February 01, 2022

Article at nydailynews.com

View original

Want fair redistricting? Use artificial intelligence

Over 200 years ago, the “Gerrymander,” a salamander-like legislative district, was mocked in a newspaper cartoon. The Massachusetts governor at the time, Elbridge Gerry, had put his signature on a plan to redistrict a state senate in 1812 in order to benefit his party and create additional seats. The Boston Gazette presciently baptized the grotesque redistricting a “new species of monster.” It has been just that ever since.

New York City congressional map
New York City congressional map (New York Democrats)

The once-a-decade freak of nature reveals itself unapologetically and prepensely in 39 state legislatures. The process of redistricting is a blighted affair marked by political meddling and backroom dealmaking in which parties attempt to maximize the number of seats they can win while also securing the incumbency of the current occupants. Attempts to create independent commissions, such as recently in New York, often fail to produce district lines that can be agreed upon. When an impasse occurs in New York State, the decision reverts back to the Legislature, which has now finessed a vagarious map of 26 congressional seats.

There is a better way to reduce, or entirely eliminate, political manipulation in order to draw fairer boundaries and rebuild trust with the electorate: automate redistricting with artificial intelligence.

Technological advances over the past decade have made it easier than ever for AI to generate lines much more effectively and less arbitrarily than partisan stakeholders. Research that has studied machine-learning algorithms for redistricting shows the potential for what these tools can do. AI models can consider inputs such as race, and identify the extent to which the quality of a map has been affected by gerrymandering, or what’s referred to as “the efficiency gap.” Oftentimes the practice of “packing” (clustering certain voters together to limit their influence to a single district) and “cracking” (separating blocs across districts to dilute their voting strength) are distinctly obvious. Other times, the manipulation within a map is so subtle that an “eyeball test” won’t reveal a salamander-like district, requiring those asymmetries to be identified through the use of an algorithm or in statistical tests.

There is a precedent for computational redistricting. In 2019, North Carolina’s state Senate moved to use a lottery machine to select a handful of maps from 1,000 AI-drawn maps. What’s more, while the Supreme Court has unfortunately ruled that partisan gerrymandering is constitutional, it hasn’t ruled out the use of technology in making these delineations, so long as state legislatures allow the computers to take over. Open-source software such as AutoRedistrict can allow mapmakers to meet criteria including geometry and equality such as contiguity, compactness proportion, minimal partisan influence and the elimination of racial gerrymandering.

With the algorithm in public view, it offers transparency to a previously murky process. The added benefit, of course, is that long and grueling legal battles that cost taxpayers tens of millions of dollars every year and add to institutional trust deficit can likely be reduced or altogether eliminated if the process for electoral cartography is improved.

At present, institutional efforts in the New York State Senate to retain power have drawn up a proposal for a congressional district that sinuously winds through the Upper West Side making its way through Brooklyn all the way to Bensonhurst. Is it fair for these communities with dissimilar needs to be joined together as incongruously as the shape of a serpent? Would any elected official be capable of best representing a district drawn as arbitrarily?

On WNYC’s “The Brian Lehrer Show” Tuesday, Sen. Mike Gianaris, deputy Senate majority leader and co-chair of the Legislative Task Force on Reapportionment and Redistricting, avoided the question of why AI can’t be used to help draw up electoral maps. Computer models can make the districts make sense more than any partisan player can.

Computational design for redistricting bestows impartiality as much as it can, and provides clarity via input variables to any otherwise political process. New York’s move to create an independent commission may have been a good-faith attempt to eliminate biases or it may have been a smokescreen. Come what it may, when democratic processes are marred by conspicuous wheeling and dealing, it erodes confidence in legislatures and Congress.

While it still can, New York should scrap its current redistricting proposal and bring in nonpartisan researchers to provide fairer, AI-drawn maps for districts so that constituents can be better served.