Right, this is me brainstorming in my lunchhour so it’s necessarily brief. But:

There’s a good article on abstractdynamics.org which basically explains one of my fundamental problems with Friendster-like systems. Namely, that they require you to be far too definite about your relationship to people. At basic, it’s binary: they’re a friend or they’re not a friend. Beyond that, you can say how much they are a friend but rarely how little they are a friend. And, as the article points out, this is all counter-intuitive, when real-life friend networks are far fuzzier.

So I was trying to think how to organise something a bit better, and this is basically what I came up with:

1. Start by dividing friends into “piles”. A pile could be, say, “all my friends from Barbelith, or “all my friends from home”; it needs to be a group that is on roughly equal standing. Essentially, the user should be willing to compare the friends within a single pile with one another.
2. The user now compares the friends in a pile with one another – starting at the top, you just decide whether a person is more/less/the same “amount” (ugh) a friend as one other. As this process is repeated, the pile settles out. Repeat this for every pile.
3. Now, the strength of ties of the piles to the user need to be explained. For instance, you might rank family or really close friends a bit above, say, friends from an email group. Not in particular – in general. A statement like “my family is probably more important to me than my friends from Barbelith”, say, might be the kind of thing we’re looking for.
4. Finally, there ought to be a way of having “satellites” – ie, people you know but would never go as far as to call a friend. They’re not connected to your network; they orbit it. As you become closer, their orbit could tighten until they were probably roughly in line with a pile. And then perhaps they could be merged into a pile.

Now, that’s only a beginning. I’m not sure how to cope with people in multiple piles – to be honest, I could avoid it quite easily as, say, friends I met on the Internet who’ve become really good friends would probably slip into a “RL friends” pile rather than a “digital friends” pile, but I can see this is a problem.

Also: part of my idea for this implementation demands that no-one bar the user ever see the rankings. What should be public is a relative chart of links – ie, you can see who/what is more important to a person, but not by how much. And so we’ve got a relative graphics that judges friends fairly – by comparing like groups with like groups – rather than by comparing all with all. And, crucially, the comparison is of most import to the user making it, not to other people – which is why they can only see a relative map of things. Also, I really like the whole satellite friends thing, because I usually acquire a lot of them.

Anyhow, there you go. I’m sure it’s riddled with flaws but as I said, it’s a half-hour lunchtime brainstorm. Anyone want to take me up on this? Anyone want to comment? Anyone read this damn thing?