How Go Vocal is rethinking online public deliberation with AI

Having a good conversation online at scale is one of the hardest unsolved problems in technology. Social media didn’t solve it. It optimized for engagement, and the result was polarization, echo chambers, and public discourse that fragments instead of converges. Civic participation inherited much of that same logic. As a result, public participation today has a listening problem.

Over the past decade, digital tools have made it dramatically easier to collect input. More residents can contribute, more ideas surface, and more perspectives are captured than ever before. That is real progress, but it has also exposed a deeper challenge.

The bottleneck is no longer collection. It’s dialogue.

Getting thousands of voices into a consultation does not automatically produce better decisions. Without structure, contributions pile up with no real connection between ideas. The deeper goal of participation is to understand where a community agrees, disagrees, and wants to go next. But when residents see fragments instead of patterns, and governments see volume instead of substance, the purpose gets lost somewhere between submission and decision.

Most platforms make this worse by surfacing contributions the way social media surfaces content: by popularity. Ideas that attract early likes dominate visibility, while quieter voices, minority perspectives, and later contributions get buried. As legal scholar Cass Sunstein argues in #Republic: Divided Democracy in the Age of Social Media (2017), digital environments can narrow the range of viewpoints people encounter, reinforcing echo chambers and contributing to polarization.

This challenge is widely recognized. A recent report by the OECD on emerging technologies and civic participation highlights that, while digital tools expand who can participate, they also introduce new complexity in how input is understood, weighted, and translated into decisions.

So the real question isn’t “how do we get more input?” It’s whether civic technology can succeed where social media failed, and build the kind of online conversation communities actually need. It is about creating space for genuine dialogue, with a clearer understanding of different perspectives, a visible breadth of views, and the potential for common ground.

How AI is reshaping public participation and deliberation

AI is starting to play a role in solving this problem, but not in the way it is often framed. Until recently, most AI in participation worked behind the scenes, helping teams summarize results, tag responses, or generate reports after a consultation ends. Useful, but limited to operational efficiency.

The shift happening now is more fundamental. AI is entering the participation experience itself, shaping how residents navigate conversations, understand what others are saying, and decide how to contribute.

“The quality of a democracy is measured by how many people spoke, but also by how well the community listened to itself.” - Wietse Van Ransbeeck, CEO, Go Vocal

A number of teams are exploring this space. Tools like Pol.is, Talk to the City, Harmonica, Echo by Dembrane, and Stanford's Deliberative Platform experiment with ways to structure large-scale conversations, using clustering, summarization, and bridge-building algorithms to reveal patterns that would otherwise remain invisible.

What these approaches share is a shift in intent. AI is not used to maximize engagement. It is used to make conversations more understandable, more balanced, and more useful for collective decision-making.

Why AI matters for public participation in 2026

This shift comes at a moment where participation itself is maturing. As explored in the Go Vocal Trends Report 2026, participation is moving from one-off projects to a continuous, cross-team capability embedded across government.

But scaling participation alone does not solve the core problem. More input does not automatically lead to better understanding. Without structure, conversations remain fragmented and difficult to navigate.

What changes this is the ability to make large-scale input understandable while the conversation is still unfolding. This is where AI becomes useful: it helps people see how ideas connect, where perspectives differ, and where there is overlap.

The implications are clear:

• For governments, success shifts from counting responses to understanding the full landscape of perspectives, including those that do not get the most visibility.
• For residents, participation becomes a structured dialogue where contributions are visible, connected, and easier to engage with.
• For the civic engagement space, the bar rises. Legitimacy depends not just on reach and transparency, but on the quality of deliberation. Explainability and auditability of AI become essential: residents need to trust not just that they were heard, but that the conversation was fairly represented.

Our approach: building AI for real public deliberation

Perspectives is Go Vocal’s AI-powered deliberation feature, and it is built around a simple idea: if people cannot see the structure of a conversation, they cannot meaningfully take part in it.

Instead of presenting participation as a feed of disconnected contributions, Perspectives organizes input into a structured view of the conversation, with themes and sub-topics that reflect how people actually think about issues.

All of this happens live, as the conversation evolves, and remains visible to participants, not just to the teams managing the process.

