This is written before the Cambridge Analytica affair.
Facebook and others use algorithms to survey what you do. They do this partly to be able to send you personalised ads. An algorithm is simply a description of a part of what the machine does, which is a calculation – presented as a flow chart in a program, as machine language, as binary calculations. It depends from what angle you see it.
So the thing is, via the Internet it is possible to see what you do, and the “things” which are created to do so can be described. With algorithms, and probably with words as well.
One person cannot follow all the users of Facebook, and also creating an algorithm for each user is too much work, so instead they make an algorithm which creates an algorithm which creates an algorithm, which is actually “your” algorithm, which tells Facebook or someone else exactly what you are doing.
A computer calculation, in a program or a similar thing, may also contain possibilities of variation, depending on your input. This is possible even if the “thinking” is machine thinking, even if the whole thing is built over the principle “if A happens, do B”. If you do certain things the machine will do certain things, it “answers” your actions in different ways, for instance it may store information about what you do, or other things. There are ways into the programs, with particular types of response or activity related to what goes in. Of course it is not alive…but some things it does resemble our ways of acting or thinking.
Random functions also exist, that the program picks a random number which plays a role in the program. This may also expand the number of possible “machine reactions”.
What for instance Facebook is after is, in short, traffic information, what you do on the Internet. Where you click, how much time you spend on the different places, where you come from and where you go. Which words you use for searching. Which browser you use and which version of it.
Cookies do some of the same things, but they “live” inside your computer or are attached to it.
With this information stored for instance by Facebook they can picture how you think in certain matters, who you are – as a customer, is what they are interested in. When this information is “translated” into marketing lingo, we have entered a perhaps better known world of customer psychology and -behaviour, assumptions of what you will buy when etc.
To groupthings is something they do – if your language is Norwegian you belong to a group of roughly 5 million, down from the x million people who use Facebook at the moment. If your operating system’s name is Linux the search is also narrowed, there are fewer Linux users compared to Mac OS or something else.
Facebook’s algorithms are secret, you won’t get to know exactly how they watch you, but these are some simple basics on what they do.