In an era where data-driven research and analytics are critical, ensuring the integrity, privacy, and safety of research outputs is crucial. The Statistical Disclosure Control – Reducing Barriers to Outputs from TREs (SDC-REBOOT) Interest Group (formerly Evaluation of Automated Output Checking and Artificial Intelligence (AI) Model Risk Assessment Group) brings together experts, professionals, and organisations passionate about advancing the field of Output Statistical Disclosure Control (OSDC) and Machine Learning (ML) model risk assessment within Trusted Research Environments (TREs).
Output checking is a core function for operators of Trusted Research Environments that ensures that any research leaving the environment does not pose any privacy risks. However, the process is time and resource-intensive, creating a bottleneck for the timely release of research findings that has sparked interest in automated approaches to assist this process.
This Interest Group builds on the accomplishments of the DARE UK Driver Project, Semi-Automated Checking of Research Outputs (SACRO), as well as the DARE UK Sprint Exemplar Project, Guidelines and Resources for AI Model Access from Trusted Research Environments (GRAIMatter). These projects set the stage for innovative approaches to ensure the trustworthiness of research outputs, spurring significant interest both in the UK and on the international stage.
Built on a foundation of collaboration, knowledge sharing, and problem-solving, the focus of the interest group is on driving user adoption of output-checking tools by identifying and removing blocks to practice. These include a combination of practical resources, agreeing on approaches to evaluating tools’ trustworthiness and efficiency, and aligning the conceptual framework and taxonomy of outputs across the field. The group will also consider the critical task of risk assessment for ML models. The group’s ultimate goal is to establish a community of expertise that can support researchers and TREs in their mission to safeguard research outputs, minimise disclosure risks, and advance the adoption of automated tools and frameworks.
Projected Outputs
Initially, SDC-REBOOT’s work will be divided into four parallel projects supported by the creation of open-source repository governance structures. These threads will:
- Align the conceptual framework and taxonomy of outputs.
- Create user adoption roadmaps and resources.
- Enable TREs to deploy and evaluate automated OSDC tools.
- Address risk assessment of machine learning models.
Participation and Collaboration
SDC-REBOOT includes members from diverse groups and organisations, including project leaders, TREs, computer scientists, AI/ML researchers, public engagement practitioners, and more. As the community evolves, representation will expand to include organisations ranging from charities to National Statistics Institutes.
Ways of Working
SDC-REBOOT meets monthly via face-to-face, hybrid, and online meetings. Subgroups focusing on specific threads may meet more frequently and host targeted meetings to engage a wider audience. The Chair and co-chairs review progress biweekly, and members are encouraged to propose new activities or changes to the existing focus. Where necessary, the group may request support for a research software technician to assist in implementing various resources and to host support sessions for researchers and TREs.
Group Co-Chairs
- Professor Jim Smith, The University of the West of England, Bristol
- Professor Felix Ritchie, The University of the West of England, Bristol
- Jackie Caldwell, Public Health Scotland
- Becca Wilson, University of Liverpool & PI DataSHIELD
- Simon Rogers, NHS National Services Scotland
Documents
For enquiries, please send an email to sacro.contact@uwe.ac.uk.
