Can Sortition Save Democracy? ISPS Democratic Innovations Program Tackles Representation and Recruitment

If deliberative bodies comprised of randomly selected citizens can reinvigorate democracy, people need to show up. They need to participate. And they need to accurately represent the wider population.

Yale’s Institution for Social and Policy Studies (ISPS) assembled academic researchers and private sector practitioners leading such citizens’ assemblies to address challenges they face, including selection bias, sampling, and recruitment of citizens.

“Representation is pretty important in deliberative mini-publics,” said Alexandra “Ali” Cirone, an assistant professor in the London School of Economics’ School of Public Policy  with a joint appointment in the Government Department and an external fellow in the ISPS Democratic Innovations program. “You don’t want massively skewed samples of participants because that could undermine the legitimacy.”

Alan Gerber, ISPS director and Sterling Professor of Political Science, praised Cirone for her commitment to Democratic Innovations, which identifies and tests new ideas for improving the quality of democratic representation and governance.

“It’s always a fantastic opportunity to convene researchers and the people out there doing the work with the public in cities around the world,” Gerber said. “Ali has built a network of experts in this small but growing field of deliberative democracy, and we are thrilled we can help pool their resources and knowledge.”

In recent years, a growing number of governments — particularly across Europe and in Canada — have turned to random selection methods to form temporary citizens’ panels focused on specific policy issues. Some municipalities, like Paris, have gone a step further by creating ongoing, randomly selected groups tasked with shaping local priorities.

But citizens’ assemblies often suffer from selection bias, particularly concerning who opts into the pool from whom participants are chosen. Cirone said that representation matters for legitimacy, especially when assemblies inform policy. She emphasized the need for systematic data collection across organizations to understand who participates and why. And to improve recruitment strategies.

“We need to be collecting more data,” Cirone said, noting that recruitment methods vary widely and lack standardization. “We need to know: If you say yes, why did you say yes? If you say no, why did you say no?”

She drew data from a database maintained by the Organization for Economic Cooperation and Development, analyzing 65 assemblies and finding an average participation rate of about 6%, a figure consistent with participation in academic studies and the challenges of finding people willing and able to participate in in-person events that require time, travel, and accommodations.

Kyle Redman, a program manager for the Federation for Innovation in Democracy – Europe (FIDE), presented a toolkit to help organizations — particularly those involved in artificial intelligence development and governance — make better, more democratic decisions. He introduced a roadmap to help companies and governments determine how much decision-making power they are willing to share with the public or randomly selected citizens in an assembly and whether the democratic process they are using is good enough to handle that responsibility.

He argued that AI companies and governments are making decisions that affect everyone, but most people have no say in those decisions.

“We hope that our toolkit will provide researchers with open research questions, help developers and engineers understand what problems exist, and show a concrete path toward democratizing institutions,” Redman said.

Chris Ellis, director of MASS LBP in Canada, discussed the logistics behind the 55 assemblies his organization has led, including the delivery of 500,000 invitations. He said that recruiting people for citizens’ assemblies can be messy, expensive, and full of trade-offs, especially when trying to make the process fair and representative. He emphasized that real-world logistics — like the cost of mailing invitations (and even the effect of envelope color on response rates), dealing with dropouts, and managing expectations — are just as important as the theory behind sortition.

He compared civic lotteries to tools that need to be sold to public servants.

“This is new tech,” Ellis said. “We need to position it as such to get more people involved.”

Tom Lord, director of sortition services for the Sortition Foundation, discussed what the United Kingdom-based nonprofit has learned after running about 300 democratic lotteries around the world.

His team has created tools for random address selection, webforms for sign-ups, algorithms to ensure demographic balance, communication tools, and software testing to catch errors. They are working with practitioners in the United Kingdom to standardize data collection for better analysis and advocacy. They are confronting challenges such as declining response rates, possibility due to political disillusionment and post-COVID fatigue. And their testing has shown hybrid in-person and online events reduce participation.

In addition, Lord said the foundation is currently working to develop a multi-lingual open-source democratic lottery platform so that organizations without technical expertise can run their own recruitment and selection processes for citizens’ assembles.

“What if we can just make this something that people can use all over the world?” Lord said.

Ansel Herz, communications coordinator for FIDE North America, explored the history of deliberative traditions in the United States, from the Iroquois Confederacy to New England town meetings still in use.

He said more Americans might be enthused about citizens’ assemblies if they saw them as part of their own democratic heritage, connected to familiar traditions such as jury duty, and not a European import.

“If there is an American deliberative tradition, it’s a little bit less technocratic,” Herz said. “A little bit more grassroots.”

Carmel Baharav, a Ph.D. student in computer science at Massachusetts Institute of Technology, presented her work designing sortition algorithms that are both fair and resistant to manipulation while still being transparent to the public.

“When quotas are too tight — requiring exactly five participants with graduate degrees, for example — people may misreport their identity to increase their chances of being selected,” Baharav said, noting that tight quotas improve representation but reduce fairness. Loose quotas improve fairness but may reduce representational accuracy.

She introduced the “Goldilocks Algorithm,” a mathematical tool that adjusts selection probabilities so they are “not too low and not too high — just right.” 

“We can give you a menu of options,” Baharav said, explaining how the algorithm can help practitioners tune quotas based on their goals. “Loose quotas with higher equality, tight quotas with lower equality. In every quota setting, Goldilocks will still work on getting participant selection probabilities as close to equal as possible.”

Bailey Flanigan, assistant professor of political science and computer science at MIT, advises Baharav. At the ISPS conference, she presented a soon-to-be release free, public online toolkit designed to help researchers and practitioners run fair, transparent, and customizable sortition processes. In addition to Goldilocks, tools include a way to select panels with quotas, an algorithm for selecting alternates when participants drop out, and visualization tools to show how quotas affect fairness.

In addition, the toolkit, dubbed Lottery Lab, is designed to be open-source and explainable so users can see how quotas affect selection probabilities and explore tradeoffs, building trust with participants and clients.

“We want to make it easy for people to understand what’s happening behind the scenes,” she said. “And to make informed choices.”