National Statistician’s Data Ethics Advisory Committee minutes: 2 June 2026

Published:

Attendees

Members 

  • Helen Boaden (Chair)
  • Ray Harris
  • Stephen John
  • Dr Rochelle Tractenberg 
  • Dr Orla McBride
  • David Walker
  • Joy Mbaulu
  • Aden Kwok

Secretariat

  • Nicola Shearman
  • Steph Jacobs
  • Heather Bryant
  • Samuel Williams

Other attendees

  • Jennifer Arkell (item 2)
  • Katrina Tyrrell (item 2)
  • Arron Maspero (item 2)
  • Iva Spakulova (Item 2)
  • April Maynard (Item 3)
  • Dani Evans (Item 3)
  • Meghan Elkin (Item 3)
  • Alex Lambert (Item 3)

Apologies

  • Dr Sarah Markham
  • Vanessa Cuthill

Minutes

1. Minutes and matters arising from the previous meeting

  1. The Chair welcomed members, including new members Joy Mbaulu, David Walker and Aden Kwok, to the meeting.
  2. Minutes from the 3 March 2026 meeting were approved, subject to minor clarifications regarding the statutory role of the Research Accreditation Panel.

2. Survey Assist within the Transformed Labour Force Survey (TLFS) (NSDEC(26)21)

  1. The Chair introduced the item on Survey Assist, noting the wider challenges facing labour market surveys, including declining response rates and the need to modernise data collection approaches. The purpose of the discussion was to explore the early-stage use of AI within surveys and to identify any ethical considerations arising from its application.
  2. Jennifer Arkell and Katrina Tyrrell presented findings from initial testing on Standard Industrial Classification (SIC) data using “Survey Assist”, an AI‑enabled prototype designed to improve the quality of SIC and Standard Occupational Classification (SOC) data within the TLFS. The approach uses a closed large language model to generate tailored follow‑up questions where initial survey responses do not provide sufficient detail to enable higher‑level coding.
  3. The Committee noted findings from testing with approximately 1,000 respondents, which showed that a substantial proportion of respondents received large language model (LLM) generated followup questions, with evidence that these improved codability. Both open and closed question formats were trialled, with open questions supporting more detailed responses but introducing some ambiguity, while closed questions were quicker for respondents to complete but offered less flexibility.
  4. The Committee noted a number of ethical and operational considerations, including the potential impact on respondent burden, the quality and consistency of AI‑generated questions, and the risk of bias in responses and subsequent coding. Members also discussed the importance of ensuring that classifications are clearly understood by respondents and expressed in accessible language.
  5. The research team outlined findings from ongoing public engagement work, which includes checking whether respondents are comfortable interacting with AI. Early findings suggested general acceptance of AI where it is used for public benefit, but highlighted concerns around data security, transparency and trust. The Committee emphasised the importance of clear communication to respondents, including how AI is being used and the provision of appropriate opt-out options where necessary.
  6. Members sought clarification on data security arrangements, including the use of a closed model and the handling of data during processing. The team confirmed that while data are not retained by the model, information is transmitted securely during processing, with mitigations in place.
  7. The Committee welcomed the innovative nature of the work and the careful, iterative approach being taken. Members noted the early-stage nature of the project and that this represents the beginning of a longer development journey. The committee encouraged continued attention to respondent experience, transparency, and potential impacts on public trust, particularly as development progresses.
  8. The Chair invited the team to return to the Committee at a later stage once further evidence and testing outcomes are available.

Action:

Survey Assist team to return to NSDEC at a later stage as the work develops.

