What does google know about you? What does it know about everyone?

[SiliconWizard]

They have no way of filtering out, for example, if you search say 10 different destinations on Google Flights, where you only actually care about one of them.

Yes they do. It's not someone looking over your shoulder looking at what you do. It's all about finding patterns in data and behaviours and those are essentially impossible to fake convincingly. You can't and won't know what they pick up on. People are stupendously naive when it comes to these things.

True. But people are also naive regarding how their data could be useful.
In the example above, why filter out? If you're compulsively looking for flights on Google, Google gets to know you are interested in travelling by plane. That's already pretty useful information about you. Now of course they can get to "know" more based on everything else you look for by finding further patterns. But even just knowing you are looking for flights is useful.

And finding patterns is easier than many think. If all of your searches for travelling are for other means than flights, it may be inferred that you either have a fear of planes, can't afford flights, or never go very far away from your home. Your other searches could quickly help favor one of these options.

Properly "tricking" Google by looking up bogus stuff can be harder than many think as well. It would need some serious discipline, and to pretty much never use Google for things you're really interested in.

And "thanks" to various means of Google tracking (unless you can disable them all), Google can get to know stuff about you even when you're not actively using their services.

[Mr. Scram]

That's easy. Pregnant women stop  buying products they used to use once a month
The more bogus searches that "you" are not really interested in is fed to google the more "your" interests get obscured. It doesn't make a dent to the data google collects from everyone else, but "you" stop being as "well defined" in google's eyes.

Nope. What likely gave it away was the girl buying less heavily scented products. The sense of smell changes when women get pregnant and the products they choose change accordingly. You buy everything you normally buy but just a different brand of say shampoo and deodorant and someone might correctly infer you're pregnant before you know yourself. Again, thinking the inferred conclusions are obvious when you look at the products is naive. That's not how this process works. You take datasets of non pregnant and pregnant women and compare them to unknowns and through the magic of modern statistics you can pick up behavioural changes the women themselves may not even be aware of. You don't even need to know what the actual change is for the data to reveal that you somehow match group X or Y. That's also why intentional noise is very unlikely to change anything as your behaviour is still present.

As many people will have figured out by now these techniques can also be used to spot illnesses or other weaknesses in people using very mundane data. Subtle changes in your diet or interests may be indicators of serious conditions. You don't even know or can't even know your profile is a match but Google does.

[SilverSolder]

They have no way of filtering out, for example, if you search say 10 different destinations on Google Flights, where you only actually care about one of them.

Yes they do. It's not someone looking over your shoulder looking at what you do. It's all about finding patterns in data and behaviours and those are essentially impossible to fake convincingly. You can't and won't know what they pick up on. People are stupendously naive when it comes to these things.

An story already half a decade old is how Target figured out a girl was pregnant by her purchases. It weren't purchases blatantly related to pregnancy but patterns they saw in other pregnant women. Data mining techniques have become a lot more sophisticated since and companies like Google have much more extensive datasets. Third parties may very well know of a serious illness you have before you do.

https://www.forbes.com/sites/kashmirhill/2012/02/16/how-target-figured-out-a-teen-girl-was-pregnant-before-her-father-did/

Data mining doesn't work by magic...  it is, when all is said and done, simply statistics...  and the results depend on the data they have.

So, I don't see how you can argue that adding noise to the data (consistently, over time) is not going to lower the "signal to noise" ratio for the data mining job?

[DimitriP]

Data mining doesn't work by magic...  it is, when all is said and done, simply statistics...  and the results depend on the data they have.

There is a chance google has managed to overcome the old adage "Garbage in, garbage out".
Then again ...they might have not.

[Mr. Scram]

Data mining doesn't work by magic...  it is, when all is said and done, simply statistics...  and the results depend on the data they have.

So, I don't see how you can argue that adding noise to the data (consistently, over time) is not going to lower the "signal to noise" ratio for the data mining job?

Read my previous post as it's explained in some detail. This isn't your basic statistics class. Data science is a quickly evolving field and even scientists are caught out by what data can be derived from seemingly harmless datasets they released to their peers. There's a reason data analysts are in very hot demand and that's because they can distill data from massive and often unstructured sets. People here seem stuck in very outdated paradigms.

https://mashable.com/article/anonymous-data-sets-easily-de-anonymized/

[Mr. Scram]

There is a chance google has managed to overcome the old adage "Garbage in, garbage out".
Then again ...they might have not.

