4 ms·

I can see the argument for ii. But the ads could just as easily be for rehabilitation clinics. For i, a single data point is not creating all of your targette

I can see the argument for ii. But the ads could just as easily be for rehabilitation clinics. For i, a single data point is not creating all of your targetted ads. And so what if it is, it is still closer to what you might be interested in than completely random ads. iii. Is your own problem for sharing devices and accounts with a 5 year old when you use it for "adult" things. iv. Sure, ads suck, but that metal detector ad is at least close to something you might be interested in. What if you searched for metal detector and got ads for high heeled shoes and candy bars?? How is the metal detector ad not better? You have made some good arguments though, and I think ii. really makes the most sense. I wouldn't be surprised if that very thing is happening.

pdkl95 12y ago
> it is still closer to what you might be interested in than completely random ads. This is projection; "interest" is not the only criteria add are judged by. For instance, I immediately killfile anything animated on a page; something moving next to what I am trying to read often makes the reading impossible. Even though "interesting" ads may be somewhat popular preference, when humans are involved the variance is going to be large. > Is your own problem for sharing devices and accounts with a 5 year old when you use it for "adult" things. You are assuming more than one device even exists. Regardless, in the real world there idea that there is any kind of clean separation between devices and accounts is patently false. The basic idea that it is even possible to predict interest from record of which ads were clicked on or what queries are searched for is very presumptuous, but it the larger problem is "data mining", especially in the long term as the data accumulates. I recommend this[1] very interesting article published couple years ago, about some of the data mining that Target has been doing for at least a decade. http://www.nytimes.com/2012/02/19/magazine/shopping-habits.html?_r=1&pagewanted=all http://www.nytimes.com/2012/02/19/magazine/shopping-habits.h...

pdkl95 12y ago

> it is still closer to what you might be interested in than completely random ads. This is projection; "interest" is not the only criteria add are judged by. For instance, I immediately killfile anything animated on a page; something moving next to what I am trying to read often makes the reading impossible. Even though "interesting" ads may be somewhat popular preference, when humans are involved the variance is going to be large. > Is your own problem for sharing devices and accounts with a 5 year old when you use it for "adult" things. You are assuming more than one device even exists. Regardless, in the real world there idea that there is any kind of clean separation between devices and accounts is patently false. The basic idea that it is even possible to predict interest from record of which ads were clicked on or what queries are searched for is very presumptuous, but it the larger problem is "data mining", especially in the long term as the data accumulates. I recommend this[1] very interesting article published couple years ago, about some of the data mining that Target has been doing for at least a decade. http://www.nytimes.com/2012/02/19/magazine/shopping-habits.html?_r=1&pagewanted=all http://www.nytimes.com/2012/02/19/magazine/shopping-habits.h...