When Bailey Flanigan began developing algorithms for selecting participants in citizens' assemblies, she was addressing a problem that cuts to the core of democratic legitimacy: how do you assemble a representative group of people when the only citizens willing to participate do not look anything like the broader population? The question, at once technical and political, now has a working answer deployed on a live platform. Flanigan, who joined MIT in fall 2025 as a shared faculty member across the Schwarzman College of Computing, the Department of Political Science, and the Department of Electrical Engineering and Computer Science, has built computational tools that balance representation, fairness, transparency, and resistance to manipulation in citizen assembly selection. Her work is already available on panelot.org, an open-access platform that walks practitioners through the trade-offs involved in assembling decision-making bodies that the public can trust.
The Algorithmic Problem at the Heart of Citizen Assemblies
Citizens' assemblies are increasingly used by governments and institutions to deliberate on complex policy questions, from climate action to artificial intelligence regulation. But the standard recruitment model suffers from a predictable skew: willing participants tend to be younger, more educated, and more engaged with the specific issue at hand. Flanigan gives the example of an assembly on AI policy, where self-selected participants might disproportionately represent tech-savvy demographics, leaving older citizens, non-technical workers, and other affected groups underrepresented despite their direct stake in the outcomes.
The algorithms Flanigan developed treat this as a constrained optimization problem. The goal is to select a panel that reflects the demographic and attitudinal diversity of the wider population, while also maintaining equal individual chances of participation, resisting strategic manipulation of the selection process, and keeping the entire procedure transparent enough for the public to accept its legitimacy. These objectives often conflict. Maximizing demographic balance under strict equality of chance, for example, can create openings for manipulation if the algorithm's decision logic is not carefully designed. Flanigan's contribution is a set of mathematical tools that make these trade-offs explicit and then optimize according to the priorities the practitioner specifies.
From Farmland to Panelot: A Cross-Disciplinary Path
Flanigan's route to this work was anything but direct. Growing up on a family farm in Wisconsin, she pursued a self-directed education that ranged across medicine, bioengineering, public health, economics, and eventually computer science and political science. At the University of Wisconsin at Madison, she worked on cancer therapeutics in a wet lab and on tumor genetics computationally, before moving into public health research on microfluidic devices for HIV detection in low-resource settings. A growing concern that the science she was doing would only reach a small, wealthy fraction of the world pushed her further afield, toward economics and then toward formal mathematics.
Encouraged by mentors at Wisconsin, Princeton, and Carnegie Mellon, Flanigan entered a computer science PhD program at Carnegie Mellon, where she began research on social choice and democratic decision-making. Her academic trajectory, she says, was simply the result of chasing the problems that felt most pressing at the time, even when that meant operating in areas where she was less conventionally qualified. That pattern, she notes, cultivated the ability to learn the languages of new disciplines quickly, a skill she considers essential to her current joint appointment across computing and political science.
How Panelot.org Translates Technical Trade-offs into Practical Decisions
The algorithms Flanigan developed are now operational on panelot.org, an open-access website that hosts the selection tools and guides users through the decision process. Rather than presenting a black-box solution, the platform surfaces the inherent tensions in assembly design. A practitioner might prioritize demographic representativeness above all else, or might decide that equality of individual participation chance is non-negotiable, or might require full auditability of every selection step. Panelot makes the consequences of each priority visible and then finds the optimal selection given the stated constraints.
This approach matters because the perceived legitimacy of a citizens' assembly depends not only on the outcome but on the process used to reach it. If the selection mechanism is opaque or appears manipulable, the assembly's decisions carry less weight, regardless of the quality of deliberation. Flanigan's tools are designed to produce panels that can withstand public scrutiny on procedural grounds, which is a prerequisite for any real political impact.
What This Means for Democratic Participation
Flanigan's research addresses a structural weakness in contemporary governance. As algorithmic decision-making becomes more prevalent in public life, the tools used to constitute decision-making bodies must themselves be technically sound and democratically defensible. Her current work extends beyond assembly selection to broader questions about how to systematically gather public input on complex policy decisions and how the framing of questions in preference elicitation affects the conclusions drawn from that input.
The joint appointment at MIT gives Flanigan the freedom to explore both the political and technical dimensions of these problems. Her research sits at the intersection of computer science and political science, treating democratic design as an engineering challenge with real human consequences. The algorithms on panelot.org represent a concrete deployment of that thinking, available now for any institution or government body seeking to build a more representative deliberative process.
Who Should Pay Attention to This Work
For researchers and practitioners in computational social choice, Flanigan's algorithms offer a rigorous framework for handling the selection problem that has long plagued citizens' assemblies. For government bodies and civic organizations considering participatory decision-making, the platform provides a usable tool that makes technical trade-offs legible without requiring a computer science background. And for technologists interested in the intersection of algorithms and democracy, this work demonstrates that the hardest problems in democratic design are not purely political but involve deep computational and mathematical challenges.
The practical takeaway is straightforward: if your institution is planning a citizens' assembly or any participatory process that requires representative selection, the tools to do that selection in a transparent, fair, and technically defensible way now exist and are freely available on panelot.org. The algorithms have been developed, tested, and published. The question is whether democratic institutions will adopt them.
