How can we harness the power of Artificial Intelligence (LLMs) to benefit children and families?

Abstract

Large language models (LLMs – a type of artificial intelligence) are developing at extraordinary speed, and have increasingly become part of everyday life. This technology has the potential to change how social care professionals and the public access, understand and communicate information. However, LLMs have been adopted at such incredible speed across the country that there has been little time to ask, let alone answer, important questions about how these technologies should be designed, tested and used in social care.

This webinar argues that we should not simply ask “Can we use LLMs in social care?”, but “How can we make LLMs work for social care, responsibly, usefully, and in ways that reflect the complexity of the work?”.

Researchers from the Centre for Social Care and Artificial Intelligence (SCALE) at Cardiff University are addressing this challenge through three exciting projects:

1. Can LLMs support social workers’ use of evidence in practice? The ‘SWEET’ project is working with a Welsh local authority to develop a bespoke tool that answers practice queries with high quality research evidence.
2. Can LLMs translate research for minority languages? The ‘TINE: Translation is not enough’ project explores translating research text into Welsh while also making complex academic evidence clearer and more accessible.
3. Can LLMs preserve the voices that matter? This project examines how LLM conversation summaries represent different voices, with a particular focus on equity and retaining the voice of the child.

Using these studies as examples, Dr Verity Bennett, Dr Fernando Alva Manchego and Dr Aleks Edwards will highlight why social care needs to be treated as a distinct domain for LLM design and development, and how through responsible and rigorous research we might better harness their potential for the benefit of children and families.

Biographies

Dr Verity Bennett is a Senior Research Fellow at Cardiff University, working in the Centre for Social Care and Artificial Intelligence Learning (SCALE). Her research includes evaluating safeguarding interventions and exploring how artificial intelligence can improve access to information, support evidence-informed practice and strengthen multi-agency working to improve outcomes for children and families.

Dr Aleksandra Edwards is a Research Associate at Cardiff University specialising in responsible AI and natural language processing in social care, health and safeguarding. Through the Centre for Social Care and Artificial Intelligence Learning (SCALE), she works with practitioners and public contributors to explore how AI can support professional judgement while preserving important context and people’s perspectives. Her recent work has focused on evaluating large language models in social care and policy settings.

Dr Fernando Alva Manchego is a Senior Lecturer in the School of Computational and Mathematical Sciences at Cardiff University, where he is part of the Cardiff NLP group. His research focuses on text adaptation, simplification, and readability assessment, with the goal of building systems that make language more accessible to diverse audiences across languages. He works on applied projects in education, health, and social care, including TINE (plain-language adaptation of multilingual science) and SWEET (an evidence appraisal tool for social work). His work combines computational methods with people-centred evaluation and collaboration with practitioners and communities.