‍

‍

What Perspectives is designed for

Revealing the structure of the conversation

‍Contributions are grouped into themes and sub-topics that emerge from the data. Residents do not see a flat list, but a conversation with shape, where different parts of the discussion are clearly visible.

• Example: In a consultation on housing, residents don't just see a category labeled "Housing". They see that within it, the community is actually having three distinct sub-conversations: zoning reform, social housing supply, and rent pressure. Each sub-cluster is a real pattern the AI detected in the data, not an editorial choice by staff.

Making sub-topics actionable

‍Sub-topics are framed as questions rather than labels. A question orients residents toward reflection, it signals that the process is moving toward a decision, and invites participants to think about what comes next, not just what's wrong. This small choice has a significant impact on the quality of what people contribute.

• Example: Instead of “Affordable housing,” residents see prompts like “How should the city prioritize new affordable housing locations?”

Surfacing representative voices, not just popular ones

Within each cluster, Perspectives doesn't rank contributions by recency or popularity. It uses a sampling algorithm that scores each contribution along three dimensions:

• Semantic representativeness and diversity: the selection covers the breadth of what's been said, not just the dominant view, ensuring minority perspectives remain visible.
• Newness weighting: recent submissions get a visibility boost to counteract first-mover advantage, the structural bias where the earliest contributions dominate a feed indefinitely.
• "Wise voices" uplift: contributors with demonstrated lived or domain expertise on the specific sub-topic get a visibility lift, so relevant expertise surfaces where it matters rather than getting buried.

• Example: A nurse commenting on healthcare, a housing officer on planning reform, or a small business owner on parking regulations may receive a visibility boost as “wise voices,” ensuring relevant expertise is surfaced where it matters.

Maintaining a stable structure as the conversation evolves

Most AI tools in civic participation do topical clustering post-hoc: after the consultation closes, as a summary for the team running the process, sometimes shared back with participants. Perspectives does it live, as contributions come in, for participants — and holds the two-level structure (themes + sub-topic clusters) stable even as the conversation grows.

Getting this right is harder than it looks. More simple approaches either reshuffle clusters every time a new contribution arrives (disorienting for residents mid-browse), or collapse into a flat layer of tags that hides the sub-conversations where the real substance sits. A stable, two-level structure that updates in real time without losing coherence is, as far as we know, unique in the space — and the same is true of the contribution sampling algorithm.

What Perspectives is (consciously) not optimizing for

Perspectives is deliberately not optimized for engagement rate, time on site, or likes as a proxy for the quality of the conversation. These are the metrics of social media platforms, and they produce the same outcomes: virality, polarization, and the amplification of dominant voices.

This is a choice worth naming, and one we are proud of. A platform that maximizes engagement will reward outrage over reflection, speed over depth, and visibility over representativeness. These incentives do not just fail to produce good deliberation, they actively work against it.

So Perspectives measures different things. Not how long a resident stayed, but whether they encountered the breadth of the conversation. Not how many reactions an idea received, but whether it reflects a meaningful pattern in community thinking. Not how much traffic a consultation generated, but whether it helped a community move closer to shared understanding and common ground.

This is, quietly, a political choice about what civic technology is for.

The difference becomes clearer when comparing how traditional participation platforms operate versus how Perspectives structures and surfaces input:

‍

‍

Perspectives does not replace human deliberation. It creates the conditions for it by making the collective conversation visible and understandable in real time.

Conclusion: The future of AI in public participation

The past decade of civic technology focused on solving the collection problem. The next phase is about solving the understanding problem.

When residents can see the full landscape of perspectives, not just the most visible ones, they are better equipped to reflect, respond, and contribute in ways that move a conversation forward. Participation becomes more thoughtful. Decisions become more grounded.

This shift also comes with responsibility. As AI shapes how participation is experienced, transparency about how conversations are structured and how perspectives are surfaced becomes essential. Trust depends on it. And that is exactly what we are building with Perspectives.

AI can help move public participation beyond volume and visibility, toward something more valuable: shared understanding between a government and its community.

Ready to transform how your community deliberates?

See how Perspectives helps residents navigate large-scale conversations, hear diverse perspectives, and contribute more meaningfully to public deliberation.

Book a demo with our team to explore how AI can improve not just the efficiency, but the quality of your community's conversations.

Book a demo