Will AI Strengthen or Weaken Deliberation? DemocracyNext Offers a Different Question

​DemocracyNext, the democratic innovation nonprofit led by founder Claudia Chwalisz, has released Deliberative Muscles & AI, a paper examining how deliberation technology can strengthen rather than erode citizens' capacities to deliberate. Written by Chwalisz, Sammy McKinney, Jorim Theuns, and Eugene Yi, the paper reframes deliberative assemblies not as recommendation-producing machines but as spaces where people build the cognitive, relational, and dispositional capacities democratic life depends on. Its central argument is that whether AI helps or harms deliberation is a matter of design, and it offers a typology of seven "deliberative muscles" alongside a practical framework for judging whether a given tool strengthens or substitutes for each one. For the dialogue and deliberation field, the paper is a rigorous guide for practitioners, funders, and technologists weighing how to bring AI into participatory work without hollowing out its human core.

---

---

As AI arrives in the world of citizens' assemblies and deliberative forums, DemocracyNext has published a paper that reframes the entire question practitioners should be asking. Deliberative Muscles & AI, authored by Claudia Chwalisz, Sammy McKinney, Jorim Theuns, and Eugene Yi, starts from a premise easy to lose sight of amid the enthusiasm for efficiency: deliberative assemblies are valuable not only for the recommendations they produce but for what they do to the people inside them. Over a thousand such processes have run worldwide, depolarizing opinion, building trust, and reconnecting people to political life. They are also slow, costly, and hard to scale, which is exactly why they have become a focus for AI integration. The paper's argument is that the efficiencies making AI attractive can quietly undermine what deliberation is for, and that whether AI strengthens or weakens the practice is not predetermined but designed.

​The paper's central metaphor is the deliberative muscle: a capacity maintained through practice, weakened when unused, and strengthened when trained. It identifies seven such muscles: self-reflection, reasoning, dialogue, vulnerability, collaboration, imagination, and facilitation, each naming a capacity participants and facilitators exercise in the room. For every muscle, the authors map AI along a spectrum drawn from David Krakauer's distinction between tools that strengthen human capacities and tools that substitute for them. The same technology can fall on either side depending on context. Real-time translation and live captioning can help participants engage directly with one another's contributions, while an AI that drafts a participant's statement and presents it as their own bypasses the work of dialogue entirely. Handing transcripts to a language model and returning its summary as "what you said" risks placing, in the authors' phrase, a layer of plastic wrap between people and their own outputs, whereas a facilitator-reviewed first-pass summary can give a group a shared starting point to revise together.

The paper's guiding question is one every convener and funder in the NCDD network can use directly: not whether AI produces a recommendation faster, but whether it leaves participants, facilitators, and the surrounding institutions more capable at the end of a process than at the start. That standard keeps human capacity, equity, and genuine encounter at the center, precisely the values that animate NCDD's mission. The authors extend the argument beyond individual tools to civic infrastructure, calling for investment in training, standards, and a "civic AI" future in which communities help shape the technology rather than simply consuming it. Their closing charge is distributed across roles: practitioners should resist efficiency gains that skip the hard work of capacity building, technologists should attend to how their tools shape users, and communities should demand processes that strengthen democratic muscle rather than atrophy it. Practitioners can read the executive summary and full paper at https://www.demnext.org/action/deliberative-muscles-and-ai.