Conversational agent voting advice applications: A comparison between a structured, semi-structured, and non-structured chatbot design for communicating with voters about political issues

Voting Advice Applications (VAAs) are online survey tools that inform voters about political parties and their standpoints. Recent research shows that these tools may be improved by integrating a chatbot function that can provide on demand background information about the political issues. This new goal-oriented tool for information-seeking tasks is called a Conversational Agent Voting Advice Application (CAVAA). In an experimental study (N = 185) during the Dutch national elections of 17 March 2021, three CAVAA designs (structured with buttons, non-structured with a text field, and semi-structured with buttons and a text field) have been compared on tool evaluation measures (ease of use, usefulness, and playfulness), and political measures (perceived political knowledge and voting intention). Also, a comparison has been made between users with a high or a low educational level. Results show that the structured design was perceived as more playful compared to the non-structured design. Moreover, lower educated users rated the structured and the semi-structured design as easier to use than the non-structured design. Finally, this lower educated group also reported higher levels of political knowledge after using a CAVAA than the higher educated users. This suggests that CAVAAs can be a valuable tool to inform voters.