Reinventing Online Engagement: AI and Deliberative Democracy

Author(s): Scott Douglas Jacobsen

Publication (Outlet/Website): The Good Men Project

Publication Date (yyyy/mm/dd): 2025/06/25

UC Riverside professor Kevin Esterling, creator of Prytaneum, talks about an AI-powered webinar platform built to foster inclusive, deliberative public discourse. Designed at the TeCD-Lab and inspired by ancient democratic principles, Prytaneum enables real-time audience participation and AI synthesis of diverse viewpoints—reshaping how civic engagement happens online. Esterling shares insights from working with Congress, the Department of Education, and public agencies, and reflects on the technical, financial, and conceptual challenges of developing such a platform. He envisions partnerships focused on high-quality communication and data collection—not just market-ready tech.

Scott Douglas Jacobsen: Today, we’re here with Kevin Esterling, professor of public policy and political science at UC Riverside, where he leads the Laboratory for Technology, Communication, and Democracy—TeCD-Lab. He developed Prytaneum, an AI-powered webinar platform inspired by ancient democratic practices that synthesize participant responses to ensure inclusive public discourse. Prytaneum enables meaningful dialogue in large online meetings by capturing and categorizing all voices, not just the loudest.

Supported by the National Science Foundation, Democracy Fund, and others, Esterling’s work advances what he calls “directly representative democracy,” fostering two-way communication between citizens and policymakers.

Thank you for joining me today. As mentioned in the introduction, what inspired the development of Prytaneum?

Kevin Esterling: About fifteen years ago, I received an NSF grant that funded a project where we worked with both the U.S. Congress and individual members of Congress to develop best practices for how elected officials could use webinar platforms to conduct more constructive public meetings with their constituents.

My background is in deliberative democracy. In my field, we focus on designing institutions that enable people to interact more deliberatively. That project used commercially available webinar technology, which was still new at the time. We developed a set of best-practice recommendations to make these public meetings more constructive.

It was a complete NSF-funded project. We published several papers and even a book titled Politics with the People, co-authored by Michael Neblo and David Lazer, my two collaborators. The project was successful.

We conducted a randomized controlled trial by having members of Congress implement our best practices to host a series of town halls. For each member of Congress hosting a town hall, we partnered with a survey firm to select a representative sample of constituents from their district. Some were randomly invited to attend the town hall, while others were not.

This allowed us to rigorously evaluate the effects of participating in such town halls—whether people were satisfied and whether there were democratic benefits. We found many positive outcomes: the meetings attracted a broad cross-section of the constituency, participants learned from one another, valued the experience, and believed the process benefited democracy.

Public officials liked the meetings. The project was a success, but when it ended, there was still room for improvement. We had been using commercially available webinar platforms, but I recognized a fundamental limitation—something built into the technology—that restricted our ability to make public meetings more deliberative.

The core issue is that existing webinar platforms were never designed with democratic engagement in mind. A typical webinar is ideal for a speaker to broadcast a message to an audience. A public official, for example, can inform constituents about their policies or positions. However, as anyone who has ever attended a webinar knows, it is a terrible format for audience members to meaningfully speak back to the speaker.

As a political scientist working in democratic theory, I saw a mismatch between our technology and the kind of democratic engagement we were trying to achieve.

That realization led me to seek new solutions. I received a small seed grant from the Democracy Fund, a funding agency based in Washington, DC, to build a prototype of what is now Prytaneum. The idea was to rethink the webinar from the ground up—to redesign it in a way that gives constituents a stronger voice in policymaking, in real-time, during a meeting. At the same time, the technology had to encourage constructive engagement with the content being discussed.

So, Prytaneum is designed to elevate constituent voices and guide them into meaningful, organized dialogue.

Since then, I have received additional grants, and we’ve developed the platform further. I initially approached it as a political scientist—a project to improve public meetings. However, once the platform was built, I realized that the underlying communication problem we were solving wasn’t limited to policymaking or government.

Any speaker, in any setting—whether online or in person—wants to understand what their audience is thinking. But no technology has enabled mutual learning or real-time insight into audience sentiment. Prytaneum solves that broader, more universal communication challenge.

