What we've got (hammer... nail...) is a precinct canvass (tells us how, at a level of hundreds of voters in a geographic area, people voted in an election). Can we do similarity matching and find geographic clusters which voted similarly? Yes, yes I think so. Before you scoff, this was tested at a larger level and met with great derision and paranoia before the principles underlying it were subsequently adapted as an organizing principle by a political organization whose turf was covered by the dataset.
So the notion here was: Plug in your precinct. Find other precincts which voted most like your precinct. Go to one of them, knock on some doors and see if you can find some like-minded people to agree / disagree with in a civil and engaging fashion.
I already walked my own precinct and was willing to do it... because I liked meeting people and found it entertaining. I thought this was a problem that people had: they didn't know where to go to meet people who they could actually engage with meaningfully about issues of the day. There's a caveat here: it's a two way street, and you might talk to people who change your mind about something or have some information that's not in your particular "bubble". Turns out that that is emphatically not what organized politics wants to do when they knock on doors.
http://elections.m3047.net/elections/perl/cluster-correlate....
For the record, I did walk some other precincts identified by this tool... just to see who lived there. It was eery, I felt like I could live in any one of those places much more strongly than just some random place in the city; the openness to conversation was similar to knocking on doors in my own precinct and much better outcomes from a standpoint of personal fulfillment.