Redistricting maps by humans gerrymandered Wisconsin. Powerful algorithms could do better.
The Wisconsin Supreme Court has received two petitions calling for the redrawing of the state's legislative map, claiming it is extremely unfair to Democrats. Both petitions observe that despite the near 50-50 split in the statewide vote in recent elections, Republicans have represented about two-thirds of the districts in the Assembly and Senate for more than a decade.
The petitions argue that the current map, drawn by Republicans, should be replaced with a new, fairer map that is just and equitable to Democrats and Republicans. The key to fairness is proportional representation: the percentage of districts represented by each party should reflect the percentage of the total vote cast for each party in recent statewide elections.
If the court decides to discard the current map, it will face the daunting challenge of identifying a good replacement. Wisconsin has more than 200,000 census blocks, each needing to be assigned to an Assembly and Senate district, with the goal of achieving an optimal balance among numerous features that define a good, fair map.
Computer algorithms should draw Wisconsin redistricting maps
There are more ways to divide Wisconsin into 99 assembly and 33 Senate districts of equal population than there are atoms in the universe. Among these possible maps, only a tiny fraction also have contiguous districts with uninterrupted boundaries. Significantly fewer maps also satisfy the federal Voting Rights Act. Even fewer also have nice looking, compact districts. Of these, a diminishingly small number of maps are likely to give proportional representation to Democrats and Republicans. And if any such maps are found, far fewer will also have a respectable number of competitive districts.
There is, however, a bright spot: map performance can be quantified. Numerical scores can be attached to all map criteria — contiguity, federal law compliance, compactness, proportionality, competitiveness, etc. — to help Supreme Court Justices decide which map is the best. This quantification places redistricting squarely in the mathematical realm, particularly within the domain of mathematical optimization.
Mathematical optimization is an academic discipline that focuses on finding the best options, among an astronomical number of possibilities, when options can be mathematically scored. Distinct from artificial intelligence and machine learning, mathematical optimization is used throughout the private sector to improve decisions in areas where managerial possibilities are nearly endless, including airline scheduling, factory machine scheduling, package delivery truck scheduling, warehouse operations management, cancer radiation therapy, and hospital staffing. When it is used, it is always implemented with a computer algorithm. Many companies could not survive without these algorithms.
This is no HAL 9000. Humans in control of process of creating electoral districts
Should computer algorithms be involved in redistricting? Yes! Currently, district maps are being constructed by humans who manually click and drag a computer mouse to build maps piece by piece as they would a huge jigsaw puzzle. Such mapmaking is by nature subjective, slow, prone to human error, and, therefore, highly unlikely to achieve a good balance among the multiple criteria that define a fair map. On the other hand, a powerful algorithm, guided by mathematical principles alone, can quickly find exceptional maps without error or bias.
Will humans lose control if an algorithm is used? Not at all. Humans still control the process. An algorithm simply gives humans the ability to make maps to order as a tool, following instructions and objectively carrying out work assigned by humans. Decisions about all input data supporting the computational process, including the historical election data needed to judge a map’s proportionality, are made by humans.
A good algorithm will produce a different map every time, even when its settings are unchanged. So when, perhaps, 100 maps are generated by computer, humans decide which one, if any, to use. Manual finishing touches may then be added if desired. Or the algorithm can be ordered to produce new maps that more strongly satisfy a particular criterion. An algorithm empowers human decisionmakers because it gives them quick insight and feedback into what is possible and what is not possible for a district map.
My redistricting algorithm, FastMap, generates a great diversity of maps based on settings decided by the human user. It allows political redistricting to be done in a new, rigorous, and fair way, and it has an excellent track record. Before the Wisconsin Supreme Court ruled that the 2022 map should be “least change” relative to the map in force from 2012-2021, FastMap made a state legislative map that scored the best in proportionality, compactness, and competitiveness among all legally acceptable maps made in 2021, including those by the People’s Maps Commission, Wisconsin Senate Democrats, and all maps submitted to the legislature’s Draw Your District portal.
After the court ruled that the 2022 map had to be “least change,” FastMap was recalibrated, and it again outperformed these maps, this time according to the least-change criterion. Don Leake, of the Wisconsin Maps Assessment Project and a professor emeritus of Mathematics at UW-River Falls, used the map analysis tools at DavesRedistricting.org to do the scoring. Dave's Redistricting is a team of volunteers who want to empower civic organizations and citizen activists to advocate for fair congressional and legislative districts and increased transparency in the redistricting process.
More perspectives on political maps, impeachment
Robin Vos' change of heart doesn't mean he champions fair election maps for Wisconsin
Republicans want to cancel the votes of more than a million Wisconsinites
Again, another map made by FastMap did what many observers — including PolitiFact and parties to the court case — said was impossible: it created a strictly proportional legislative map. View the map online here: DistrictSolutions.net/Wisconsin-Maps.html.
In the 21st century, it is unimaginable that algorithms common to the private sector are not also being used in what is arguably our nation’s most important optimization problem — political redistricting.
As Wisconsin considers redrawing its legislative map, let’s replace an outdated and vague method of mapmaking with one that introduces power, precision, clarity, and transparency to the process. A computer algorithm can provide voters with a legislative map that is rigorously fair in all aspects. Computer algorithms are the best solution to end gerrymandering.
Matt Petering is owner of District Solutions LLC, a redistricting consulting company, and an Associate Professor of Industrial & Manufacturing Engineering at the University of Wisconsin-Milwaukee.
This article originally appeared on Milwaukee Journal Sentinel: If Wisconsin tosses unfair electoral maps, use algorithm for new ones
