During elections, millions of voters consult online Voting Advice Applications (VAAs) to gain insights into political party stances. However, voters often struggle to understand VAA statements. Previous studies show that incorporating a Conversational Agent (CA) that addresses voters’ inquiries about VAA statements enhances the users’ factual and perceived political knowledge, as well as their evaluation of the tool. The current paper reports on two experimental studies (N = 141; N = 164) delving into the prospect of enhancing these CAVAAs by investigating how the choice of an anthropomorphic communication style (conversational human voice vs. formal) influences tool evaluation and political measures among users with higher and lower levels of political sophistication. Results suggest that especially lower politically sophisticated users could benefit from CAVAAs since they felt more informed to vote after using a CAVAA, and because they answered more factual knowledge questions correctly after using the anthropomorphic CAVAA. Implications of these findings are discussed.

This Dataset includes information on "study 2" of the article “Hi chatbot, let’s talk about politics!” Examining the impact of verbal anthropomorphism in Conversational Agent Voting Advice Applications (CAVAAs) on higher and lower politically sophisticated users. There is also a first study in the article and the data related to this study have been archived here.