Data, Statistics and AI Literacy for Meaningful Public Involvement – Rapid fire presentations

Authors: Elizabeth Waind (1), Katie Porter (1)

1 DARE UK

Abstract

There has been significant exploration of public attitudes towards the use of people’s sensitive data for research in the UK. This has been crucial for maintaining the social license for data research in the public benefit. However, artificial intelligence (AI) has introduced a new area of complexity, and applications to access sensitive data held in Trusted Research Environments (TREs) for AI research are on the rise.

Previous studies have engaged small groups of public contributors on specific AI research projects, or have explored attitudes towards sensitive data and AI in the health context. However, there is a distinct gap in our understanding of the views of the wider public towards the training of AI models on sensitive data held in TREs, and beyond the scope of health. In line with the PEDRI Good Practice Standard of Equity, Diversity and Inclusion, it is essential that we embed the views of the wider public in our work, and effective public dialogue is key to achieving this.

DARE UK (Data and Analytics Research Environments UK), in partnership with organisations from across the sector, is working to address this gap. In late 2025, we began with a pilot public dialogue with 22 members of a local community in Birmingham, England with limited knowledge of data research and AI. Although participants could acknowledge the benefits of AI trained on sensitive data, there was concern about data representativeness, AI misuse, regulation and oversight. There were also challenges maintaining a focus on TREs, with conversations drifting towards topics dominating the media discourse at the time, like job losses and ChatGPT.

This presentation will share key findings and lessons learned from the pilot project, and outline ongoing work and crucial next steps for enriching our understanding of public attitudes towards sensitive data and AI.

Authors: Elsie Makachiya (1), Tom Billins, Laura Venn, Eleanor Hall, Jack Adamson, Kshitij Nemade, Lola Ogunlusi, EE-SDE Core Public Advisory Group, Laura Clarke

1 Health Innovation East

Abstract

Introduction

The Eastern England Secure Data Environments (EE-SDE), part of the national NHS Research SDE Network, enables safe and secure access to sensitive, de-identified NHS health data for research, including the development of artificial intelligence (AI) and machine learning models. As the use of AI expands, there is a growing need for robust governance frameworks that can manage privacy risks while maintaining public trust. Within the EE‑SDE, the VISTA (Viable Implementation of SATRE, SACRO and Tools for AI) programme was established to develop practical tools supporting the safe training and release of ML models. Our talk showcases how the EE-SDE Core Public Advisory Group (CPAG) played a pivotal role in developing and evidencing the value of the AI checklist tool. Their input helped ensure the tool is understandable to non-specialists and emphasised the need for transparency, particularly in communicating how risks are managed when using AI models thereby strengthening public trust.

Methods

The VISTA AI Risk Assessment Toolkit was developed as a structured governance framework informed by established standards, including the Five Safes and national AI governance guidance. The development was initiated following a request from CPAG in April 2025, when, during the project’s introduction, the group raised the need for an AI policy to govern use of AI within the EE-SDE. Although the team moved away from developing a formal policy and instead adopted a checklist-based approach, development involved multidisciplinary collaboration across technical, governance, and operational stakeholders, alongside Patient and Public Involvement and Engagement (PPIE) through the EE‑SDE CPAG. Public contributors informed key design considerations, including transparency, accountability, and acceptable risk levels. This was all validated through the CPAG.

Results

The toolkit provides a standardised approach to identifying and mitigating disclosure risks associated with AI models trained on sensitive data. Key components include a researcher questionnaire, a structured risk assessment process, defined control measures, and a model registry to support transparency and oversight. Public input highlighted the importance of clear safeguards, demonstrable public benefit, and ongoing involvement in governance processes to maintain trust.

Conclusions

The VISTA AI Risk Assessment Toolkit applies a consistent, evidence-based approach to managing AI-related risks within SDEs. Embedding PPIE in its development strengthens alignment with public expectations and enhances transparency. The toolkit offers a scalable model for other organisations seeking to enable trustworthy AI while safeguarding sensitive data and maintaining public confidence.

Authors: Camilla Irvine-Fortescue (1), West Lothian Libraries

1 Heriot-Watt University

Abstract

In October 2023, the Public Engagement Team at Heriot-Watt University undertook a deliberative engagement programme in West Lothian (Scotland) to explore public perceptions of Artificial Intelligence (AI) and Robotics.

As part of our place-based engagement approach, we partnered with West Lothian Libraries to transform library spaces into sites of exchange between university research and local community perspectives. Libraries have been a vital entry point into these communities and since 2018 we have developed a sustained partnership that has been central to building trust around innovative and evolving technologies.

This collaboration has enabled the delivery of a wide range of community activity, from public talks for adults, to inclusive family-focused programmes. Within the AI and Robotics themed focus groups, we explored issues including trust in AI, digital lives and the future of medical technologies.

Building on the findings from the initial focus groups, we designed and delivered an AI & Robotics Youth Panel for young people aged 16–18. Through a series of workshops, participants developed AI Literacy, digital confidence and critical thinking skills. The programme provided behind-the-scenes insights into current research across technical disciplines, introduced the youth cohort to various researchers and culminated in a co-designed public panel event, shaped by the young people themselves.

