隨著審議式民主概念的引進，包括參與式預算在內的不同公民參與機制陸續在我國被實踐。公民參與機制雖然多以公共審議作為制度設計的基礎，但是民眾依循我們所規劃的流程進行參與，是否就能在參與過程中實踐審議，過去卻很少在學術上或是實務上被檢視。換句話說，我們對於參與者在公共討論過程中所展現出來的行為與互動其實了解不多，對於制度與流程設計是否有效促成審議也未積極評估。基此，本研究參考DQI（Discourse Quality Index）、IDQ（Index of Deliberative Quality）以及Knobloch之架構，發展出具有14個評估面向的審議討論過程品質評估架構，並以2017年臺北市參與式預算住民大會為觀察場域，希望評估參與者在分組討論過程中的互動與行為是否實踐了審議。研究發現在臺北市參與式預算住民大會的分組討論，多數能夠營造出積極融洽的討論氣氛，且未出現不文明行為，而參與者在討論過程中，除了能夠正常參與而不被他人干擾之外，也能夠對等討論並彼此合作；不過，研究也發現分組討論的過程中，參與者較少基於共善論點，提出多元意見並相互回饋，且在不同論點與意見間進行相互連結、進行有品質與多重論點的舉證、進而透過理性說服，以尋求共識方案。

With the emergence of the concept of deliberative democracy, various kinds of citizen participation mechanisms, including participatory budgeting, have been adopted in Taiwan. Although the procedures of citizen participation events are often designed based on the principles of public deliberation, few studies have been conducted from theoretical or practical perspectives to examine whether citizens have deliberated during their participation when they follow the procedures developed for those events. In other words, we know little about participants’ behavior and interaction in public discussion, and we do not actively make attempts to evaluate whether the procedures designed can effectively lead citizens to deliberate. Based on the Discourse Quality Index (DQI), the Index of Deliberative Quality (IDQ), and the evaluative framework developed by Knobloch et al. in 2013, this study developed an evaluative framework for a deliberative dialogue process consisting of 14 evaluative dimensions. Participants’ behavior and interactions occurring in group discussions in neighborhood assemblies of Taipei City’s 2017 Participatory Budgeting were evaluated. The findings suggest that in most group discussions in neighborhood assemblies, a relaxed atmosphere can be sustained and incivility is absent. In addition, participants can also speak without interruption, treat each other as equal discussants, and cooperate in the discussions. However, on the other hand, this study also finds that in group discussions, participants are less likely to make statements based on the common good, provide a range of distinctive opinions and feedback, refer to others’ positions and opinions, make qualified and sophisticated justifications, persuade rationally, or make attempts to reach a consensus.