3. Safety During Childhood Survey (SDCS) update (NSDEC(26)22)

  1. The Chair introduced the item, noting the sensitive nature of the research and the significant ethical considerations involved in collecting data on experiences of abuse among young people. The purpose of the discussion was to consider proposals for the second field test and, in particular, the approach to safeguarding and anonymity.
  2. Dani Evans provided an update on the first field test, which took place between January and February 2026. The test included a screener questionnaire only and received positive feedback from respondents, with no complaints raised. A response rate of approximately 19% was achieved prior to reminder activity, and a final response rate of 31.4%.
  3. The Committee was informed that a second field test is planned for July 2026, involving a larger sample and the inclusion of the full questionnaire. This will include follow-up questions to capture further detail on experiences of abuse. Proposed changes include an updated incentive process and the introduction of additional reminder letters to improve response rates.
  4. The research team outlined the proposed approach to anonymity. While the survey is designed to be anonymous, operational processes relating to reminders and incentives mean that identifiable information may be temporarily held during the fieldwork period. The Committee noted that, although there is no intention to link survey responses to identifiable data, this remains technically possible until the end of the field period.
  5. Alex Lambert, Chief Safeguarding Officer, provided context on the organisation’s safeguarding responsibilities. He noted that ONS has a clear policy and moral responsibility to act where a risk to wellbeing or life is identified. He advised that not acting on information falling within the remit of the safeguarding policy would represent a departure from standard practice and therefore constitute a significant organisational decision requiring senior-level approval.
  6. He further emphasised that safeguarding intervention is not automatic in any case. Where information is collected under the Statistics and Registration Service Act (SRSA) 2007, decisions would be escalated to the Permanent Secretary or National Statistician, who would make the ultimate judgement on balancing confidentiality against wellbeing and safety considerations.
  7. The Committee noted that this raises a complex ethical consideration. The team explained that the approach has been informed by extensive engagement with stakeholders, including children, survivors and safeguarding organisations. This indicated that anonymity and no safeguarding action based on survey responses will likely increase participation, improve honesty in responses, and reduce the risk of harm to respondents due to unwanted referrals or perpetrators becoming aware of them taking part in the survey. This approach also gives respondents the autonomy and control to choose how and when to meet support which was deemed important in the research.
  8. Members discussed the balance between safeguarding responsibilities and the public value of producing robust data on an under researched issue. Members emphasised the importance of clearly framing this as a considered ethical judgement, rather than a decision to take no action. Members also emphasised the importance of a clear stakeholder and communications strategy, including how the approach would be explained publicly and whether key partners (e.g. NSPCC) would support the position.
  9. The research team outlined the proposed differential incentive approach for the second field test where those aged 16 to 19 continued to be offered £15, whilst those aged 20-25 would be offered £20 with the aim of improving response for those aged 20-25.
  10. The Committee noted that while this approach may support representativeness, it could give rise to perceptions of unfairness. Members also discussed the importance of ensuring that incentives do not act as undue inducement, particularly given the sensitive subject matter. It was noted that the prominence and positioning of incentive information within communications should be carefully reviewed, ensuring that participation is clearly presented as voluntary and not driven by financial incentive.
  11. The research team also presented proposed updates to survey materials, including shorter invitation and reminder letters and the introduction of an information leaflet containing key information and signposting to support services. Detailed privacy information will be made available online rather than included in full within invitation materials.
  12. The Committee welcomed the efforts to simplify communications and reduce respondent burden, while emphasising the importance of ensuring that key information, including the voluntary nature of participation, the use of anonymity, and the availability of support, is clearly communicated. Members also highlighted the potential reputational risks associated with the safeguarding approach and the importance of stakeholder engagement.
  13. The Committee acknowledged the complexity and sensitivity of the issues and the thoroughness of the work undertaken to date. Members supported the continued development of the survey, subject to ongoing consideration of safeguarding implications, appropriate senior oversight of the proposed approach, and clear, proportionate communication to respondents.

Action:

Research team to review the presentation and prominence of incentive information within survey communications.

Action:

Research team to escalate safeguarding policy decision to the Permanent Secretary.

Action:

Research team to provide NSDEC with an update following the next field test, including further detail on the safeguarding approach and respondent experience.

4. NSDEC Ethical Principles Review (NSDEC(26)23)

  1. The Chair introduced the item, noting the importance of ensuring that the Committee’s ethical principles remain relevant in the context of evolving data practices and emerging technologies.
  2. Steph Jacobs provided an update on the planned review of NSDEC’s ethical principles. The Committee was informed that the review will begin with a discovery and evidence‑gathering phase, drawing on existing ethical frameworks from across government, academia and international organisations.
  3. The Committee noted that the review aims to assess whether the current principles remain appropriate or require updating to reflect developments in areas such as artificial intelligence, data linkage and public trust. The intention is to ensure that the principles remain sufficiently broad and durable, while continuing to provide clear and practical guidance.
  4. Members discussed the importance of producing principles that are technology‑agnostic and future‑proof, while still being accessible to non‑specialist audiences. The Committee emphasised the need to balance clarity with flexibility, ensuring that principles can be applied consistently across a wide range of use cases.
  5. Steph Jacobs outlined the proposed timeline for the review, with evidence gathering to take place over the summer, followed by the development of recommendations for the Committee to consider later in the year. Subject to agreement, a draft set of updated principles will be developed with the aim of publication in 2027, following appropriate governance and approval processes.
  6. The Committee welcomed the proposed approach and emphasised the value of engaging both Committee members and external experts as part of the review process.

Action:

Secretariat to lead the evidence gathering phase for the review of NSDEC’s ethical principles and provide an update to the Committee at the September meeting.

5. Any other business

  1. Nicola Shearman provided an update on the recruitment process for the National Statistician, noting that interviews had been rescheduled and that an announcement is expected in due course. The Committee noted the importance of continuity in leadership during this period.
  2. The Chair informed members that she intends to step down in early 2027. The Committee heard that recruitment for a new Chair will take place over the summer, with the intention of appointing a successor with appropriate expertise, including in data and technology. Members noted that the process will be conducted in line with standard public appointments procedures.
  3. The Committee discussed plans for an in‑person workshop later in the year. The Secretariat confirmed that a meeting is scheduled for September and that further details will be shared with members in advance.
  4. The Committee welcomed the updates and noted the importance of maintaining strong governance and continuity as the Committee’s work evolves.
  5. The next meeting will be Wednesday 30th September 2026, 2 Marsham Street, London.