In whose interest? The formation of AI policy in Sweden

Today, Sweden is one of the most extreme examples of a country where public activities are outsourced to private business, and where market rationalities has become the rationalities of the public sector. Neoliberal forms of governing have forcefully entered the public sector, and through different forms of new public management – mainly through marketisation and corporatization – the organization of the Swedish welfare state has been transformed (Öjehag and Granberg, 2019). At the same time, Sweden to a large extent still is perceived as a country characterized by a strong welfare state, and a country where there still is a strong trust in the state (Granberg et.al., 2021). Another Swedish characteristic is a long tradition of technology optimism, where technology is seen as an effective and neutral way for societal improvement. The history of social engineers as the “builders of the welfare state” is also a strong legacy that combines trust in technology and trust in the state (c.f. Hirdman 1992). Today, Sweden – like several other countries – aims to be world leading in AI, and AI is also put forward as the solution to many of the challenges that the welfare state currently faces. In this paper, we are interested in how this interplay between neoliberal governing and trust in both the state and in technology as a driving factor for societal change plays out in the strategies of how AI technologies should be implemented in the public sector. Thus, our empirical focus is to analyse how public ambitions and strategies are formed in relation to the ambitions and interests of industry. What purposes and interests are the implementation of AI in the public sector supposed to serve? Through an analysis of empirical material consisting of both of policy material from the Swedish Government and the Agency for Digital Government (DIGG), webpages and written material from Private Public Partnerships like AI Sweden, and interviews with key informants at DIGG and AI Sweden, we aim to scrutinise the formation of the public AI agenda in Sweden. Inspired by Carol Bacchi’s (2009) critical take on policy analysis, we will surface the problem representations produced in this policy, as well as the different effects these result in. We end the paper with a discussion on the democratic implications of the formation of AI policy in Sweden.