First of all, I have to apologize to any data scientists and experimental psychologists that might see this article. I know that what I am about to tell is grossly simplified and reduced. For you guys, this is not going to be a pleasant read. You might even feel nausea at some point. I still think, however, that what I’m going to tell is kind of important, so just bring yourself a bucket and let’s carry on.
So, senator Al Franken asks a valid question:
“How did Facebook, which prides itself on being able to process billions of data points and instantly transform them into personal connections for its users, somehow not make the connection that electoral ads paid for in roubles were coming from Russia? Those are two data points! American political ads and Russian money: roubles. How could you not connect those two dots?”
And the answer is: because computers don’t think this way.
Data mining is when you take a lot of raw data and with a fair amount of calculations retrieve some useful patterns from it.
Let’s say we have data Facebook does. Then, for instance, we could measure the time our users spend on games and an amount of friends they have. Let’s put it on a plane.
Now what computers can do is they can find a pattern in this data. The more games they play, the more friends there have or (sic!) the more friends they have, the more games they play. Computers don’t speculate, they don’t imply causation, and they don’t produce titles for yellow press, they only seek for the patterns.
Computers can strip down all the names, computers usually don’t think of people as persons with names, and build a regression model — a kind of equation that characterizes the pattern there is.
Computers can evaluate this model. The closer the points to the line — the better it is.
But the line may not necessary be straight. The dependency may not be linear. It could be quadratic with a visible minimum.
Or it can be logarithmic.
It can be of any form you can imagine. It all depends on data.
Now, what can be the pattern for these two data points?
Like I just said, it can be of any form you can imagine!
You can’t tell which one it is with so little data, so you can’t really pick a pattern. But, chances are, you just did. Because you are not a computer, you are a human being. And people think differently from computers.
We are, of course, capable of building regression models, but it takes too much computation, too much conscientious effort. We let computers do that kind of work. What we do instead — is we recognize patterns that we already know. And for this kind of activity we don’t need that much of data points, all we need is for them to fit precisely into the pattern we expect.
If anything, adding more points can ruin the clarity.
And that’s why what is clear for Al is not visible for A. I. They just think differently.
I’m not saying that one way of thinking is inferior to other. They are both great, and they are both flawed. If you think that people are always capable of connecting the dots, think again.
As some of you may know, in 2014 Russia invaded Ukrainian east. At first, it was shown as an insurgency, but we all knew where are the “insurgents” came from. I mean, until the August they were actually proud to be Russian. They were literally waving flags and anything. But in August Russian regular army crossed the border and attacked Ukrainian anti-terrorist forces near Ilovaysk, and the long-lasting armed conflict emerged.
Russia may deny its part, but I just happened to work in eastern Ukraine in 2014 and I saw how the whole thing progressed with my own eyes. What still amazes me is the number of people who also lived there and still don’t believe that the “insurgency” was actually the badly masked invasion. They have all the data points they need. They have evidence, they have testimonies, they know the motive, the only thing they lack — is the brain pattern that will allow connecting the dots.
It is unbelievable that in the XXI century one European country would invade another. It just doesn’t fit into the head. And ironically that’s exactly what allows Russia to deny it for so long. The most bizarre thing you do — the easier it is to get away with.
It’s easy to blame A. I. for not seeing things. After all, A. I. is easy. It’s just a machine made by people and programmed by people. But what programs people, what makes them believe they believe, what makes them act like they act,—this is the real mystery.