Transcribing trust
Evaluating the use of AI in social care
Project background
The UK government is increasingly pointing to the potential of AI to improve economic growth, productivity and public service delivery. Central and local government bodies are already rolling out AI systems in a range of contexts, from summarising policy consultations to detecting potholes.
AI adoption has proved popular in sectors with a high volume of routine administrative tasks that also require meaningful engagement with service users and making critical decisions. One such sector is social care. Councils in England and Wales are rapidly conducting pilots for the use of AI-powered transcription and summarisation tools by social workers in service delivery.
What is AI transcription?
AI transcription tools use audio recordings to generate text transcriptions that can be used in important documentation processes. This is sometimes described as a ‘speech-to-text’ capability. Some transcription tools use AI foundation models, but they can also use other technologies such as older algorithms or statistical models.
In social care, these tools can be used for transcribing meetings and creating meeting notes from interactions with people. Other sectors such as healthcare may also use and benefit from these tools – for example, practitioners might use them to create clinical notes.
Some tools also have additional AI features that can create summaries or recommendations based on these transcripts. Other AI transcription tools can generate text documents that are pre-formatted to fit into existing public-sector systems and procedures.
There is optimism that the use of AI for transcription and summarisation will free up social workers to spend more time with families and residents – and therefore improve care delivery. Prime Minister Keir Starmer recently claimed that AI tools could ‘almost halve the time social workers spend on paperwork’.
However, there is a pressing need for evidence on the real-world impacts of use of AI systems, and how public-sector bodies can use AI tools safely and in the public interest. Empirical evidence of how social workers use AI transcription tools, and the impacts of these AI systems on practice can support local authorities to make evidenced-based decisions when adopting and evaluating these tools.
Project overview
The Ada Lovelace Institute will examine how AI tools are being incorporated into social work in the UK and develop principles for evaluating these tools more effectively.
Working with social workers, technology leads and managers from a variety of local authorities already piloting AI transcription tools, we will hold interviews and an expert workshop to answer the following research questions:
- Why are local authorities adopting AI tools in social care and what are their metrics of success?
- How are social workers using AI transcription tools?
- Have social workers adapted their care work to accommodate to AI transcription tools?
- How are local authorities currently evaluating the use of AI tools in social work?
- What should the core principles of an evaluation framework for AI in social work be?
This project draws on existing Ada research:
- documenting evidence of a case study in data analytics in public services in a local authority
- exploring adoption of emerging technologies in public sector bodies
- examining guidance and support for procurement of AI in local government in the public interest.
If you’re interested in hearing more about this project, please contact Oliver Bruff and Lara Groves.
Image credit: sturti