Except that it's not just garbage. As you're a real person your real behaviour is in there.

[SilverSolder]

Data mining doesn't work by magic...  it is, when all is said and done, simply statistics...  and the results depend on the data they have.

So, I don't see how you can argue that adding noise to the data (consistently, over time) is not going to lower the "signal to noise" ratio for the data mining job?

Read my previous post as it's explained in some detail. This isn't your basic statistics class. Data science is a quickly evolving field and even scientists are caught out by what data can be derived from seemingly harmless datasets they released to their peers. There's a reason data analysts are in very hot demand and that's because they can distill data from massive and often unstructured sets. People here seem stuck in very outdated paradigms.

https://mashable.com/article/anonymous-data-sets-easily-de-anonymized/

I have re-read your posts, and you haven't actually explained anything in detail at all -  you are saying (repeatedly) that it is possible to extract information from patterns in the data - subtle patterns, even - which nobody is disagreeing with.   Furthermore, nobody is disagreeing that you can pick up useful information from unexpected "side channels" that most people are not really consciously aware of.

But this kind of analysis is not immune to noise.  The cleaner and more trustworthy the input data for the analysis is, the more accurate the results will be.  Conversely, the more noise you add (including into the "side channels"), the harder the job gets.  This is just common sense and I don't understand why the notion is contentious at all.

[MagicSmoker]

....People here seem stuck in very outdated paradigms.

I was going to tell you to dial back on the condescension, but upon reading some of your post history I suspect that's a wasted cause.

After all, most of here are engineers, not statisticians, and shouldn't be expected to be capable of doing Google-level data manipulation.

I don't care how Google, Amazon, etc., figure out what I want before I even know I want it; I don't want them figuring it out in the first place.

[Mr. Scram]

I have re-read your posts, and you haven't actually explained anything in detail at all -  you are saying (repeatedly) that it is possible to extract information from patterns in the data - subtle patterns, even - which nobody is disagreeing with.   Furthermore, nobody is disagreeing that you can pick up useful information from unexpected "side channels" that most people are not really consciously aware of.

But this kind of analysis is not immune to noise.  The cleaner and more trustworthy the input data for the analysis is, the more accurate the results will be.  Conversely, the more noise you add (including into the "side channels"), the harder the job gets.  This is just common sense and I don't understand why the notion is contentious at all.

Your real behaviour is still in there. Everything other than what they're comparing you to was filtered out anyway so adding a tiny bit on top makes no difference. Of course your "noise" is likely blatantly obvious.

[Mr. Scram]

I was going to tell you to dial back on the condescension, but upon reading some of your post history I suspect that's a wasted cause.

After all, most of here are engineers, not statisticians, and shouldn't be expected to be capable of doing Google-level data manipulation.

I don't care how Google, Amazon, etc., figure out what I want before I even know I want it; I don't want them figuring it out in the first place.

I don't think many people are comfortable with being tracked and profiled in great detail but it's how the world works right know. Knowing how data is mined is the only chance you have preventing it from happening. Please refrain from personal insults. 

[DimitriP]

Except that it's not just garbage. As you're a real person your real behaviour is in there.

I'm sure my social security number can be composed by page numbers of "War and Peace" too. But I'm not worried

[SilverSolder]

I have re-read your posts, and you haven't actually explained anything in detail at all -  you are saying (repeatedly) that it is possible to extract information from patterns in the data - subtle patterns, even - which nobody is disagreeing with.   Furthermore, nobody is disagreeing that you can pick up useful information from unexpected "side channels" that most people are not really consciously aware of.

But this kind of analysis is not immune to noise.  The cleaner and more trustworthy the input data for the analysis is, the more accurate the results will be.  Conversely, the more noise you add (including into the "side channels"), the harder the job gets.  This is just common sense and I don't understand why the notion is contentious at all.

Your real behaviour is still in there. Everything other than what they're comparing you to was filtered out anyway so adding a tiny bit on top makes no difference. Of course your "noise" is likely blatantly obvious.

