The House of Representatives is made up of individuals representing state districts apportioned based on population. In the colonial era, districts were defined by the borders of towns and counties, and as the population grew, some districts kept abreast by increasing the number of representatives while others did not. This uneven districting translated into uneven House representation for residents in counties where the state constitutional requirements to readjust district lines were either neglected or misinterpreted.
To learn more about the history of House districts, ESAL spoke with Kyle Kondik, a managing editor at the University of Virginia Center for Politics. Kondik told us:
“In the mid-1960s, a Supreme Court ruling (Baker v. Carr) mandated that congressional district boundaries be redrawn after each census, typically every 10 years. This is termed redistricting and enforces two ideas: equal population across all districts and, importantly, one person, one vote. In 2026, the House of Representatives is facing a combination of unprecedented events disrupting historical norms for House representation: mid-decade redistricting, which has not occurred since 2005, and the emergence of AI analysis tools.”
Redistricting is necessary to reflect population changes after a census. When conducted properly, it is a fair and equitable process that ensures voters pick their representatives, not the other way around. Gerrymandering occurs when redistricting is used strategically to manipulate district lines to favor a political outcome or discriminate against certain groups through vote dilution. This can be done by “packing” high numbers of voters into a small number of districts or by “cracking,” when similar voters are split across many districts to prevent them from having a majority. In such instances, redistricting ceases to be equitable. While gerrymandering based on political belief is legal, Section 2 of the 1965 Voting Rights Act prohibited gerrymandering based on race. However, the Voting Rights Act was reduced in scope by the 2026 Supreme Court case Louisiana v. Callais, making racial gerrymandering significantly harder to prevent.
The introduction of AI has provided mapmakers with new tools and algorithms that can draw and analyze maps of a state with unmatched power. Legal and political experts foresee their use by both political sides in gerrymandering attempts. However, while AI can review and analyze maps, there are limitations to what it can achieve because human input will remain essential due to vital conversations between mapmakers and party leaders that kick off redistricting.
As Kondik explained, the most useful data for drawing maps depends on their purpose. With partisan gerrymanders, election data is the most useful. Party registration has been used in places where people have registered to vote for a particular party, though this doesn’t always reflect actual voting, because people change their minds. In 2024, we saw that Republicans did much better with nonwhite voters, particularly Latino Americans; however, using the 2024 results to draw new maps for 2026 might lead to disappointment. In Texas, Republicans drew a new map where the number of districts that voted red was increased by almost 5 points, but these might not translate in the upcoming elections. Additionally, there are tools such as Dave's Redistricting App, which professionals and hobbyists have used to create and share redistricting maps. Potentially, there is also demographic data, but the court has banned the use of racial data for redistricting (this power has been diluted by Louisiana v. Callais).
As with every data set, the information used to draw districts has blind spots and biases. “COVID caused an explosion in vote-by-mail, but in Virginia, for instance, the mail-in ballots weren’t reallocated by precinct within counties, and so results for within counties were not as accurate. So long as the ballots are allocated properly, you generally have a good idea of what goes on, even with a little bit of fudging. I don’t think there are biases in the data, but of course, the person or entity drawing the map will be biased against the other party. Some states have a redistricting commission to acquire fair maps, but everyone’s idea of fairness is different, and so any bias that occurs is intentional,” Kondik said.
ESAL also probed about the future success of AI in this field. Kyle said, “There are so many things that are being outsourced to AI already. Academic literature before AI shows simulations, and AI could be useful for such purposes. Ultimately, there is a group of people who decide what the map looks like, and so these decisions will always come down to people. There could be a future where there is legislation on Gerrymandering and set up stringent criteria on what districts should look like - AI could play a role here. A fair way to do it is to outsource this data to the machines rather than having one side or the other decide on these maps. I would say that AI-human collaboration is to be expected in the coming years.”
While AI can revolutionize the speed of redistricting, regulatory red flags are being raised to curb bipartisan gerrymandering, prevent dilution of minority representation, and prohibit data manipulation. One remedy would be establishing court-appointed nonpartisan experts who validate scientific and technical information and aid the courts in reaching decisions on redistricting cases, and that selected experts be reviewed by nonpartisan judges after each redistricting cycle. Another solution is the introduction of independent redistricting commissions, which would prevent partisan politicians from leading redistricting efforts that give their own party an unfair advantage. While these commissions could still use AI tools, they would do so in the name of equity. In Michigan, using an independent redistricting commission has already resulted in one of the fairest maps in the U.S. today.
