In this dataset the data regarding the research on text vs voice vs combined conversational agent voting advice applications can be found.

Conversational Agent Voting Advice Applications (CAVAAs) have been proven to be valuable information retrieval systems for citizens who aim to obtain a voting advice based on their answers to political attitude statements but desire additional on-demand information about the political issues first by using a chatbot functionality. Research on CAVAAs is relatively young and in previous studies only the effects of textual CAVAAs has been examined. In light of the positive effects of these tools found in earlier studies, we compared different modalities in which information can be requested to further optimize the design of these information retrieval systems. In an experimental study (N = 60), three CAVAA modalities (text, voice, or a combination of text and voice) were compared on tool evaluation measures (ease of use, usefulness, and enjoyment), political measures (perceived and factual political knowledge), and usage measures (the amount of information retrieved from the chatbot and miscommunication). Results show that the textual and combined CAVAA outperformed the voice CAVAA on several aspects: the voice CAVAA received lower ease of use and usefulness scores, respondents requested less additional information and they experienced more miscommunication when interacting with the chatbot. Furthermore, given the fact that the predefined buttons were predominantly used and stimulated users to request also more and different types of information, it can be concluded that CAVAAs should make information accessible in an easy way to to play into CAVAA users’ processing mode of low elaboration.

These data are gathered by Celine Aerts as part of her MA thesis project.