Mathematicians are deploying algorithms to stop gerrymandering

For decades, one of the users was Thomas Hofeller, “Michelangelo of the modern gerrymander,” has long been the official re-election director of the Republican National Committee, and he died in 2018.

The Gerrymandering plan includes “cracking” and “packaging”-spreading a party’s votes to different regions, thereby diluting their power, and squeezing like-minded voters into one region, wasting the power they have elsewhere. The city of Austin, Texas was cracked and divided into six districts (it is the largest non-districted city in the United States).

In 2010, the Republican Party’s Redistribution Majority Project (REDMAP) made the full power of the threat a reality. It spent $30 million in state legislative campaigns and won victories in Florida, North Carolina, Wisconsin, Michigan, and Ohio. “The victory in 2010 gave them the ability to map in 2011,” said David Daly, who was, Ratf**ked: The true story behind the secret plan to steal American democracy.

“What was once a dark art is now a dark science.”

Michael Lee

Since the last re-division of constituencies, the technology has made rapid progress, which will only exacerbate the results. “This makes the gerrymanders drawn that year more durable and lasting than any other gerrymanders in our country’s history,” he said. “The complexity of computer software, the speed of the computer, and the amount of data available allow partisan cartographers to perform 60 or 70 different iterations of their maps and truly improve and optimize the partisan performance of these maps.”

As Michael Lee, a redivisioning expert at the Brennan Judicial Center at New York University School of Law, said: “It was once a dark art, but now it is a dark science.” Almost impossible to overcome.

Mathematics microscope

Martinly and his Duke Having stayed up all night to write code, they expect to produce a “huge victory” in the algorithm-to prepare for the practical application of their latest tool, which debuted in a paper (currently under review) under the title of Technical Order Excited”Multi-scale merge split Markov chain Monte Carlo for repartition. “

However, advancing technical discussions is not a priority. Mattingley and his colleagues hope to educate politicians and the public, as well as lawyers, judges, mathematicians, scientists-anyone interested in the cause of democracy. In July, Mattingly gave a public speech, the title is easier to understand, and the cut-in is quick: “Can you hear the will of the people in the vote?“

Deformed areas are usually considered to be Grimand’s hallmarks. On the 2012 map of North Carolina, the congressional districts are “very weird beasts,” Mattingley said that he (and his main collaborator, Greg Herschlag) provided expert testimony in subsequent lawsuits. Over the past decade, all parts of the country have faced legal challenges—Illinois, Maryland, Ohio, Pennsylvania, and Wisconsin.

But even though these disfigured areas “made very beautiful posters, coffee cups, and T-shirts,” Mattinley said, “but the truth is that stopping strange geometric shapes does not stop classifying.” In fact, all technologies are used.​​ With complicated techniques, a messy map is hard to find.

Many tools developed by mathematical scientists at the same time provide so-called “extreme outlier tests.” Each researcher’s method is slightly different, but the result is as follows: Compare the map suspected to be classified with a large collection or “collection” of unbiased, neutral maps.Mathematical methods at work-based on the so-called Markov chain Monte Carlo algorithm— Randomly generate map samples from possible maps and reflect the probability that any given map drawn meets various policy considerations.

The integrated map is coded to capture the various principles used to map regions, taking into account the way these principles interact with the country’s geopolitical geometry. Principles (which vary from state to state) include standards such as keeping areas relatively compact and connected, making them roughly equal in population, and protecting counties, cities, and communities that share common interests. The regional map must comply with the U.S. Constitution and the Voting Rights Act of 1965.

As the Census Bureau releases 2020 data, Mattingly and his team will load the data set, run their algorithms, and generate a series of typical non-partisan district plans for North Carolina. From this huge map distribution, and taking into account the historical voting patterns, they will identify what should be the benchmark for the guardrail. For example, they will evaluate the relative likelihood of these maps producing various election results—for example, the number of seats won by Democrats and Republicans—and what advantage they have: 50-50 voting, and give a reasonable voting model, neutral map It is unlikely to give 10 seats to Republicans and only three seats to Democrats (like the 2012 map).

“We are using computational mathematics to determine our expected results for unbiased maps, and then we can compare against specific maps,” Mattingly said.

By mid-September, they will announce their findings, and then hope that state legislators will pay attention to the guardrail. Once new regional maps are proposed later in the fall, they will analyze the results and participate in the ensuing public and political dialogue-if these maps are again suspected to be wrong, there will be more lawsuits, among which mathematicians Will file a lawsuit again to play a central role.

“I don’t want to just make someone believe that something is wrong,” Mattingly said. “I want to give them a microscope so they can look at the map and understand its properties, and then draw their own conclusions.”

Courtesy photo

When Mattingly testified in 2017 and 2019, analyzing two subsequent iterations of North Carolina’s regional maps, the court agreed that these maps were excessive partisanship and discriminated against Democrats. Wes Pegden, a mathematician at Carnegie Mellon University, used a similar method to testify in a Pennsylvania case; the court agreed that the map discriminated against Republicans.

“The court has long been trying to solve the problem of how to measure party constituencies,” Li said. “But then there seemed to be a breakthrough, when one court after another used some of these new tools to delete maps.”

When the North Carolina case entered the U.S. Supreme Court in 2019 (together with the Maryland case), Eric Lander, a mathematician and geneticist who is a professor at Harvard University and MIT and is currently President Biden’s top scientific adviser A briefing stated that “computer technology has solved the problems it created.” He believed that the extreme outlier criterion-a simple question, “Which part of the redivision plan is not as extreme as the proposed plan?”-a ” Direct and quantitative mathematical problems, there is a Correct Reply. “

Most judges Another conclusion.

“The five judges of the Supreme Court seem to be the only ones who are confused about how mathematics and models work,” Li said. “State and other federal courts have managed to apply it-it’s not beyond the intellectual capacity of the courts, like complex sex discrimination cases or complex securities fraud cases. But the five Supreme Court justices said,’This is for us It’s too difficult to say.'”

“They also said,’This is not what we are going to solve-this is the state to solve; this is For amendment by Congress; This is not something we can solve,'” Li said.

Does it matter?

As Daley saw, the Supreme Court’s decision gave state lawmakers “the green light for the kind of partisan sheriff that can be made when the map is drawn later this month, and there is no speed limit.” At the same time, he said, “ Technology has improved to where we can use it now [it] See through the technology-driven gerrymander created by the legislator. “