Facts

Contact person:
Kettil Nordesjö
Financer:
  • Forte
Responsible at MaU:
Kettil Nordesjö
Project members at MaU:
External project members:
  • Rickard Ulmestig Institutionen för socialt arbete Linnéuniversitetet
Time frame:
01 July 2026 - 30 June 2030
Faculty/department:
Research environment :
Research subject:

About the project

Formal and informal use of large language models in social assistance: shaping social workers’ professional discretion and co-producing actionable implementation strategies

Despite high expectations that AI and generative language models (LLMs) such as ChatGPT will reduce administration and support decision-making in social work, there is a lack of knowledge about how LLMs affect social workers' discretion. Their use is often informal and lacks oversight, which can weaken transparency and legal security in the administration of social assistance, with implications for people in marginalised positions. The aim is to explore how the formal and informal use of LLMs shapes social workers’ discretion within social assistance, and to co-produce actionable implementation strategies together with municipalities. The project takes an ethnographic approach. The first part of the project focuses on informal use among social workers, while the second part concerns formal use managed by managers and IT professionals. Annual workshops are held with municipalities, during which implementation strategies are co-produced. The study is based on interviews, observations, document analysis, and workshops conducted over four years.