It's like a stock exchange -- you say you want to sell 30 shares of AAPL, and you're anonymously matched with someone that wants to buy 30 shares of AAPL. Only the exchange knows who both parties to the transaction are, protecting you from evil transaction partners.
Now, if you don't trust the exchange, that's a problem. But if you are worried about using the exchange because you don't trust every possible trading party, that's incorrect.
Only sort of. Let's say I'm doing general advertising but want to know the age and gender of people who click on my ads. So I place ads for "men age 18-20", "men age 21-24", "men age 25-35", ..., "women age 18-20", ... . I give each a different landing url. Then I as the advertiser learn more about my incoming users than I would without targeting. I can't pull all the demographic information the ad network knows because if the buckets are too small things like CTR estimation work poorly, but if I'm high volume I can get good resolution.
Since it's not in Google's or Facebook's best interest to allow a direct connection between their advertiser's and users (if they no longer act as the proxy, where's the money?) I'm not terribly worried about it.
I would like there to be more than just a vested interest in not revealing my identity though. Something legal ensuring they take it seriously would be nice. Like being able to sue them should they mistakenly expose my identity via some third party's clever fingerprinting etc. As long as I stay anonymous and protected (from big brother etc.) though I really don't care. I just think we need more assurances that we are indeed anonymous, gov included, which currently isn't the case.
If you see a random ad for a company on Google, click on it and then supply your email. Then they would know a lot more about than just your email. They would know every demographic filter they used to create that ad. That could be quite specific.
Browser fingerprinting is far more alarming: http://panopticlick.eff.org/
Let's say there's a SASS service you can tie into that has a massive database of browser fingerprint activity. Potentially you could get quite a significant amount of browsing history once enough websites are using the service. You can tie this to an email address etc. when a site requests a signup.
It's a bunch of servers that know lots about you. Google then uses these servers to match ads to you. There is no team of Google employees who knows every detail of your life and personally matches ads to you. It's servers and algorithms all the way down.
A server has as much use for your likes and dislikes as it does for a piece of pizza (which you may or may not like).
I think it's important to realise that the advertisers don't necessarily know about you. Google/Facebook do and they're the ones who provide the 'filtering' service for folks who want to place ads in front of certain groups.
The better they know their users, the more specifically they can target ads to them. For example, if I wanted to reach 20-somethings, who really like pizza and motorbikes, travel widely and currently reside on the west coast, FB can put ads in front of that group (I don't know if Google can, but I assume it's more tricky for them).
Since there are likely a large number of people that use the web while still logged in to FB, then FB could feasibly start offering to place highly-targeted ads out on the web at large. This could obviously have a dramatic effect on Google's dominance.
In other words, it's not just social per se, but being able to learn more about users in order to ensure that they can keep their hold over advertising revenue. Once I looked at it this way, Google's 'need' for social made more sense.
To test the above, we could look at what Google offers in terms of demographics/interest filtering when you buy/place ads around the web. If the ability to refine who sees your ads is really part of the plan then there should be signs that it's happening. I've never used any of Google's ad stuff so I wouldn't know where to look.