Coalesce: An Accessible Mixed-Initiative System for Designing Community-Centric Questionnaires

DOI

Effectively incorporating community input into civic decision-making processes is crucial for fostering inclusive governance. However, public officials often face challenges in formulating effective questions to gather meaningful insights due to constraints such as time, resources, and limited experience in questionnaire design. This paper explores the potential of leveraging large language models (LLMs) to address this challenge. We present \textit{Coalesce}, a novel mixed-initiative system that utilizes LLMs to assist civic leaders in crafting tailored and impactful questions for surveys, interviews, and conversation guides. Guided by best practices in questionnaire design, Coalesce improves question readability, enhances specificity, and reduces bias. To inform our design, we conducted a formative interview study with 30 civic leaders and implemented an iterative human-centered design process involving 14 feedback sessions. We built a fully-functional system before evaluating it through a real-world user study with 16 participants who applied the platform to their own community engagement projects. Our findings show that Coalesce improved participants' confidence in questionnaire design, supported diverse workflows, and fostered learning while raising important questions about human agency and over-reliance on AI. These insights highlight the potential for intelligent user interfaces to reshape how civic leaders engage with their communities, fostering more informed and inclusive decision-making processes.