Workshop on HRMT in October 2024.

Employer branding

IN PROGRESS

This sub-team also had productive meeting and considering to prepare a paper and a survey that will be sent to the countries. It was  prepared an outline of the work of this sub team that was circulated to everyone by e-mail and will be discussed during the next meeting on 14 March. In the meanwhile they will also be gathering documents related to employer branding.  We will also attempt to map this work to Generic Growth Model.

The plan is to present output of this sub-team at the HRMT workshop in October and Modernization Workshop in November.

Ethical management (Data and Business)

IN PROGRESS

The team had a meeting. We can observe progress on the Reference book on ethics: https://docs.google.com/document/d/1CzfMZuTOERZm4DqGHB4-h_Km_TTyuEDuofsqL6oMb9M/edit?usp=sharing 

Ethics Workshop 2024 (26-28 March 2024)

IN PROGRESS

The work is on the track. We are finalising agenda and analysing received contributions.

Workshop on HRMT OC

IN PROGRESS

First meeting will be organized by the end of February.

AI for official statistics - communication perspective

IN PROGRESS

The group met and had a round table discussion on topics participants wish to work on and outputs (all)

  • Providing examples and case studies demonstrating the use of AI and LLM tools for communication purposes in official statistics would be beneficial for both team members and the wider statistical community.
  • LLMs could introduce inaccuracies and biases in public discourse, National Statistical Offices (NSOs) should dedicate efforts to continuously increase the statistical literacy of their users.
  • AI has the potential to support communication objectives through automated and faster visualization, text editing, and maintenance tools.
  • Are there any standardized prompts or repositories available? Writing a prompt as a statistics communicator persona may yield different results compared to when a simple user makes the same request.
  • Utilizing AI for audience segmentation and message targeting, for message creation and insights derived from user interactions, user and media monitoring, and sentiment analysis.
  • Are there existing policies or guidance on the use of AI tools, taking into consideration accuracy and confidentiality issues?
  • Incorporating a "show and tell" component in each meeting where someone presents an example or case study.
  • The task team is a safe space, where participants can present any cases, challenges and involve in discussions.

ACTION POINT: provide an example of AI application for communication in our organisations, focusing on the things that worked well, and those that didn’t (!).  

Supporting Standards

 



To date (as of 19th Feb), the Supporting Standards Group can report the following:

  • The work on version 2.0 of GSIM was presented to the CES bureau, which endorsed its presentation to the CES plenary session in June.
  • For the ModernStats World workshop, the group has defined the scope of the topics to be discussed.
  • Arrangements are being made for a side-meeting to the smart metadata conference in April: This includes discussions about how to scope and structure those discussions (on the complementarity of SDMX and DDI, and the role of transformation languages like VTL and SDTL).
  • We are seeking a leader for the activity on Common Statistical Data Architecture (CSDA). – Suggestions are welcome!
  • Work is being undertaken to finalise the work of the activity on linking GSBPM to SDMX and DDI.
  • Work on the revision of GSBPM is continuing as planned. (There are 2 phases left to consider comments about, then GAMSO comments will be reviewed.)




  • No labels