Biases in Administrative Service Encounters: Transitioning from Human to Artificial Intelligence

Introduction

Artificial intelligence (AI) holds the potential to assist or even replace civil servants in citizen-state interactions, which refer to those instances in which citizens seek or receive assistance, information, or support from a public agency. In Germany alone, around 130 to 150 million such public service encounters (PSE) occur each year, with the current proportion of digital encounters, facilitated through email or chatbots, standing at 41 percent. This is only the beginning of a revolution that will accelerate due to technological innovation and scarcity in human resources. The goal of this project is to study the consequences of shifting from human to AI-powered conversational agents in PSE.

Background

Paradoxically, while the transition to artificial encounters is already taking place in practice, we still lack understanding of what happens during conventional human encounters. This prevents researchers from evaluating the consequences of shifting from one venue to the other.

Research Questions

Therefore, the project’s main research questions are:

1. How can we conceptualize and measure communicative biases in PSE?
2. What biases emerge in the verbal communication of humans and AI agents in PSE?
3. How does variation in communication affect citizen perceptions of the encounter?

Potential Impact

AI has the potential to improve service delivery, but there are also risks for the legitimacy of democracy. Without further research, the introduction of AI threatens to merely shift biases from one venue (civil servants) to another (AI agent).

Project Contributions

As its ground-breaking contributions, the project:

1. Develops natural language processing tools to operationalize verbal communication.
2. Collects data from human encounters in PSE and realizes experiments on AI-based encounters.
3. Analyzes such data to formulate and test a theory about communicative biases in PSE.
4. Produces and disseminates urgently needed practical insights on the application of AI conversational agents in the public sector.