Jacobsen: Have there been real-world applications—such as with the California Mental Health Services Oversight and Accountability Commission?

Esterling: Yes. I received a grant to help fund the data science component of Prytaneum, specifically for integrating large language model artificial intelligence to support facilitators in curating participant input. We piloted this with the California Mental Health Services Oversight and Accountability Commission. They regularly host public meetings and were looking for ways to improve the quality of public engagement. Our collaboration helped them do just that.

The California Mental Health Services Oversight and Accountability Commission team advised us as we explored how to integrate AI into the curation process. We then hosted several demonstration town halls last summer to test the system.

I also have a separate grant from the Department of Education, through which we used Prytaneum without the AI component. In that case, we focused solely on the engagement features—the front end—as part of a curriculum module developed through the grant. The program enabled high school seniors to meet with their members of Congress as a capstone event that culminated in a three-week curriculum.

The students went through the curriculum, learning about the issue they would be discussing, the role of representatives in a democracy, and key elements of analytical reasoning. The final event was a virtual town hall on Prytaneum, where they debated the issue with their member of Congress. It was a multi-year project, and thousands of students participated using the platform.

Jacobsen: How does the AI distinguish between popular opinions and minority perspectives during these discussions?

Esterling: That’s a great question. One of the core problems with public meetings is that the people who show up are almost always unrepresentative of the broader community that the speaker wants to reach. You especially see this in local government settings—city council or planning department meetings—where the loudest or most affluent voices often dominate.

We designed Prytaneum to ensure a broader range of community perspectives are heard. One of the primary tools we use is our polling feature. When a speaker launches a poll, participants not only select their preferred option but also write a reason explaining why they support that choice.

So, we are not just collecting answers—we are capturing the reasoning behind those answers. The AI then synthesizes and summarizes these written justifications for each available option. Even if only a small minority selects a specific response, their reasoning is surfaced and shared with the entire group.

This helps ensure that all perspectives—not just the most common ones—are represented. It also allows participants to encounter reasoning that may differ from their own, helping them step outside their echo chambers.

Jacobsen: And things are often said differently depending on the language and cultural context. The texture of language can emphasize different nuances. Is Prytaneum planning to incorporate multilingual or accessibility features to support that?

Esterling: Yes—thank you for asking. Currently, Prytaneum offers real-time translation between English and Spanish. Participants can choose to participate in either language.

Yes, everything is translated in real-time. And thanks to large language models, machine translation today is much more accurate than it was a few years ago. It works pretty well.

If we can secure more funding, our long-term hope is to make Prytaneum fully language-neutral. Ideally, users would select their preferred language from a dropdown menu, and everything would be presented in that language. I believe that’s absolutely essential to making the platform globally inclusive.

Jacobsen: What were some of the challenges you faced in building Prytaneum?

Esterling: There were several—three major ones.

The first is funding. My lab operates on a very tight budget—a classic underfunded academic lab scenario. If I had a full team of engineers, there would be so much more we could build, but development moves slowly because of limited resources.

The second challenge was technical. Building a platform where many people can simultaneously submit content and then have that content distributed in real time to everyone else’s computer is complex. We had to build a backend that supports real-time updates across many users without lag or data conflict.

Prytaneum is also fully collaborative, meaning multiple users interact with the same content simultaneously. This introduces another layer of complexity, so we had to design the system to prevent conflicts when people engage simultaneously with the interface.

The third and perhaps the most significant challenge is conceptual. We hold a weekly design team meeting to explore how to translate our normative aspirations—inclusivity, deliberative equality, and constructive engagement—into actual design features.

What we’re really trying to do is solve a problem that no one has attempted before: how to make it possible for everyone in a meeting to understand each other at the same time. That’s a conceptual puzzle that has taken a lot of thought and iteration. Much of our work has gone into understanding the problem before translating that into technical design.

Jacobsen: Labs are perpetually underfunded. Even businesses face scale-related limitations. What kind of business partnership would help you scale Prytaneum as a research platform and as a broader product?

