Supporting Online Discussions: Integrating AI Into the adhocracy+ Participation Platform To Enhance Deliberation

TL;DR

This work tackles persistent challenges in online deliberation, notably information overload, polarization, and incivility, by extending the adhocracy+ platform with two AI-driven modules. The Comment Recommendation Module uses stance detection to present opponents' viewpoints, boosting reciprocity and participation, while the Deliberative Quality Module uses the AQuA scoring system to highlight high-deliberative comments. A large-scale field study (N=1,356) reveals that the CR module increases engagement and perceived diversity in discussions, whereas the DQ module yields more modest effects, with some evidence of improved quality in certain topics. The extensions are implemented in an open-source framework, Discuss2.0, and point toward future work to refine AI predictions, integrate modules, and broaden language support for more inclusive civic participation.

Abstract

Online spaces provide individuals with the opportunity to engage in discussions on important topics and make collective decisions, regardless of their geographic location or time zone. However, without adequate support and careful design, such discussions often suffer from a lack of structure and civility in the exchange of opinions. Artificial intelligence (AI) offers a promising avenue for helping both participants and organizers in managing large-scale online participation processes. This paper introduces an extension of adhocracy+, a large-scale open-source participation platform. Our extension features two AI-supported debate modules designed to improve discussion quality and foster participant interaction. In a large-scale user study we examined the effects and usability of both modules. We report our findings in this paper. The extended platform is available at https://github.com/mabehrendt/discuss2.0.

Figures (2)

• Figure 1: We propose two AI tools that we integrate into adhocracy+. (Left) Comment Recommendation Module: Participants are confronted with a comment that contradicts their own opinion and are asked if they want to respond. The AI tool determines the stance of the comments, which is used to propose opposing comments. Translation: The following comment has already been added to the discussion. Do you want to reply to it? (Right) Deliberative Quality Module: We predict a deliberative quality score (AQuA score) for each comment. Comments with a high AQuA score are sorted to the top of the discussion and highlighted in bright green and marked as "top comment".
• Figure 2: Overview of the architecture to extend adhocracy+ with our AI tools. (Left) The debate module imports both the stance detection and deliberative quality AI's as Python modules. (Right) The Django database model sends out an event when a new comment is added to the database. The event is handled in signals.py where the new comment is passed either to the stance detection or deliberative quality model. These send a response (either a stance or quality score) back to the database where the corresponding response is stored.