Suvudu Universe

One thought behind the sort of political manipulation that goes on at the highest levels of both major parties, and what I was thinking about a lot when I wrote my book, is the idea that human beings are statistically predictable. That variables like the policy preferences we feel and the candidates we vote for typically are not the product of some grand individual calculus, or even mere idiosyncratic choices. Rather they're the products of worldview and demographic variables that can accurately be predicted, then even modeled out. Modeled out though statistically, at the aggregate level, not the individual level. Political experts, in academia or political campaigns, won't say that they know that every single rural truck driver who has a gun is conservative, or that every single urban Prius driver who owns solar panel roof-shingles is liberal. But they know that when modeling an election they can reliably predict that, say, 98% of the former are conservative and 98% of the latter are liberal.

A lot of us would assume these same things, partly based on our own tendency towards stereotyping. But have you heard that the Democratic Senatorial Campaign Committee is analyzing the fact that 'in Iowa people with German and Scandinavian surnames are likely to vote slightly differently than those with English and Irish surnames.' Major political operatives aren't just looking at the basic and obvious variables, and not just analyzing such things in a cursory way. They're looking at a lot of variables, ranging from obvious to little understood ones, and looking at them very rigorously. The 2012 Obama Campaign in particular showed a remarkable ability to figure out almost exactly what voters would do using these techniques. For example Dan Wagner, one of the chiefs of analytics for the Obama 2012 Campaign, predicted that Obama would win the key swing area of Hamilton County, Ohio (home of Cincinnati) with 56.4% of the vote. Obama got 56.6% of the Hamilton County vote. For the entire state of Florida, another key battleground, Wagner was 'as close to the mark; Obama's margin was only two tenths of a percent off' what Wagner predicted. This is just one example of the extent to which our collective political behavior seems to be susceptible of a very detailed statistical understanding.

If all of this is true it would mean that, using one technique or another, elections are not some mysterious process whose result can only be known after the fact, but rather a nearly predetermined exercise that can be understood before the election happens. Perhaps years before. Before the election is even scheduled or the candidates for the respective parties chosen to run. As I said in my last post such an idea raises the specter of many disturbing notions. For one, it seems to insult humanity to suggest that our political behavior is so reducible and understandable. It takes a lot of the mystery out of it, and suggests that democratic choice is something sort of moronic and ossified in our mind. One takeaway could be that if indeed our political choices are so predictable, maybe we're not making them very intelligently. Maybe what we're doing is just putting the final resolution on a decision whose calculus we really had finished a while before with a real lack of any new information. Maybe our conclusions are too fixed in time and we all need to be more receptive to allowing them at least the possibility to be more dynamic, to take into account new information or ideas. Or at least examine our old ideas to gauge how much they still seem to have reason and empirical support.

However, there are more menacing implications than the idea that we ought to think a bit more critically about our political choices. That's part of why many, like myself  when I first encountered this set of ideas in a college political science course, are quite perturbed by it. If our collective political behavior can be modeled, ultimately that facilitates it's manipulation. If you want to convince or trick the American people into doing a certain thing that will be much easier if you actually know exactly what they will do or would have done at certain times. But the good news for those who fear such manipulation, or refuse to accept that people are so readily predictable, is that it apparently isn't all quite so simple.

In the first place, as mentioned above, the idea is not that every single person's political behavior is predictable but that in the aggregate political behavior is. This is a key difference between hard sciences like chemistry and physics and social 'sciences' like political science and sociology. When a chemist mixes acids and bases they know what the reaction will be every time, or at least somebody in their field knows. If for some reason the reaction they get varies from what was expected they often either will figure out why that is the case soon or in the near future. Or learn that the reaction didn't really vary, they just had conducted their calculations incorrectly. But when a political scientist encounters a truck-driving, gun-toting rural man they don't know to a certainty that the man is a conservative. They may think they're almost certain that he would be, but in a large sampling of men who fit this description there will, once in a while, be a liberal or a man who otherwise doesn't fit the expected mold. The political expert trying to understand those exceptions likely won't be able to say precisely why those exceptions happened. They might be able to posit some intelligent theories, i.e. 'This rural, gun owning pickup truck driver happens to have a Ph.D. and is a retired professor' or 'might have been raised an 'old line' rural Democrat'. But such subjective guesswork is hardly going to be exact, and in political affairs the time and energy to figure out such exceptions isn't undertaken very often.  After all, could it ever be figured out exactly why all the exceptions to the patterns occur? And what is the need to spend time or money doing so if the major parties and others who control major political resources already know so much about political behavior in the aggregate?

But while political scientists claim to be able to understand predict political behavior in the aggregate, no political prognosticator (not even Nate Silver or Larry Sabato or Gallup) has a 100% track record of doing so. They all have, at the aggregated level, called some elections incorrectly. Even the 2012 Obama Campaign, for all of it's incredible success, apparently had believed it was worth investing a great deal of effort in winning North Carolina, but couldn't win the state on Election Day of 2012. More big picture, both the Romney and Obama campaigns, and many other big budget political efforts have not used election modeling to say that electoral outcomes are unchangeably written in stone. Rather they have used election modeling to see what a sort of natural shape of key elections will be to decide how to allocate their resources in the many places in which they could decide to compete. They choose to do this in some places because they believe that the election results could be bent toward their side, or they know they need to prevent the other side from doing so. With both sides doing this some complexity emerges that makes it hard for any one person or party to know exactly what the future of it all will be. 

Although the parties take into account a lot of modeling and analytics in their decision making, in making their predictions and modeling they still place a great deal of value on polling of random samples of voters, or the aggregation of such polls. The idea of parties predicting political behavior by asking voters what they think is probably, to many of us, far less disturbing than the idea of parties modeling our behavior without asking anyone what they actually think. But it seems that a synthesis of both asking people what they think and modeling it using demography is the most accurate way to see what electoral outcomes will be. The aforementioned predictions by the Obama Campaign that were so accurate at estimating results in Ohio, Florida and many other states, indeed took a great deal of polling into account. Using this synthesis of techniques the Obama campaign was remarkable in how much about the electorate they ended up knowing. Their campaign achieved this level of accuracy by making individual level predictions about every eligible voter in America. Obama's campaigns in 2008 and 2012 used algorithms that trawled for patterns between voter's opinions and 'the data points the campaign had assembled for every voter - as many as one thousand variables each, drawn from voter registration records, consumer data warehouses and past campaign contacts.' 

I don't know if it's entirely wrong to do what the Obama campaign did, but I do find it fascinating. I find it to be the sort of stuff of interesting science fiction (Political Science Fiction!?). The sort of thing that is not obviously and undeniably bad, but obviously and undeniably glib and manipulative. Those who recoil at it should be aware that we're likely to see more of this sort of thing to come. It should make for a lot of interesting artistic fodder, and I hope some of that came across in my book.