As lifelong students of democracy who argue for improving the participatory quality of our public institutions through the benefits of new technology, we are frequently asked why evolution is so slow and why our institutions prove so stubbornly resistant to change.

Part of the answer is political. Entrenched interests benefit from the status quo. But part of the answer is intellectual. For decades, influential strands of democratic theory have treated participation as a problem to be managed rather than an asset to be cultivated.

The "realist" view of democracy casts ordinary citizens as too busy, too uninformed, or too incapable to participate meaningfully in public life. The task of institutions, therefore, was not to harness collective intelligence but to contain popular involvement within narrow channels.

That assumption still shapes much of our thinking today.

The Wrong Question

A recent project called Habermolt (see the related study) is emblematic of this elitist position that treats the public as spectators rather than participants, consumers rather than co-creators, respondents to surveys rather than partners in solving problems.

Rather than asking how institutions can become better at listening, learning, and solving problems with the public, Habermolt seeks to circumvent public participation altogether. Habermolt’s goal is to learn how “current AI agents learn user preferences and represent them in an online, agent-only deliberation setting.”

The system works by creating an AI agent that speaks and votes on behalf of a participant. The agent writes statements, ranks proposals, and helps generate a consensus position.

But what problem are they trying to solve?

As a thought experiment, Habermolt is provocative. As a research agenda for democracy, it points us in precisely the wrong direction, away from approaches that strengthen democracy or produce effective governance.

Instead of asking how AI might help institutions incorporate distributed knowledge and experience into decision-making, these creators are asking whether AI can replace citizens altogether. 

Instead of devoting their energy to asking how AI might help institutions incorporate distributed knowledge and experience into decision-making, these creators are asking whether AI can replace citizens altogether. As governments struggle to solve pressing public problems, it is difficult to see why replacing citizens rather than empowering them should command scarce intellectual attention.

If all we want is agreement, prediction, or preference aggregation, then there is no need for democracy.

If all we want is agreement, prediction, or preference aggregation, then there is no need for democracy. We do not need AI agents to do what dictatorships will do. 

An object lesson in condescension

Habermolt is an object lesson in condescension. It continues a long tradition of research that starts from the premise that citizens are the problem to be managed rather than the source of knowledge, legitimacy, and collective capacity on which democracy depends.

What is troubling is that projects like this continue to attract attention while we devote comparatively little intellectual energy and investment to the far more urgent question of how to make institutions better at listening, learning, and solving problems with the public.

Democracy Is a Way of Governing Together

The purpose of deliberation is to create legitimacy, build civic capacity, surface lived experience, distribute power, and enable people to participate in shaping the decisions that affect their lives. 

The purpose of deliberation is not to produce a statistically efficient answer. It is to create legitimacy, build civic capacity, surface lived experience, distribute power, and enable people to participate in shaping the decisions that affect and improve their lives. Democracy is a way of governing together. The legitimacy of a decision comes both from the process by which people are involved and the outcomes of their collaboration.

At a time when trust in institutions is collapsing and democratic participation is declining, it is difficult to understand why we would devote scarce intellectual energy to creating synthetic substitutes

At a time when trust in institutions is collapsing and democratic participation is declining, it is difficult to understand why we would devote scarce intellectual energy to creating synthetic substitutes for citizens rather than finding new ways to empower them. 

Innovation, but for What?

To be sure, projects like Habermolt demonstrate that we can innovate. They show that democratic institutions are not fixed and that new technologies enable experimentation. That part is valuable.

 A more urgent question is not how to simulate participation, but how to improve institutional capacity and address problems that matter.

Given the scale of the challenges we face—from climate change and inequality to failing public institutions and declining public trust—the more urgent question is not how to simulate participation but how to improve institutional capacity and work on problems that matter.

We have spent decades documenting distrust, polarization, and democratic decline while investing far less effort in building the practices and institutions that might reverse them.

Trust is earned. If higher education is to retain the public confidence it still enjoys, we must do more than diagnose democratic dysfunction. We must help design its remedies.