google reader sharing: conjecture or inference.
Posted by detour1999 | Filed under google, nerdly, programming, reader, social
I’ve been thinking a lot about google reader and the sharing fiasco over the past couple of days.
It seems to me that the new sharing feature doesn’t really work for anyone. For some, (who communicate primarily through their @gmail.com account) it was way too easy to share with their contacts [NB: I said "Contacts" not "Friends", that's a rather important semantic distinction.] For others (who don’t really use their @gmail.com address as their main email address) it became a series of emailing actual friends for their @gmail.com account. Personally, I had already bothered my friends once (a while back) for their shared items address… I didn’t need to bother them again for their gmail address.
In theory, better integration of friend’s shared items is an awesome idea. I was (and still am to an extent) incredibly excited about it. The problem is in the definition of friend.
The criteria for “Friend”ship in google reader is the ability to chat user@gmail.com to user@gmail.com. That ability comes about either by explicitly choosing to allow chat between users; or by exchanging emails. That is really the core problem with google’s implementation. Conjecture.
I’ve been wrestling with the idea for a while. In my mind, google’s implementation is based on flawed logic, but that flawed logic seemed to stem directly from trying too hard to apply DRY principles.
Say I had a dataset representing a person. I wouldn’t want to store both birthdate AND age. Age can conclusively be inferred by birthdate.
Say I had a dataset representing social interaction. I wouldn’t want to store social connectivity if it can be inferred from existing data, right?
In this case, no.
Given the age example, age can ALWAYS be calculated given a birthdate. Friendship can NOT be calculated based on email exchange. For example, 60% of my “friends”, according to google reader, are people I have bought used volkswagen parts from. Exchanging emails about buying used headlights does not infer friendship. Interaction, yes. Friendship, no.
Inference is good practice. Conjecture is not.
If you can’t get reproducible results ALWAYS (or at the very least, most of the time) it’s not inference… it’s conjecture.