In election times, many (young) voters consult a voting advice application (VAA) to inform themselves about the political issues at stake. Recently a new type of VAA has been introduced that includes a Conversational Agent in the form of a chatbot voters can address with questions about the political attitude statements in the tool. In this Dataverse we include data of two studies that have examined the effects of different types of avatars in these new Conversational Agent Voting Advice Applications (CAVAAs).

Study 1 reports about an experiment (N = 81) with a between-subjects design with three groups (a disembodied CAVAA; an embodied CAVAA with a renowned digital human avatar; an embodied CAVAA with a non-renowned digital human avatar). Results show that a renowned digital human does not exert effects on the usefulness, perceived political knowledge and voting intention of participants compared to a non-renowned digital human and a disembodied CAVAA. For ease of use, the disembodied CAVAA was rated more favorably than the embodied renowned CAVAA. No effect of eeriness was found.

Study 2 reports about an experiment (N = 199) with a between-subjects design with five groups: a disembodied CAVAA and four embodied CAVAAs that differed in the avatar’s familiarity (renowned vs. non-renowned) and expertise (political expert vs. not a political expert). No differences were found between the conditions for expertise on tool evaluation measures ease of use and usefulness, and the political measures political knowledge and voting intention. There was a significant effect on eeriness: the non-renowned avatars were seen as more eerie than the renowned avatars. This was in line with expectations based on prior research. The studies’ implications for theory and practice are discussed.

The survey data were gathered via Qualtrics and the data of the conversations have been collected through the chatbot developer Genius Voice.

Ethical clearance: Study 1 (REDC2024.11a) and study 2 (REDC2024.11) were approved by the ethics committee of Tilburg University. Method: Data were collected through two experimental studies. Universe: The population consisted of theoretically educated students from Tilburg University (study 1) and practically educated students from ROC Tilburg and SintLucas Eindhoven (study 2). Data sources: The survey data were gathered via Qualtrics and the data of the conversations have been collected through the chatbot developer Genius Voice.