Public value positions and design preferences toward AI-based chatbots in e-government. Evidence from a conjoint experiment with citizens and municipal front desk officers

Developing a chatbot to handle citizen requests in a municipal office requires multiple design choices. We use public value theory to test how value positions shape these design choices. In a conjoint experiment, we asked German citizens ( n = 1690) and front desk officers in municipalities ( n = 267) to evaluate hypothetical chatbot designs that differ in their fulfillment of goals derived from different value positions: (1) maintaining security, privacy, and accountability, (2) improving administrative performance, and (3) improving user-friendliness and empathy. Experimental results show that citizens prefer chatbots programmed by domestic firms, value chatbots taking routine decisions excluding discretion, and strongly prefer human intervention when conversations fail. While altering the salience of public sector values through priming does not affect citizens' design choices consistently, we find systematic differences between citizens and front desk officers. However, these differences are qualitative rather than fundamental. We conclude that citizens and front desk officers share public values that provide a sufficient basis for chatbot designs that overcome a potential legitimacy gap of AI in citizens-state service encounters. • Conceptualization of how public value positions translate into chatbot design choices. • Citizens view chatbots as intermediary tools rather than human replacements. • Citizens and front desk officers prefer chatbots from domestic IT developers. • Front desk officers emphasize efficiency gains for municipal staff. • Priming public values was ineffective in altering citizens' design choices.