Generative AI for Reducing Language and Administrative Barriers for Immigrants in Public Services

Abstract

Generative artificial intelligence (AI), including large language models, is increasingly explored to support communication and navigation in public services. This paper presents a case study of how immigrants in Norway use and perceive generative AI when interacting with digital public systems.

We adopted a mixed-methods design, combining twelve individual interviews with immigrants, two focus groups with six NAV employees, one digital professional meeting with about twenty NAV employees, four interviews with staff from the municipal Refugee Unit, and a survey of 55 immigrants in Trondheim.

The study highlights language and administrative literacy as key barriers, including challenges with complex procedures and secure authentication (e.g., BankID). Generative AI was valued for translating, simplifying, and explaining bureaucratic content, but its use was mainly concentrated among participants with moderate to high digital skills. Concerns about accuracy, trust, and privacy shaped adoption. We find that AI should complement rather than replace human support; equitable deployment requires inclusive design, verified official sources, and training for both users and staff. The study contributes to discussions on digital inclusion and the responsible use of AI in the public sector.