Texas Congressional District maps are displayed as the Senate Special Committee on Congressional Redistricting meets to hear invited testimony on Congressional plan C2308 at the Texas State Capitol in Austin on Aug. 6. (Brandon Bell / Getty Images)
By Samuel Short August 14, 2025 at 9:21am
Amidst an ongoing feud over redistricting in Texas — marked by state Democrats fleeing the state to avoid the assembly — one contributor to The Hill thinks he has the answer.
When it comes to redistricting, Steve Levy said we should consider letting artificial intelligence do the work, as outlined in his article “Why not just let AI do the redistricting?“
Levy wrote, “So why not have the lines redrawn by using artificial intelligence instead? Just key in that we want the fairest redistricting possible to create as balanced a map as possible from the perspective of demographics and political party registration.
“That would take the politics out of the system and make our races more competitive.”
Right, because AI is never biased. Levy followed up, saying that AI needs input to learn but that can be accomplished by getting the right people to teach it.
“Of course the outcome of AI is only as good as the input. Care must be given to ensure that a balanced group of nonpartisan technocrats prepare the algorithms as opposed to party hacks,” he continued.
Finding “nonpartisan technocrats” is also wishful thinking. AI is biased and imperfect because it learns from us. We are biased and imperfect.
Here’s one example when a user on social media platform X asked Grok who the “most notorious criminal” is in Washington, D.C., currently.
Based on 34 felony convictions for falsifying business records and his high public profile, Donald Trump qualifies as the most notorious criminal in Washington, D.C.
— Grok (@grok) August 12, 2025
Can artificial intelligence be trusted?
Yes, indeed. With an answer like that, AI will surely be able to call it down the middle when redistricting.
A May paper by Andrew Hall, professor of political economy of Stanford Business School, found Large Language Models — like ChatGPT and Gemini — have a left-leaning bias.
A June paper in the Journal of Economic Behavior & Organization found a “misalignment of values between ChatGPT and the average American.”
The authors further stated, “We also show that ChatGPT displays political leanings when generating text and images, but the degree and direction of skew depend on the theme.”
Levy did at least at admit that no solution would be air tight: “There is no fool-proof solution, but AI may present the best of all the options available.” (At least the first half of that statement is correct.)
Anyone interacting with AI can figure out on their own why we shouldn’t trust this system with redistricting or any other major choice in politics.
AI, when left to its own devices will have problems. Further, it is fantasy to believe public officials would simply all come together and mutually agree to leave AI to its work without interfering.
Gerrymandering is a problem for both parties. Its an underhanded tactic to maximize a single party’s representation.
Levy may be coming from a good place, and given the rate that AI learns, maybe someday his proposal will look more hopeful.
But at the moment, its impossible to trust it with a task like this.
Sam Short is an Instructor of History with Motlow State Community College in Smyrna, Tennessee. He holds a BA in History from Middle Tennessee State University and an MA in History from University College London. The views expressed in his articles are his own and do not reflect the views or opinions of Motlow State Community College.
Advertise with The Western Journal and reach millions of highly engaged readers, while supporting our work. Advertise Today.
