The Politics of Using AI in Policy Implementation: Evidence from a Field Experiment

The use by government agencies of AI in guiding important policy decisions (e.g., on policing, welfare, education) has generated backlash and led to calls for greater public input in AI regulation. But what does the public’s input on this topic entail? Does personal experience with the technology or learning about its implications change people’s views on using AI for guiding policy implementation? We study these questions experimentally in an online labour marketplace. We track the attitudes of over 1,500 workers, where the boss who allocates them to tasks, the tasks’ content and valence are all randomly assigned. Over a three-wave panel, we find that personal experience with AI-as-boss affected workers’ behaviour on the job but not their policy attitudes. In contrast, exposure to information about the technology generated significant attitudinal change. The findings provide insight into the formation of attitudes on AI regulation and the potential politicisation of this issue.