Esterling: If you look at the webinar market today—and I will not name names—you’ll notice something striking: there’s almost no differentiation between the major platforms used for large-scale meetings. They all function in essentially the same way.

They all essentially do the same thing. So, Prytaneum offers a chance to fundamentally rethink the webinar industry. It has the potential to shift how the entire sector conceptualizes the purpose and design of communication platforms.

For me, the ideal partner would be a corporate collaborator interested in reimagining communication—not someone chasing short-term profit, but someone who cares about making communication more effective. Ideally, this would be a company that values high-quality data collection—a company that sees the benefit in creating environments where people interact meaningfully, leading to deeper insights about how they think and make decisions.

Such a company could use those insights to improve its own products or services. So, rather than seeking a quick return on investment through a flashy app, this would be a partner committed to data collection the right way—using a platform that encourages authentic, constructive engagement.

Here’s another way to think about it: On most webinar platforms today, audience members—everyone who is not a panellist—are expected to be passive observers. There are very few opportunities for active participation. And when people are passive, they are not cognitively engaged; they are simply watching, not processing or reasoning deeply about the content.

Now, imagine using AI to mine audience input in real time on a traditional webinar platform. The data you collect will be superficial if the audience is not mentally engaged. It becomes a classic “garbage in, garbage out” problem. The AI might process those inputs, but the outputs will lack meaningful insight.

By contrast, Prytaneum’s design actively fosters deeper engagement. The audience is asked to think critically, contribute reasoning, and reflect on different perspectives. As a result, the platform can generate much richer, higher-quality data. That’s the core of our pitch to potential partners: this is not just a better communication tool—it is a tool for collecting better data because it encourages better thinking.

Jacobsen: What else can I ask? How can organizations and communities access Prytaneum?

Esterling: Well, right now, Prytaneum is a research platform. We use it primarily for continued development—especially to improve the large language model integrations.

And to clarify, everyone other than me calls it Pry-TAY-neum. My collaborators say it that way. Everyone I speak with says it that way. But for some reason, I say PRY-taneum. I named it, but apparently, I have no control over how it is pronounced! That’s okay—I’ve accepted it.

We are currently reviewing grant proposals to take the next step: using Prytaneum as a platform to collect better data and further improve the language models themselves. It also functions as a testbed—a proof-of-concept platform for organizations that want to experiment with more inclusive and intelligent online engagement.

I say “proof of concept” intentionally. My lab has one software developer. We’re not a tech company. So, we cannot promise that Prytaneum is ready for market deployment with full customer support and all the bells and whistles.

Organizations using it would need to accept it as-is. That said, they can visit the website—I believe you have the URL for it—where there are video demos showing how the platform works in practice.

If an organization is interested in using Prytaneum, they can contact me directly. If the interest aligns with our goals, we could issue a license that would allow them to create and host meetings. A permit is required to initiate new meetings, so collaboration is key.

Jacobsen: Kevin, thank you very much for your time today. It was great to meet you, and I appreciate you walking me through this cutting-edge development in the evolution of webinar platforms.

Esterling: Awesome. Yes, thank you so much, Scott.

Last updated May 3, 2025. These terms govern all In Sight Publishing content—past, present, and future—and supersede any prior notices. In Sight Publishing by Scott Douglas Jacobsen is licensed under a Creative Commons BY‑NC‑ND 4.0; © In Sight Publishing by Scott Douglas Jacobsen 2012–Present. All trademarks, performances, databases & branding are owned by their rights holders; no use without permission. Unauthorized copying, modification, framing or public communication is prohibited. External links are not endorsed. Cookies & tracking require consent, and data processing complies with PIPEDA & GDPR; no data from children < 13 (COPPA). Content meets WCAG 2.1 AA under the Accessible Canada Act & is preserved in open archival formats with backups. Excerpts & links require full credit & hyperlink; limited quoting under fair-dealing & fair-use. All content is informational; no liability for errors or omissions: Feedback welcome, and verified errors corrected promptly. For permissions or DMCA notices, email: scott.jacobsen2025@gmail.com. Site use is governed by BC laws; content is “as‑is,” liability limited, users indemnify us; moral, performers’ & database sui generis rights reserved.