5 ms·
It's all about understanding behavior. Humans, all humans, anywhere in the world are creatures of habit and routine ... Once you understand that you can start t
by gdsdfe 6y ago
It's all about understanding behavior. Humans, all humans, anywhere in the world are creatures of habit and routine ... Once you understand that you can start to exploit it.
And it doesn't matter if you're doing it now, the data they're collecting is there forever, as ML algorithms are getting better your understanding of the data is getting better.
Let's take a concrete example : the cambridge analytica case ... Let's say for the sake of argument I've determined that blue collar workers between the age of 30 and 40 that are in major cities have a 40 to 60 percent chance of voting Republican now if take all of those that have been to a church once or have a close relative that have and bombard them with ads showing how the democrat candidate is supporting abortion for example (wether that's true or not) now the Republican vote have a much better odds. Of course I'm oversimplifying but this is not science fiction this was done with a very good efficiency in different countries and it's just one example ...
- treis 6y agoYour concrete example is nothing more than voters getting the information most pertinent to them. That's not a bad thing.
- gdsdfe 6y agoyou're kinda proving my point ... but let me ask you this : are you ok with mass manipulation and profiling ?
- IX-103 6y agoUmm that's only some of the information that pertinent. It might also be pertinent that the other candidate cheated on his wife with a male prostitute and used campaign find to pay for an aide's abortion, but I'm pretty sure they wouldn't put that in the ad as well. Incomplete or biased information can be worse than no information...