Despite the widespread usage and positive impact of Voting Advice Applications (VAAs), users of these tools often struggle to understand the political attitude statements they are asked to evaluate. Since users rarely resolve these comprehension issues by searching for information outside the tool, one promising solution is to enrich VAAs with additional contextual information, either through clickable explanations (VAA+s) or via AI-driven chatbots (CAVAAs).

Responding to calls for such enriched VAAs from both user-centric and normative perspectives, current study investigates how additional information can be integrated in a way that is both neutral and ethically sound. It does so by drawing on the perspectives of politicians, who are key stakeholders in the co-creation of VAAs in many countries, including the Netherlands. In 20 semi-structured interviews with local politicians from two large Dutch municipalities, we explored (1) the perceived acceptability and feasibility of four different types of contextual information—semantic clarifications, status quo descriptions, summaries of key arguments in the political debate, and party positions—and (2) how AI could be used to disseminate this contextual information to the user. Discussions addressed appropriate resources for information, language use, and ethical concerns such as the risk of political bias or misinformation.

Findings show broad support for including all four types of information, including argument summaries. Roughly half of the politicians emphasized that VAA developers should not evaluate the quality of arguments but could instead summarize the arguments most frequently mentioned by political parties to create concise and balanced overviews. Politicians also highlighted the importance of clear, accessible language, including the usage of concrete examples and pictures to make complex issues more understandable for a diverse audience. Opinions on AI use were more mixed. While some politicians saw its potential, most underscored that AI-generated content must be accurate, politically neutral, and transparently sourced. These insights underscore the need for a cautious, ethically grounded integration of AI into VAAs. By combining normative reflection with input from political stakeholders, the current study contributes to the responsible development of next-generation VAAs.

This dataset contains only materials that can be shared publicly. Sensitive interview data, such as full transcripts and the complete codebook, are not included. One anonymised sample interview is available, along with the interview protocol, informed consent, and additional documentation (all materials are in Dutch).

Ethics approval: The study was approved by the ethics committee of Tilburg University: REDC2024.75

Method: Data were collected through 20 semi-structured interviews

Universe: Participants were politicians from two large Dutch municipalities

Data sources: Interviews were recorded through MS Teams and a manual recorder.