Matador
Well-known member
I attended a fascinating lecture of a group of computer scientists who were using machine learning to inform redistricting efforts and to help draw alternate congressional districts which are the subject of intense gerrymandering scrutiny. The supreme court basically said it wasn't able to rule without a definitive method to understand which districts were 'fair', and this group is trying to discover what that looks like. The work is based on this original research:
http://www.ncsa.illinois.edu/news/story/blue_waters_supercomputer_used_to_develop_a_standard_for_partisan_gerrymand
Which used distributed computing to generate millions of potential congressional maps.
A side issue was related to the electoral college and proportionality vote tallying methods: essentially altering how states apportion votes based on individual and state-wide results. Interestingly enough, a state-level proportional vote would have placed Clinton and Trump nearly neck and neck, which is much closer as the national popular vote predicts (and Clinton ended up winning).
They make the interesting point that for the presidency, you can swap the electorate into different geographical places and get wildly different results, which doesn't make much sense from the point of view that the presidency is a national Federal election, and (in theory) the President is supposed to represent the interests of the entire country (again, in theory). As a case in point, you could have all Republican votes trade living arrangements with all Democratic voters, and with the exact same vote tallies get the opposite result (Clinton winning 308 electoral votes), which again isn't sensical for a national election.
They also reminded me of the fact that John Kerry came very close to beating George Bush despite Bush having nearly 3 million more votes nationally. If John Kerry has won about 50,000 more votes in Ohio, he would have won the EC despite losing the popular vote. I'm wondering if the rhetoric might have been different had that happened.
http://www.ncsa.illinois.edu/news/story/blue_waters_supercomputer_used_to_develop_a_standard_for_partisan_gerrymand
Which used distributed computing to generate millions of potential congressional maps.
A side issue was related to the electoral college and proportionality vote tallying methods: essentially altering how states apportion votes based on individual and state-wide results. Interestingly enough, a state-level proportional vote would have placed Clinton and Trump nearly neck and neck, which is much closer as the national popular vote predicts (and Clinton ended up winning).
They make the interesting point that for the presidency, you can swap the electorate into different geographical places and get wildly different results, which doesn't make much sense from the point of view that the presidency is a national Federal election, and (in theory) the President is supposed to represent the interests of the entire country (again, in theory). As a case in point, you could have all Republican votes trade living arrangements with all Democratic voters, and with the exact same vote tallies get the opposite result (Clinton winning 308 electoral votes), which again isn't sensical for a national election.
They also reminded me of the fact that John Kerry came very close to beating George Bush despite Bush having nearly 3 million more votes nationally. If John Kerry has won about 50,000 more votes in Ohio, he would have won the EC despite losing the popular vote. I'm wondering if the rhetoric might have been different had that happened.