Conceptual challenges of researching Artificial Intelligence in public administrations

Research has been advancing on the development and deployment of Artificial Intelligence (AI) in public administrations. However, there is limited consensus and agreement on what is considered Artificial Intelligence, as different understandings and approaches in research and practice exist. This paper explores and compares the varying ways AI has been described and understood in previous Information Systems and eGovernment research. Following, a survey amongst Belgium civil servants is analysed to assess what they associate with the term Artificial Intelligence. The findings show that many civil servants tend to associate AI with being able to conduct intelligent tasks, have certain capabilities or are specific applications they are familiar with. Specific algorithms or learning methods, often included in research papers, are not associated with the term AI. These results show that researchers and policymakers may have opposite or even paradoxical views on what is or is not AI, which could have significant consequences for researching the adoption of AI in government, as well as comparing different research findings. In this respect, the paper proposes to use an integrative lens to study AI in government, by including different dimensions and understandings.