Newswise — Last winter, 80 residents of Washington State convened virtually to discuss the best ways for their state to tackle climate change. Their final recommendations were shared with state legislators, who are now considering some of the ideas in their policymaking. But the participants of the Washington Climate Assembly were neither climate experts nor politicians. Instead, they were randomly selected citizens from all walks of life, chosen carefully to reflect a range of demographics and views on climate change.
Such citizens’ assemblies are an increasingly popular way, around the world, of engaging average people in their democracies. But ensuring that participants are truly representative of society at large is a daunting analytical challenge.
That’s where Bailey Flanigan, a Hertz Fellow and a graduate student at Carnegie Mellon University, comes in. Flanigan and colleagues at Carnegie Mellon and Harvard University have developed a new algorithm for selecting the participants in citizens’ assemblies, a process called sortition. The goal of their approach, she says, is to improve the fairness of sortition—and it’s already been published in Nature and used to select participants for dozens of assemblies, including the Washington Climate Assembly.
“Our algorithms offer provable guarantees of fairness, permit new kinds of transparency, and introduce mathematical structure that we hope will let people ask new questions about how we can use citizens’ assemblies in our democracy,” said Flanigan.
An Engineering Toolkit
As an undergraduate, Flanigan studied biomedical engineering at the University of Wisconsin-Madison. She had always been interested in solving important technical problems, but during her engineering coursework, Flanigan realized her interests were broader.
“I started to develop a better understanding of sociology and politics and realized that a lot of problems require not just engineering solutions, but a more holistic approach,” she said. “To really have an impact, you also need to rely on insights from the humanities, as well as people working on the ground.”
So as she began her graduate work, Flanigan transitioned from biomedical engineering to theoretical computer science, where she could use modeling approaches to study questions of fairness, transparency, and efficiency in politics. The Hertz Fellowship allowed her the freedom to pursue this new line of interest. Moreover, her graduate advisor, Ariel Procaccia, had just become interested in citizens’ assemblies, an area of research right up Flanigan’s alley.
A Modern Town Hall Meeting
In ancient Greece, men from all social classes were invited to participate in large political debates to shape the future of their cities. It was the beginning of democracy and the earliest citizens’ assemblies. Today, citizens’ assemblies are being used in dozens of countries at municipal, regional, and national levels, and some of the decisions reached by these panels are having large implications. In Ireland, for instance, two citizens’ assemblies commissioned by the legislature led to the legalization of same-sex marriage and abortion.
Compared to ancient Greece, however, it’s harder to engage many people in direct policy debates today. Groups who try to recruit volunteers for citizen assemblies in the United States report less than a 5 percent success rate; if they send letters to 10,000 randomly selected citizens, a few hundred will reply, and volunteers are likely to be wealthier and more educated than average.
“This means you have a pool of volunteers that is quite skewed,” said Flanigan. “If you simply randomly draw the panel from this pool, there’s a pretty good chance the panel won’t look like the population and their conclusions won’t necessarily reflect the interests of the population as a whole.”
To help solve this problem, organizations commissioning citizens’ assemblies set quotas reflecting the community they’re drawing from—perhaps they dictate that half the people selected to be in an assembly must be female, and a quarter must be from the lowest quartile income bracket, for instance.
Better Panel Selection
Selecting participants for a citizens’ assembly, while meeting a long set of quotas, is too complex for pen and paper. In the past, computer algorithms to solve this problem have worked by choosing one person at a time, beginning with people who have the rarest qualities needed to meet the quota. Because of the way these algorithms work, some volunteers had essentially no chance of being selected for the panel.
So Flanigan and her colleagues came up with a different approach. Rather than select participants one at a time, their algorithm generates a list of the many possible panels that all meet the quotas. Then, they tweak the likelihood of selecting each panel to even out the odds that any individual person will end up in the assembly. In addition to spitting out a panel at the end, the approach offers the opportunity for full transparency—letting volunteers see the odds they could end up on the final assembly.
The researchers have made their algorithm, which they dubbed Panelot, available for public use, and Procaccia said it’s already been used in selecting more than 40 citizens’ assemblies.
“It’s testament to the potential impact of work in this area that our algorithm has been enthusiastically adopted by so many organizations,” Flanigan said. “A lot of practitioners were using their own algorithms, and the idea that computer scientists can help centralize efforts to make sortition fairer and more transparent has started some exciting conversations.”
What’s Next?
Receiving the Hertz Fellowship helped boost Flanigan’s confidence that she can make a difference with her line of research, she said. And citizens’ assemblies aren’t the only thing she’s been working on. She also has other ongoing projects related to voting and political polarization.
There are, however, questions she still might tackle when it comes to citizens’ assemblies. How many quotas should be imposed, and of what types? If a volunteer backs out of an assembly after the panel has been chosen, what’s the fairest way to select a replacement? And can a similar method be used to select people for clinical trials that have diversity quotas?
“I’m still reading about issues and talking with people, trying to find the next best questions to focus on,” Flanigan said.
RSS