AI holds the solution to redistricting issues
Could robots solve gerrymandering? That might sound like a question from a C-SPAN version of Black Mirror, but in an age of AI and algorithms it is a question worth pursuing. Besides, it certainly couldn’t be worse than our current system of redistricting.
There is no system in our democracy more steeped in partisan politics than the act of drawing political boundaries. The United States is unique among global democracies in its practice of allowing elected officials the ability to draw their own district lines. In effect — politicians select the very voters who will elect them. The result of this bizarre process has been increasingly non-competitive districts that are designed to elect members from a particular party. Here in Houston we see that in almost every district in the area — with neighbors often finding themselves in completely different political districts.
It is an issue being confronted by the U.S. Supreme Court, and hearings about a Texas-specific case are scheduled for later this month.
Politicians are supposed to apply “traditional redistricting principles” which primarily consist of 4 factors — compactness (related to district shape), contiguity (all parts of a district touching or connected), communities of interest (specific groups with shared interests/identity), and political competitiveness. In actuality, lines are often drawn to protect incumbents and preserve single party domination. The more information that is known about voters — the stranger the districts appear.
Technology has in many ways contributed to the proliferation of single-party districts. With the wide scale adoption of geographic information systems (GIS) technology along with algorithms analyzing voter preference, map drawers are able to design districts at a block level using information from census data, voting history and neighborhood-specific demographic data. As technology allowed the identification of specific voters in a given area, voting districts were increasingly forged into extreme shapes and dimensions splitting immediate neighbors in favor of far flung ones.
Against this backdrop the U.S. Supreme Court is poised to rule on several cases that could lead to profound changes in the redistricting system. Challenges to redistricting maps based on ethnic and racial discrimination is not uncommon. This year alone, Texas and North Carolina have seen lower courts strike down district maps for discriminating based based on race. District lines can not pack or divide ethnic and racial communities in an attempt to minimize their voting strength. Under the doctrine of “one-person one vote,” districts must not be constructed to intentionally dilute the ability of a voter to elect a candidate of his or her choosing.
This year, the Supreme Court will take that doctrine and decide if extreme partisan gerrymandering should also be prohibited. The Wisconsin case of Gill v. Whitford challenges the legality of Wisconsin maps based on maps that have packed and divided communities based on partisan affiliation — again looking at detailed voter data to determine lines. The outcome of this case could profoundly change how we approach redistricting. With a new census just around the corner, we are quickly approaching a new round of redistricting beginning in 2020. Regardless of what the court ultimately decides — the larger question is whether the system itself needs to be reformed, and if so what can best be done to solve the problem?
While many states have experimented with non-partisan voting commissions or citizen panels — neither approach is immune to the machinations of politics in our increasingly partisan and political world.
A potential answer to the dilemma of redistricting may lie in technology. The same technology that turbocharged gerrymandering may hold the key to redistricting reform. Algorithms, or more precisely artificial intelligence, hold the possibility of a process to have truly independent and fair redistricting process. Inputs can be clearly defined and a “blind” construction of districts can be developed that deemphasizes provincial politics from the process.
Data and algorithms can provide a level of transparency in the previously closed world of redistricting by specifically detailing the various inputs and parameters that led to a district being drawn. More important with these inputs and parameters made transparent, the public can have a more direct role in the process, holding politicians more accountable to the outcome.
Traditional redistricting has been a closed process, with politicians shifting boundary lines based entirely on political calculations. A machine based system functions entirely based on broad inputs, governed by a defined code, not politics. While a system like this may seem years away, working prototypes have already been developed. Data scientists from the University of Illinois applied an AI based system to redistricting in four states — Arizona, Massachusetts, New Mexico, and New York. The result were less politically polarized districts that were contiguous, diverse and better representative of the general population.
The presence of political bias in our redistricting process runs the risk of undermining our democracy and core values. Any process that reduces that bias in the system needs to be considered. In the end, our system can only function if our democratic institutions are open and competitive.
Aiyer is an assistant professor of public administration in the political science department, Barbara Jordan Mickey Leland School of Public Affairs at Texas Southern University, where he teaches urban politics and policy.