If you happened to have access to infinite monkeys typing on infinite computers for an infinite amount of time, you might eventually end up with the perfect redistricting map.
Or you might end up with infinitely frustrated monkeys, because drawing these political maps is a painstaking, tedious ordeal, and monkeys generally have better things to do with their time.
Recently, finding myself without better things to do, I tried my hand at crafting a simple four-district congressional map and after four hours of work ended up with something decent. Not perfect, but fine.
It’s an arduous task, though, because districts have to have a precise number of residents, be reasonably compact, avoid slicing up cities and towns, and every time you move a cluster of people from one district to another it can create a cascading effect in surrounding areas.
Now imagine having to create multiple maps, not just for four congressional seats, but also for 29 state Senate districts, 75 state House districts, and 15 state school board districts. It’s almost incomprehensible.
So the two commissions responsible for drawing new maps are relying to some extent on crowdsourcing — getting the public to submit maps for consideration. And that’s fine. The public is smart and capable and should be encouraged to provide input.
But there is a better way — one that is faster, more precise, data-driven, might not cost anything, and offers a level of transparency that can alleviate (or at least shine a light on) some of the political gamesmanship that permeates the process.
Think of drawing these maps as a big math problem, not one with a correct answer, but where we’re trying to find the best option. We can spend dozens of hours solving it ourselves or we can tell a computer to try to solve it millions upon millions of times and give us the best results.
“The benefit of computational redistricting is it takes a lot of the politics out of the process,” Sheldon Jacobson, a computer science professor who runs the Institute for Computational Redistricting at the University of Illinois, told me.
“We simply have algorithms where you give us information and inputs based on the constitutional requirements — compactness, contiguity, whatever you want to put in the maps — and we’ll come up with maps that we’ll also score based on different metrics,” he said.
Jacobson’s team recently created 15 Illinois congressional maps, each one with strengths and weaknesses and each score reflecting those trade-offs. Maybe one keeps communities of interest together while another is more compact. Whatever the criteria, it gives the public and policy makers actual quantifiable data so they can decide which they prefer based on their priorities.
Facing tight deadlines to get the maps done, there’s another huge advantage: Jacobson said that, depending on the rigidity of the constitutional requirements, his team can churn out maps in as little as a half hour to a few hours.
They can also score maps offered by the redistricting committee or the public to identify weaknesses and sound alarms if it departs from the expressed criteria enough that it appears to have been intentionally gerrymandered. This kind of computer analysis has been key evidence presented in court challenges to redistricting in other states.
After a court challenge in North Carolina, officials had a computer create 1,000 new maps that met the legal criteria, narrowed the field to five, and had a Bingo machine randomly spit out a ping pong ball to decide the districts. As wild as that seems, there are a lot worse ways to run a democracy.
Jacobson’s group is now working with Arizona’s redistricting commission and, he said, is willing to work with any state — for free (although donations to support the work are appreciated).
“We’re trying to be a public service to the country,” he said. “Any state that comes to us, as long as they are serious and willing to, in fact, use what we give them, we will do it at no cost.”
Rex Facer, the chair of Utah’s Independent Redistricting Commission, told me that commissioners looked at using artificial intelligence early on, but decided it would be a better tool for evaluating maps submitted by the public, rather than creating maps at the outset.
Tremonton Sen. Scott Sandall, the co-chair of the Legislature’s redistricting committee, said they considered using algorithms to draw maps, but at the end of the day decided that it still required making subjective judgments on which criteria to prioritize — is a compact district more important, for example, than one that respects city or county lines?
He’s right, but those subjective judgments are always part of redistricting. The algorithms just make the trade-offs more transparent.
Something else to consider: Even if the committees don’t use artificial intelligence in redistricting, you know the political parties trying to manipulate the outcome to their advantage absolutely will. That’s because artificial intelligence is — to borrow from Star Wars — like The Force, a power that can be used for good or evil.
“We could create gerrymandered maps better than the gerrymanderers,” Jacobson said, although they choose not to. “It’s a valuable tool but, yes, if it’s put in the wrong people’s hands it can be used nefariously.”
The best way to keep technology from being weaponized is to use it proactively to put forward maps with transparent scoring and without the taint of backroom political dealing.
Even then, perhaps it won’t matter what tools we use. Ultimately, the Republican majority in the Legislature still has — as they have always had — the power to adopt whatever maps they want, whether those are in the public’s interest or their own.
But that doesn’t mean those of us who want fairer maps should give up and default to doing things the way they’ve been done for the past 30 years. We have better tools at our fingertips that could give us empirically better boundaries, more competitive and meaningful elections, hopefully yielding better representation. That’s the democratic ideal.
At the very least, it might make life easier for infinite monkeys.
Follow @RobertGehrke