Trump and Big Data Analytics

Trump’s victory was many things. I think of it as a failure of big data analytics on many fronts. My personal theory is that there is more to it than meets the eye, but whatever does meet the eye seems to indicate a massive failure.

I heard on CNN that Hillary Clinton had a 100 million dollar data science operation in her campaign. (But then, I also heard Trump say that Clinton was going to bring in 650 million refugees in the first week of her presidency – so what doI know!) The work of this data science team clearly didn’t yield anything real or useful. They asked her to prepare confetti for a victory party while she got such a walloping that even her opponents had to feel sorry for her. All I can say is that their big data analytics failed, and failed badly.

The pollsters also got it quite wrong, which is not quite a failing of big data analytics, but of statistical modeling. But it can fall under the umbrella of data science, I guess. It turned out that some people lied to the pollster phone calls, or refused to take the survey. The sampling of the population may not have be representative either. All this is very surprising because the US pollsters are supposed to be the most sophisticated in the world, employing smart people with advanced degrees in statistics and mathematics. Could they have gotten it so wrong?

While these two failures are spectacular, I wanted to point out another kind of defect in applying big data analytics to behavioral data. As you know by now, whenever you like or share a post on Facebook or talk about on WhatsApp, they learns a little bit about your political leanings. And when combined with your friends and contacts, they gets= a decent idea of what it is that you are likely to support. This violation of your private thoughts is, of course, unpleasant, but it is a fact of life now. But what does Facebook do with this piece of information? It then feeds you more posts and ads that you are likely to share and like, slowly reinforcing your current world view and isolating you from any competing voices. Personalization of your news feeds inevitably leads to polarization among the population.

Similar trends toward polarization take place as a result of data analytics efforts by search engines as well. You are more likely to see ads and links that reinforce your belief system. I think this phenomenon is perhaps the most insidious failing of big data analytics.

I know it won’t get too far, but I would implore big tech companies like Facebook, Twitter and Google to do the exact opposite. If your data-crunching engines tell you that I hate Trump (which I am sure they do), feed me more links about why his supporters adore him. And feed them some info about his pussy-grabbing techniques.

I wrote about it a while ago. Polarization of this self-reinforcing kind is the real tragedy in geopolitics, response to terrorism, conflicts like Indo-Pakistan or Israeli-Palestine situations, and so on.

About twenty years ago, while waiting for Moroccan fried at an intersection in Marseille, I saw this poster with the title “Ouvrez vos yeux contre l’immigration” below the pair of beautiful blue eyes. I was reading the text on this Front National when my friend pulled up and scolded me for wasting time with this kind of nonsense. I felt that he was wrong. If you didn’t even know what they were complaining about, how would you fight them, much less make peace with them? I think it is time to open our eyes toward one another, and to see, hear and listen.

Comments