Well, you real behaviour - as in your actual purchasing transactions - is something that you cannot really hide.  This data has been used for advertising purposes for years. 

E.g. see:  https://www.businessinsider.com/credit-cards-sell-purchase-data-to-advertisers-2013-4

But what we are talking about here is ads that follow you around depending on what you have recently searched for - at least, that's what I thought we were talking about! :-)

I have noticed that searching for stuff while logged in on eBay means Google ads will follow you around from then on.  So eBay is obviously passing data on to Google, based on your account ID.   The ads are completely based on what you searched for on eBay.  There isn't much intelligence involved, it seems.

Data is also passed from Google to Facebook, and vice versa.  I have proven that by using 100% isolated browsers, where a search on eBay in browser A results in being stalked on Facebook in browser B (on a different computer).

Everyone that has data, is in the business of selling it...  it seems.   Surveillance capitalism, gotta love it.

What's new with Google/Facebook/eBay etc. is that they are tracking people at the individual level, not anonymizing anything, it would seem.

Of course, we live in a free country, and you can simply choose not to use any modern Internet services, if you don't like it...  right?

[eugenenine]

  I have a gmail account on my cellphone but apparently it is required by Android.  I

You can run Android without a gmail, its not required.  Some phones will have an extra setup that asks for it but for Android itsself you van get away without it.
I ran my old table for about a year with no google by using f-droid to install apps, and using my own e-mail and file.calendar service (Owncloud).

[eugenenine]

If using Chrome or the Microsoft fork in Edge, there are hard coded IP links to Google servers, used for both tracking, and for analytics, along for detecting DNS poisoning and redirects.

If your stuck on windows and chrome then use Chromium which is the same source as chrome without any google.

[Beamin]

Rather than blocking Google there is another approach to use if your Internet connection is not charged by the Mb. Some browsers have plugins that send a continuous stream of enquiries to Google using your account, one moment it may be asking the melting point of asphalt and the next enquiry could be the population of Manchester NH.

It's a better approach, poison the well, submerge them in so much information that they can't tell good from bad. The moment the servers think that you're a thirty year old housewife from Phoenix or a twenty something from Paris you will know that you've won.

I used to live in Manchester NH. Its NH's largest city with two tall buildings... well 2 1/2 if you count the one with the microwave antennas on it. Why did you pick that city? No one has ever heard or been to manchvegas as the locals call it. My ex girlfriend Mel, was actually "Miss Manchvegas" one year which is like a state wide beauty pageant. God she was dumb and every time I went out with her guys would hit on her, and the best part was when drunk guys would try and fight me because I was short. Yes that  will win her love. So glad I don't live in that state anymore. 

[NiHaoMike]

Well, you real behaviour - as in your actual purchasing transactions - is something that you cannot really hide.  This data has been used for advertising purposes for years. 

E.g. see:  https://www.businessinsider.com/credit-cards-sell-purchase-data-to-advertisers-2013-4

There is a way out of that - buy stuff with cash or cryptocurrency.

But what we are talking about here is ads that follow you around depending on what you have recently searched for - at least, that's what I thought we were talking about! :-)

I have noticed that searching for stuff while logged in on eBay means Google ads will follow you around from then on.  So eBay is obviously passing data on to Google, based on your account ID.   The ads are completely based on what you searched for on eBay.  There isn't much intelligence involved, it seems.

Data is also passed from Google to Facebook, and vice versa.  I have proven that by using 100% isolated browsers, where a search on eBay in browser A results in being stalked on Facebook in browser B (on a different computer).

Everyone that has data, is in the business of selling it...  it seems.   Surveillance capitalism, gotta love it.

What's new with Google/Facebook/eBay etc. is that they are tracking people at the individual level, not anonymizing anything, it would seem.

Of course, we live in a free country, and you can simply choose not to use any modern Internet services, if you don't like it...  right?

Simply use adblock to break the last link. Now their effort is wasted. It would be harder to sell data if there's less profit in doing so.

[Halcyon]

I deliberately feed Google (and other sites) false information, that even if they could join the dots between them all, the information is often fabricated or contradicts one another. One moment Google thinks I'm in Melbourne, the next it puts my location as in a totally different state.

I like to keep the algorithms guessing ;-)