The intersection of public policy and public data is fertile ground for data scientists to practice their craft. Two new use cases that have recently emerged in the United States are drafting political districts and predicting political unrest.
After each decennial census, all 50 states must redraw political boundaries to ensure that each congressional district has an equal population, as “practical”. Hundreds of additional state and local districts are also being redrawn as a result of the 2020 census.
Redistricting is fraught with political dangers, as there is potential for the new districts to benefit one party or the other. Accusations of gerrymandering are never far away, especially when one party controls the process. Now, a group of people in Minnesota are taking a new approach to redrawing the lines fairly: let the algorithms do it.
According to the (Minneapolis) Star Tribune, a dozen mathematicians and data scientists embarked on a project that allowed a computer model to redesign the state’s eight congressional districts. The group, which calls itself the Citizen Data Scientists, uses priorities set by a judicial commission as input.
These requirements include avoiding cards that seek to protect, promote, or defeat any incumbent, candidate, or political party, depending on the story. It also means maintaining voting blocs among minority communities, as well as trying to hold together population centers with a common economic, cultural or economic heritage.
“The number of combinations is astronomical,” said Sam Hirsch, an attorney who helped form the group, according to the Star-Tribune story. “Ten years ago the kind of things we were doing were not known and not possible.”
Armed with a large amount of computing power, the group scoured millions of maps before settling on one, which is one of five maps the Judiciary Committee will consider. According to Hirsch, it’s a computer-aided achievement.
“The court sets the priorities, but a computer just does a better job of processing data and experimenting with different combinations,” Hirsch told the Star Tribune.
Political Unrest Tracking
Researchers are also investigating the potential for algorithms to provide early warning of potential political unrest in the United States, such as what occurred on January 6, 2021. This event, which local law enforcement does not have not predicted, could be a precursor to more political unrest. violence in the future.
One such group is CoupCast, a University of Central Florida project that uses machine learning techniques such as gradient boosting and deep neural networks to analyze a variety of societal factors to predict the likelihood of hits. of state and electoral violence in dozens of countries each month.
According to an article from Washington Postthe folks at CoupCast are considering running the same type of analysis for the US, which would be a new use.
“We now have the data – and the opportunity – to follow a very different path than we have followed before,” said the To post quote Clayton Besaw of CoupCast as saying. “It is quite clear from the model that we are heading into a period where we are at greater risk of sustained political violence. The building blocks are there.
This type of sentiment analysis has become quite common overseas. The company that is now OmniSci began as an MIT CSAIL project to use social media posts to track the Arab Spring in 2011. Other groups seeking to protect political violence, such as PeaceTech Lab, also began to focus on the United States.
The latest attempts seek to go beyond social media, which some say is not a reliable indicator of impending unrest. Other types of data, such as income inequality, economic disruption, climate change and levels of social trust, could provide a better forecast of a political storm on the horizon, according to the Post article. .
Next step for AI: running for mayor
Six data science lessons from the epic polling failure
Why winning politics is now tied to big data analytics