Taking learnings from the Youth Panel, we have since continued community engagement work by collaborating with artists to co-design family-friendly AI Literacy workshops delivered across multiple library spaces. We have also worked with theatre makers and science communicators to design engaging activities and events exploring algorithmic manipulation. These activities aim to build public understanding and enable public trust and confidence in evolving AI technologies.

As universities, we have a responsibility to promote transparency, accessibility and meaningful dialogue around technological change. In this talk, we will share key learnings, co-design approaches and best practice from across these activities.

Authors: Tomas Ince (1), Sara Lewis (1)

1 Universal Care Plan Programme, NHS South West London Integrated Care Board

Abstract

The Universal Care Plan (UCP) is a London-wide digital care planning solution that enables people to record and share information about what matters to them, including care preferences, communication needs, and wishes for future treatment. New NHS App functionality allows people to directly create and edit personal information in their care plan, increasing transparency and involvement in their care.

Through the PEDRI Good Practice Standards Funding Call, the UCP programme delivered targeted public engagement workshops with underserved communities, including older adults who may be digitally excluded, people from ethnic minority backgrounds, and people living with multiple long-term conditions. Delivered in partnership with organisations including Age UK and supported by the UCP People with Lived Experience Group (PLEG), the workshops explored how to improve understanding of digital care planning, data sharing, consent, and the role of the NHS App using plain language, demonstrations, and discussion-based engagement.

This presentation will share practical learning from the workshops, including what helped build trust and confidence in digital health tools, common barriers to understanding health data and care planning, and how public feedback directly shaped communication materials and engagement approaches. The session will also reflect on the importance of trusted community partnerships, transparent communication, and meaningful lived experience involvement when engaging communities who are often underrepresented in digital health initiatives.

Subject to availability, we hope to involve a member of the UCP People with Lived Experience Group in the presentation to share reflections on accessibility, trust, and meaningful public involvement from a lived experience perspective.

Authors: Meanu Bajwa-Patel (1), Sima Rafiei

1 University of Warwick

Abstract

The Creating Healthy Jobs (CHJ) project aims to lay the foundations for improving job quality by building shared knowledge about the relationship between work and health. The project is organised across linked work packages combining stakeholder engagement, evidence synthesis, large-scale data analysis and knowledge mobilisation. It involves collaboration with employers, policymakers, government bodies, employer and employee representatives, public health organisations and members of the public.

Our proposed workshop focuses on Public Involvement and Engagement (PIE) within WP4, which uses large UK datasets, including Understanding Society and, subject to access approval, data from the Longitudinal Linkage Collaboration, to explore patterns in job quality and health over time, across sectors and demographic groups. Analysis and interpretation will be informed by a Lived Experience Advisory Panel (LEAP), made up of contributors with experience of poor-quality work and its effects on health and wellbeing.

A key challenge is how to involve public contributors meaningfully in a data-intensive work package. Longitudinal datasets, data linkage, governance and statistical analysis can be difficult to explain accessibly, yet lived experience is essential for shaping relevant questions, sense-checking interpretations and ensuring outputs are meaningful beyond academia.

The WP4 team will explore what LEAP contributors already understand about data, statistics, data linkage and governance, and use this to develop accessible training and discussion resources based on real CHJ examples. The workshop will share these emerging resources and reflect on the process of involvement, including: where public contributors can influence the research cycle; what has already been decided and what remains open to change; how academic and lived knowledge can be brought together; and what training, support and compensation are needed for meaningful engagement.

By the conference, we expect to have completed the early stages of LEAP engagement, so the session will offer an early reflective account rather than a final evaluation. We aim to co-present with LEAP contributors, potentially virtually, depending on availability.

Authors: Elizabeth Remfry (1), J Chaturvedi, S Markham, E Ford, M Ramasawmy

1 Queen Mary University of London

Abstract

Introduction

The increasing development of large language models (LLM) in healthcare research is taking place without patient and public involvement and engagement (PPIE). Part of the challenge is the lack of accessible educational resources to promote literacy around LLMs.

Methods

We employed a co-design approach with 6 PPIE contributors from Tower Hamlets, London to develop educational animations about LLMs. We conducted 7 pilot sessions that included hands-on interactive activities to develop scripts and storyboards for the animations. Animations were external validated by additional PPIE groups and experts before professional animation production.

Results

Two accessible 2-minute animated videos were successfully co-designed in English and Bengali. The first explains LLMs in the context of daily life and the second on LLMs in healthcare research. Both animations use a factory analogy to describe LLMs functionality while avoiding anthropomorphisation that could mislead audiences about LLMs inner workings and capabilities. Both animations are freely available for reuse.

Discussion

The co-design process allowed us to prioritise diverse PPIE voices throughout development, ensuring that animations are appropriate for a wide audience. Challenges included incorporating sometimes conflicting perspectives and achieving balanced portrayals of LLM benefits and risks. We hope these animations will support PPIE contributors to have a say in how LLMs are used in healthcare and research.