Data Science Can Solve Gerrymandering | Frank Evans | TEDxUCO

Determining congressional districts is traditionally a behind-the-scenes process with a track record of interesting results, even though it is crucial to our democracy. However, using the tools of data science it is possible to define open ways to cluster states into congressional districts in a fair way.
Frank D. Evans is a Data Scientist with Exaptive, Inc., an Oklahoma City based data analytics software company. He works heavily with big data technologies with a primary interest in machine learning and feature engineering. Frank additionally is a 2008 graduate of the UCO College of Business with a Masters in Business Administration, and a 2014 graduate of Johns Hopkins University with a Masters Specialization in Data Science.
This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at ted.com/tedx

[@1locust1]

Why the news media doesn't discuss this solution to gerrymandering baffles me to no end.

[@MusikCassette]

he doesn't either, or did he say anything about proportional voting?

[@theseelig]

This is a very good, and thought provoking talk. And you do not get nearly as many laughs are you deserve. That said, I have many dissentions (which, for me, is the sign of somthing inspiring). First and foremost, and I assume this was meant as an example, but going by ZIP code would completely eliminate propotional representation in many, many areas. With that would go the purpose of the lower house of Congress, which was intended to provide proportional representation as opposed to the upper house's equal representation. Furthermore, and this does not have anything to do with data, but I would argue Gerrymandering continues to exist as a result of a desire by those in power to remain in power. For more, see George Orwell's 1984 "Power is the ends, not them means." All of this is said with respect and love. Thank you for this.

[@frankdevans1]

+Alex Seelig Thank you for the compliments, particularly on the laughs part--by far the best compliment I've gotten in my mind. I honestly do not think using ZIP codes would eliminate proportional representation. There is no requirement that each district have the same number of ZIP codes, just that ZIP territories are used in full for their open data nature, and that population balance among the final districts is preserved--which is a Constitutional requirement. Plus, ZIP code territories are really small. They have to be small enough that all the mail for that area can be sorted by a few people and delivered on a truck's route each day. Populated states have thousands of them, which easily cut up a state into enough possible territories that districts can be very well proportionally balanced. I completely agree with regard to those in power. But some ideas are just too hard to put back in the box once they are out and people have seen them. I hope this is one of those ideas.

[@dagordon1]

It’s interesting to see what Michigan is doing now, with an Independent Citizens Commission drawing the maps. They couldn’t use an algorithm because Communities of Interest were difficult to program, so it’s all hand drawn maps, in public!

[@frankdevans1]

@AlecSoltes: A state would adopt standards for being able to determine which division is more favorable given any two prospectives. In machine learning, this is called a "cost function". The goal would be to minimize that cost function to search for the best one. Then given a deadline, the most optimal found would be adopted. The standard I proposed was based on geographic distance, but I am sure there are other concepts I haven't thought of and I would love for those to make it out in the open as well.

[@glpt4491]

Do you have a link to the free software you were talking about? I want to submit it to my legislators in North Carolina. We are currently in the process of redrawing our maps and need all the unbiased help we can get.

[@5imian]

I loved everything about this talk!

[@holidayymann1769]

nice story,

[@bobpratt7112]

Just an addendum: The shape of the Illinois 'earmuffs' district is NOT affected by some central geographic feature (like a bay, hills or river). Many recent GOP examples are more egregious, but this IS democratic gerrymandering to lump two different 'hispanic' clusters into a single voting bloc.

[@rungavagairun]

@@bobpratt7112 Yeah, I think Frank Evans (the speaker here) neglects the matter of people having common interests or values having a political voice. He talks about this being objectively fair, and I agree it would be an improvement, but he's solving for population/geographical balance instead of how the people would best be represented in the government.

[@cateybarnish3345]

Brilliant

[@MusikCassette]

9:46 That is just plane and objectivly wrong.

[@ChadGorshing]

Good talk @frankdevans1

[@glpt4491]

Where is the link to the free software he is talking about?

[@wiimooden]

How exactly would this work for state legislatures?

[@niggaflies]

@ Alec Soltes, The same way!

[@noahseethaler1158]

Cool Video

[@MusikCassette]

He thinks, Gerrymandering is a problem of drawing district boundaries. That is the mistake.

[@Apd766]

Using ai to prevent gerrymandering

[@holidayymann1769]

im hungry. u mad bro?

[@nealiesargent7548]

Everyone vote should count and be dicided by proof action truth not lies make up your own mind not by what a dam computer says research facts 💯💯💯 not what fake news says