Interesting take by ifindkarma:
http://ifindkarma.posterous.com/pandas-and-lobsters-why-google-cannot-build-s

The main thesis is that "with Google applications we return to the app to do something specific and then go on to something else, whereas great social applications are designed to lure us back and make us never want to leave."

Thanks for the link, milivella. However, I don't accept their thesis. It was true ten years ago when Google only did "search", but it doesn't apply to Google Groups, Orkut, Knol, Lively...

The linked article includes this sentence:
"Consider this example: Google Answers focused on answers and failed; Yahoo! Answers focused on social and succeeded."

This is clearly wrong. If you look beyond the answers, the Google Answers service built a community of regular users with a level of social interaction far more engaging and enduring than you find at Yahoo Answers. Many of the Google Answers regulars (I mean readers, not just researchers) are still friends today; how many Yahoo Answers regulars still interact with someone they answered a question for a few years ago?

People have speculated at length about why Google Answers failed. Whatever the reasons were, they were related to Google being unable to successfully scale the service (whereas Yahoo could and did successfully scale theirs). The reasons are unrelated to the merits of the services as social applications.

Google Answers was people (and thus labour) intensive. With Google being a "automate everything" kind of company, it was somewhat predictable that they would eventually drop this service.

Dear Roger, thanks for your comment. Of course the article is too straight in its conclusions. But on the other hand I guess that it's true that Google has not been that successful in its attempts in the social sphere; we just need to find a more sophisticated answer (either this, or I am just plain wrong! :)).

Also see Ionut's take:
http://googlesystem.blogspot.com/2010/07/google-and-value-of-social-networking.html