"If you are a robot, may I speak to an adult?" - An exploratory case study of NAVs chatbot Frida

Abstract

In recent years, digitalization of public services has been a priority for the Norwegian government. This initiative has been pursued with the objective of making digital interaction the preferred first choice for the Norwegian citizens in need of governmental services. One of these initiatives is the chatbot implemented by the Norwegian Labour and Welfare Administration (NAV) which this study revolves around. In this study, the focus is on understanding user interaction with a chat bot in the public sector by exploring two specific aspects: (1) how the chat bot interprets the queries posed to it, and (2) what the limitations of the chatbot are, accompanied by suggestions for improving user satisfaction. This is done by applying two different methods of analysis’, one exploring the internal chat bot system, and one focusing mainly on the dialogue. The analysis resulted in four main findings: (1) The complexity of the questions posed to the chatbot was often a determining factor on how accurate the chatbot was in its responses, (2) the citizens often gave thumbs up on their interactions when they were able to get relevant help in the form of escalation to a human advisor, (3) The citizens frequently inquired about their specific cases, but the chatbot provided response to more general questions (4) The chatbot showed a lack of contextual understanding, failing to pick up on and address the specific contexts in the conversations. Lastly, the discussion explores how the two analytical methods compare against each other and offers suggestions for potential improvements to the current solution.