ARTICLE

Better than Human? AI-Assisted Moderation Sparks Innovative Dialogue on Deliberative Democracy

The Friedrich Naumann Foundation’s Global Innovation Hub hosted the 7th episode of its Innovation for Democracy (I4D) Café webinar series— How Can AI Facilitate Public Participation and Deliberation? Featuring Dr. Alice Siu, Associate Director of the Deliberative Democracy Lab at Stanford University, and Mr. Yen-Ting Hsu, co-founder of Taiwan Reach-Out Association for Democracy (T-ROAD), the webinar explored how AI-assisted participation platform, like Stanford’s Online Deliberation Platform, could enhance a democratic dialogue and citizen’s public participation. Beyond showcasing its features, the event prompted critical discussions on the role of AI in public discourse and its implications for democracy.

What Is an AI Moderation Platform and Why Do We Need It?

As Merriam-Webster crowned “polarization” as its 2024 Word of the Year, the desire and urgency on facilitating people to rationally, inclusively, and consecutively discuss complex public affairs in their community and around the world has never been greater in every democratic countries. Since this task is obviously challenging, people start seeking for the help of emerging technologies; for example, as people become more and more familiar with using AI to summarize and analyze text and data, can it also be a helping force on how we participate in public deliberation? On the other hand, not surprisingly, skepticism surrounds the use of AI in public discussions due to concerns about transparency, fairness, and whether algorithms can genuinely represent human perspectives, etc.

This debate triggered our curiosity to invite two experts to find the answer together in this episode of I4D Café.

Dr. Siu, the lead collaborator of the Stanford’s Online Deliberation Platform project, introduced what this AI-assisted deliberation Platform can offer and has achieved. This platform is designed to ensure a fair, inclusive, and transparent discussion environment that can be scaled up and cope with the limitations of time and space. Developed based on the concept of Deliberative Polling®, the Platform uses AI to assist moderation, manage speaking turns, decrease hatred discourses, and encourage participation, making it to be one of the innovative approaches to tackle the challenging situation. The Platform has been successfully deployed the platform in over 35 countries, supporting over 40,000 hours of deliberations in multiple languages.

What motivated the Stanford team to develop this Platform? Dr. Siu explained that the limitations of human moderators—inconsistencies in quality and the challenges of scaling up discussions—inspired the creation of the platform. Even after many trainings, there will be always some moderators who not only didn’t fulfill expected performance of facilitation but also added too much personal viewpoints, which influences or even deter citizens from speaking out their true opinions. And when a discussion needs to be implemented with more participants, this problem will make the discussion more challenging. Designed to enhance democratic discussions by using AI to assist moderation and deliberations, the platform leverages AI to scale up discussions while maintaining consistent level of fairness and inclusivity.

The Online Deliberation Platform: Features and Functions

Dr. Siu emphasized that the goal, is not to replace human moderators but to complement them to ensure a consistent, equitable, and scalable facilitation in an easier way. To achieve that, the platform ensure those inevitable principles of an ideal discussion to be fulfilled in the platform by the following key features:

1. Speaking Queue: Participants are given a fair turn to speak, with a structured 45-second time limit. This ensures that quieter individuals have the opportunity to contribute without feeling pressured to interrupt.

2. Nudge Functionality: Use the help of AI to encourage quieter participants to share their views, fostering inclusivity and reducing dominance by more vocal individuals.

3. Toxic Speech Detection: Using a library database and API, the platform identifies and mitigates offensive language in real time.

4. Scalability: Facilitators can monitor multiple group discussions simultaneously, making it possible to conduct large-scale deliberations.

5. Data Insights: The platform visualizes participation trends, toxic speech occurrences, the times that each participant talks, and other metrics, offering valuable feedback for future discussions and a quick barometer for human moderator to know what needs to be done now to facilitate the quality of discussion

According to Dr. Siu, AI moderation holds immense potential for advancing democracy. Here are some key outcomes:

1. Promoting Public Participation: By offering non-intrusive features like speaking queues and nudges, AI can encourage participation and boost confidence, especially for introverts or those hesitant to voice their opinions.

2. Ensuring Fairness: Technology can help create a friendly, equitable environment where everyone has an opportunity to contribute, regardless of their background.

3. Reducing Polarization and Fostering Inclusivity: The platform also encourages empathy result in participants' increasing desire for deeper engagement and sustained participation in democratic processes.

A Practitioner’s Lens: Challenges and Feedback

While Dr. Siu showcased the platform’s capabilities,  Mr. Yen-Ting Hsu, a seasoned practitioner in deliberative democracy provided critical insights from his hands-on experience.

Both sides acknowledged the platform’s strengths, while Mr. Hsu indicated that moderation is not only technical skill but also sometime an art of expressing understanding and encouragement. He underscored the disparities in cultural capital that influence participants’ persuasiveness, which AI may overlook when focuses on equal rather than equitable participation. For example, even given an equal time to talk, people with higher cultural capital may be able to come up with persuasive output faster than others. Mr. Hsu also highlighted that diverse facilitation methods can also be the key accelerator of an ideal discussion, such as using sticky notes or other interactive tools during in-person discussions since different participants may be more used to encourage to talk by different method, and he would like to know whether AI platforms could incorporate such flexibility when ensuring consistency.

In response, Dr. Siu acknowledged these challenges, emphasizing the need for further research and development to integrate better large language models and other AI capabilities in order to sophisticatedly detect and tackle the gap of abilities of expression of everyone. She also emphasized the importance of balancing hands-off facilitation with structured intervention. Adopting a hands-off approach, the platform allows discussions to flow naturally while providing gentle guidance. This design aims to minimize the pressure often associated with heavy-handed moderation. She noted that while the platform’s current design prioritizes participant autonomy, diverse facilitation styles could be one of the features to explore in the future research.

Conclusion: What Should be Expected on AI Innovation for Democracy?

One of the most compelling aspects of this episode was its exploration of how AI can make empathy and mutual understanding possible in online discussions. Dr. Siu’s findings from the Deliberative Democracy Lab revealed a promising intersection between AI and democratic participation. The research suggested that in these structured digital spaces, participants report feeling more comfortable sharing their thoughts and learn from one another. The platform’s environment, supported by tools like speaking queues and real-time nudges, also helped reduce anxiety and empower diverse voices to contribute. Importantly, Dr. Siu shared that a space designed for openness also ignited a stronger desire among participants for continued engagement, which implicates growing empathy, mutual understanding and meaningful changes in opinion.

Both sides agreed that AI should complement rather than replace human efforts. As Dr. Siu pointed out, advancements in large language models can possibly enable AI to summarize discussions consistently and efficient. By automating repetitive tasks and providing real-time insights, AI allows human moderators to focus on more complex aspects of facilitation, such as understanding the context of discussions and encouraging meaningful contributions.

AI, when thoughtfully applied, holds the potential to nurture precisely the empathetic and inclusive dialogue that democracy urgently needs. Still, as discussed in this newest episode of I4D Café, the balance between equality and equity, autonomy and intervention will continue shape AI’s role in facilitating public participation and deliberative democracy. With cautious optimism, we hope to envision a future where AI not only enhances democratic dialogue but also strengthens the foundations of public participation.

Note:
Serves as a vital platform for exploring advance technologies with innovative angles and critical questions, “The Innovation for Democracy (I4D) Café” series is brought to you by Friedrich Naumann Foundation Global Innovation Hub. Watch the whole episode here if you feel like drinking more cups of coffee after reading this article!