top | item 23611157

Show HN: Antimander – Optimize Congressional Districts with Genetic Algorithms

68 points| joelS | 5 years ago |antimander.org

50 comments

order
[+] splitrocket|5 years ago|reply
"I cut, you choose" is really the only politically viable alternative to the current state of affairs.

https://slate.com/news-and-politics/2018/02/how-the-i-cut-yo...

[+] chrisbrandow|5 years ago|reply
It would require a more fundamental recognition of political parties than the constitution provides.

Gerrymandering is not a technological problem.

[+] specialist|5 years ago|reply
I've been trying to dream up ways to use general purpose strategies like this to mitigate regulatory capture. Applicable to two party systems with no clear obvious third party to forge a balance of powers, a durable trilemma.
[+] antishatter|5 years ago|reply
Ah that's very clever, like when my sister and I split a candy bar in grade school. It really worked then as it would now.
[+] jojobas|5 years ago|reply
>Some interesting patterns emerge, such as how with you can have compact and republican-biased but not compact and democrat-biased maps for some numbers of cities. This is a direct result of one party being forced into a small geographic area (cities).

That's quite an interesting finding. I always understood that gerrymandeting is bad because the artificially winning parties get congressmen that are not representative of their districts. This guy uses "competitiveness" as a property worth optimizing for, achieving exactly that.

[+] specialist|5 years ago|reply
I've long advocated maximal competitiveness. Mostly to motivate voter participation.

I love how antimander illuminates the tradeoffs.

Along the lines of "preserving community", preserving continuity can also be a factor. On the presumption that voters don't want to have their congressional district designation changed every 10 years. For example. I know this is a consideration during redistricting.

--

Antimander is real progress. The cites are terrific. I'm delighted that I even learned some new things, like the Seats-Votes Curve.

I've spoken with the Dave Bradlee many times over the last 15 years. He created a redistricting app that got some national media attention last cycle. https://en.wikipedia.org/wiki/Dave%27s_Redistricting

Some of those talks were pretty bleak. I'm so glad so many more people have engaged with this issue.

[+] c54|5 years ago|reply
This is really well done! Do you have analysis of what this would do nationwide or across multiple states? Optimizing for competitiveness is a metric I hadn’t considered but seems particularly valuable especially towards increasing voter turnout.
[+] jojobas|5 years ago|reply
How is it valuable? Having a "red" or "blue" district means the congressman is somebody most people can get behind, "our guy in the parliament".

Being around 50% gives you a compromise that nobody trusts and nobody feels the stake in their country.

[+] rpedela|5 years ago|reply
On Android and Chrome, I can't read the page. The graphic takes up half the screen and the text underneath is barely visible.
[+] pdxandi|5 years ago|reply
I had to scroll all the way to the bottom to see if anyone could even read the site on mobile.
[+] gillesjacobs|5 years ago|reply
Gerrymandering is a danger to democracy and it is great that researchers shine light on the different trade-offs of districting criteria.

This website seems like a great demonstration of the fairness and competitiveness trade-offs.

[+] usrusr|5 years ago|reply
I'd argue that every voting system that has gerrymandering as a possibly error mode is inherently dangerous. No amount of antimandering could ever be more than a temporal band-aid. There are many examples that successfully combine local representation with proportional outcome.
[+] steve_gh|5 years ago|reply
Interesting. GA's have some interesting use cases. We used a GA to optimize assignment of work crews to areas as part of a scheduling optimization, and achieved significant benefits

Steve

[+] kwaldman|5 years ago|reply
Seen similar benefits - many years ago applied GA's to schedule technicians to fix computers (based on location, part availability, skill set, etc) and also for manufacturing production lines (labor rules, parts, color changeovers, demand, margin)
[+] IgorPartola|5 years ago|reply
As Congressional elections become more and more a national affair, why do we still have districts? Say a state gets 24 reps. Give me a ballot with all the candidates and let me mark the 24 I want. Yes the ballot will be more complicated, but this way we entirely avoid districting and the process is entirely fair (well until you consider states not being fairly drawn, but that’s one level up).
[+] xahrepap|5 years ago|reply
IMO, lifting the cap on the number of representatives we can have as a nation is the best way to fight this issue. Representatives are supposed to represent the people in their district. If each representative represented no-more than, say, 200k people, then gerrymandering would almost be impossible and the people would be better represented.

Yes, we'd have a LOT more representatives. But we also now have the technology to support such a thing.

> As a result [of the Reapportionment Act of 1929], the average size of a congressional district has tripled in size—from 210,328 inhabitants based on the 1910 Census, to 710,767 according to the 2010 Census. https://en.wikipedia.org/wiki/Reapportionment_Act_of_1929

[+] coldcode|5 years ago|reply
Likely it means that big cities would get all the reps and the rest of the state gets nothing.
[+] antishatter|5 years ago|reply
That of course defeats the purpose of smaller districts getting a voice and not being crushed by larger districts.
[+] VWWHFSfQ|5 years ago|reply
Why stop there? Why do small states get the same number of Senators as large states? All the Senators should just be from California, Texas, New York, and Florida.
[+] charleskinbote|5 years ago|reply
Along the same thread, genetic algorithms have been used to evaluate fair land allocation in Brazil [1], which have also historically been done by hand. I agree with another user that pointed out the very same tools can be used to gerrymander even more effectively than those districts are currently, perhaps in less obvious ways, without the oddly shaped borders that stretch wildy from place to place. It would be important not to have this occur behind closed doors but left open to scrutiny where the fitness function and biases are in full view.

[1] https://www.lume.ufrgs.br/bitstream/handle/10183/174950/0010...

[+] xorfish|5 years ago|reply
How hard would it be to change the congressional election system to proportional voting by party?

In California you would vote for 53 representatives, each representative would be part of a party. First the seats are divided among all parties based on some system. Then within each party the seats are given to the representatives that have the most votes within the party.

This would introduce new parties to the system, no party would be able to govern alone.

[+] cbuq|5 years ago|reply
Would this require me to make 53 different votes? This sounds a like pretty complicated situation to expect educated votes for all 53 votes.
[+] FiReaNG3L|5 years ago|reply
Interesting to note that the very same tool can be used to optimize the 'worst' maps as well.