The following project is shared with: Petra van den Bekerom, Tanachia Ashikali, Nadine Raaphorst, Andrei Poama, Johan Christensen, and Martin Sievert.
Project Title
Representative Bureaucracy in the Age of Artificial Intelligence (AI)
Behavioral change(s) addressed
Artificial Intelligence might impact public administrations and societies in many different ways. On the one hand, AI can support and even self-perform the analysis of large amounts of data and identify patterns that augment street-level bureaucrats’ decision-making by reducing biases, and thus enable a more accurate and granular representation beyond established categories (e.g. ethnicity, gender, class). AI can also automate routine tasks, and thus free up time for street-level bureaucrats to focus on work that requires human expertise and judgment. These proposed positive effects imply that AI might increase active representation of public organizations and, thus, benefit public service delivery broadly and disadvantaged citizens. But, on the other hand, there is a growing concern that AI-informed practices reproduce, entrench, or compound some social biases and prejudices. Such applications can misrepresent and harm the members of specific, and often disadvantaged socio-economic groups. Consequently, it is unclear whether AI enhances or reduces adequate representation of citizens by street-level bureaucrats. If AI tools become more relevant in decision-making and for communication in citizen-state interactions, addressing how this impacts citizens is crucial. Such changes could affect, both positively and negatively, outcome variables such as perceived fairness and trust.
Theoretical approach
The research question is both theoretical (how does AI affect the meaning and scope of representation?) and empirical (how does AI affect street-level bureaucrats and service recipients?). To answer it, we examine the relation between different forms and uses of AI tools and different forms of bureaucratic representation (i.e., passive, active, and symbolic representation). Furthermore, we examine the relationships between different forms and uses of AI tools and outcome variables related to bureaucratic representation (e.g., trust, responsiveness, blame). Thereby, we account for the interplay of representative bureaucracy and AI tools in affecting citizens’ perceptions, attitudes, and behaviors. We combine different streams of literature and theory such as representative bureaucracy, political theories of representation, and digital governance.
Empirical research strategies
Survey experiments will be used to test the effects of realistically specified AI-informed decisions on perceived representation by both citizens and civil servants (and cognate outcome variables). We additionally plan to carry out focus groups/interviews to examine how AI affects active representation by civil servants. We will apply these research designs across various services and policy areas (e.g., healthcare, education, taxes, criminal justice, and security) and identities (e.g., ethnicity, gender, socioeconomic class). Furthermore, we plan to utilize the findings from the survey experiments and focus groups to design a series of field experiments that further investigate and test our hypotheses.
Possibilities for inter- and transdisciplinary collaboration
The project combines several perspectives on how digitization affects public administrations and society as a whole, including risks and challenges. It will be great to seek collaboration with other projects that contemplate some of the concepts we employ (bureaucratic representation, social and political biases, disadvantageous citizens), as well as that take digitization as an important phenomenon to observe.
Social transition(s) addressed
The project revolves around explaining the impacts of the new Artificial Intelligence (AI) developments on representative bureaucracy. Representative bureaucracy, in its different forms, has shown to be important for responsiveness and equitable service delivery across disadvantaged groups in society. There are different types of representation: a) active representation, where street-level bureaucrats actively engage in representing the interests of disadvantaged societal groups; b) passive representation, referring to the representativeness of the workforce vis-à-vis society; and c) symbolic representation, which implies a positive effect of representation on the attitudes of disadvantaged groups. Recent developments, however, require the combination of representative bureaucracy with a focus on digitization and technology. For example, public service delivery is being transformed by a series of relatively new technologies and practices, which include automated processing of big data, machine-issued predictions, and screen-intermediated state-citizen interactions.