What’s Covered?
This resource forms part of the AI Ethics and Governance in Practice series developed by the Public Policy Programme at The Alan Turing Institute. The workbook zeroes in on explainability—how and why decisions made by AI systems can be understood by the people they impact.
The content is grounded in the UK public sector context and connects directly with earlier national guidance developed with the Office for AI and Government Digital Service. It builds on that foundation with expanded, updated insights and practical exercises aimed at putting ethics into action.
Key features include:
- A conceptual breakdown of explainability: what it means, why it matters, and how to do it responsibly.
- The Explainability Assurance Management Template for guiding structured implementation.
- Case studies (like AI in children’s social care) and fictional scenarios (e.g. The Smith Family) to contextualize explainability in real-world public service challenges.
- Clear articulation of four “Maxims” for explainable AI: Be Transparent, Be Accountable, Consider Context, Reflect on Impacts.
- Walkthroughs of explanation types: Rationale, Responsibility, Data, Fairness, Safety, and Impact.
- Step-by-step workshop activities such as information gathering, evaluating explanations, and content discussions.
This is more than just a guide—it’s a hands-on workbook aimed at improving capabilities in ethical AI use across government bodies. The broader series includes eight workbooks covering domains like healthcare, education, policing, and urban planning, and themes such as accountability, sustainability, and fairness.
💡 Why it matters?
Explainability isn’t a buzzword here—it’s operational. In settings where AI can directly influence lives (like social care), understanding how decisions are made isn’t optional. This workbook translates high-level principles into concrete actions, making it easier for public servants to embed explainability into everyday AI practice. It’s not just about ticking compliance boxes—it’s about making sure systems can be trusted, audited, and challenged when needed.
What’s Missing?
There’s also little discussion of the trade-offs between transparency and other factors like privacy, security, or performance. Readers seeking sector-specific legal context (e.g. GDPR obligations around automated decisions) might find this absent. Some newer AI modalities (e.g. generative models or foundation models) aren’t yet clearly addressed, which could leave out use cases that are becoming increasingly common.
Best For:
Ideal for civil servants, AI project leads, and policy professionals working in or alongside government bodies. Also useful for academics or consultants designing AI ethics training for the public sector. Less relevant for private-sector tech teams unless their work intersects with public infrastructure or regulation.
Source Details:
AI Explainability in Practice (2024) is authored by a multi-disciplinary team led by David Leslie at The Alan Turing Institute, with contributions from researchers including Mhairi Aitken, Christopher Burr, and Janis Wong. The workbook series was produced with input from UK public sector partners and supported through EPSRC grants under the UKRI Strategic Priorities Fund. David Leslie, the lead author, is known for his work on AI ethics, explainability, and responsible innovation. The team draws from social science, ethics, AI policy, and public service experience, making the workbook both research-informed and practically grounded.
