I am smack in the middle of an experience I have thought about a lot over the past few years. I am getting the opportunity to work with an impassioned group of entrepreneurs who are trying to design a new social application -- details omitted to protect my liabilities under NDA -- and unlike my usual technology consulting, this is really, really early stage.

We have been talking about various well-known solutions that incorporate social elements -- like friends, groups, collaborative filtering, tagging, and so on -- and stargazing about what the hypothetical users will want and care about (we even flew in a few to get their insights and thoughts). And what I have realized, after the first day, is how hard this is. I mean, I have designed lots of software in the past, and used a lot of different approaches to doing it, but this is somehow more complex: exactly because it is all about the social aspect.

I feel that we don't know enough about social tools to have the necessary design patterns defined to construct the social architecture that will surround all future successful social applications. Based on the events of the past day, I am offering a few -- perhaps obvious and overgeneralized -- observations:

• People Are The Living, Breathing, Beating Heart Of The Universe -- Those folks that I know and form my social reality are the center of my universe. Therefore, activities involving interacting with them, learning about them, and perceiving the world through their eyes should be the centerpoint of social applications. I am strongly biased toward the instant messaging buddy list metaphor as a central motif around which social interaction can swirl, but it's not essential, I guess. The motif of an address book can serve, I suppose, or the network models that underlie social networking apps, but seems to me less helpful than buddy list aggregation into groups. (See Nerdvana posts, for more on this.)
• Artifacts Bind Us Together and Define Us -- People create and leave a trail of their social activities, like creating blog posts, comments, tags, links, ratings, posting pictures, even the path that they have taken through a series of pages on a service. These artifacts are actually a more interesting way to characterize people than simply written stuff in a profile. More importantly, some of these artifacts -- links to people and posts, ratings, testimonials, and so on -- represent the social glue that links us, and is a reflection of emergent value in social networks. When dozens of people link to a book review I post in a hypothetical service, and rate it highly, they are -- in essence -- suggesting that my post matters, that it should be read, that others would find it helpful, and, by extension, that I, as the author, have made a contribution that is valued. This accretion of meaning through the tens of thousands of individual activities within a social application is larger than the 'content' generated: the whole is truly greater than the sum of the parts.
• Social Interaction Is Bottom Up, And So Is Everything Else -- Features like tags, user defined groups, arbitrary user-defined relationships, and user defined categories make sense and work because social context is very personal and local, not universal and general. All fans of what I think of as downtempo music will not use that term, and even if they did, they may still not agree with my characterization of a specific album as downtempo. But when you surrender to the bottom-up nature of social applications, that ceases to matter. Over the long haul, those that care about downtempo-ish music will converge toward a more-or-less consistent use of the term, at least consistent enough for the term to be useful in an approximate sense. Which is really all we can hope for in a subjective universe, anyway.
• Social Stuff Absorbs and Trumps Domain Stuff -- Imagine a social application dedicated to the love of wine (I'd join!). There is a natural information schema in the world of wine, based on things like country of origin, regions, vintners, vintage, kinds of grapes, and so on. That information structures an intuitive schema that we all adopt regarding wine. The same holds in other areas, like music, books, blogs, and nearly anything that people can obsess about. But you shouldn't imagine that this domain schema should the primary axis for people's interaction. That just leads to a giant catalog experience, which basically sucks. The primary axis of user interaction in social applications is human interaction -- people communicating, or individual looking at what others say about their obsessions, or finding new potential friends based on shared obsessions, or finding new wines, books, or music based on the recommendations of friends. The natural schemas of the world should be leveraged in the social sphere, but should be subsumed by it. People will want to live in a coffeeshop, talking to people about books, not in the stacks at the library or the warehouse at Amazon.

Tolstoy wrote in War and Peace Anna Karenina [update and correction thanks to Steve and Alex: see comments], "All happy families resemble one another, each unhappy family is unhappy in its own way." I bet that all successful social applications will be based around the same, shared social architectural patterns, while ten thousand failed solutions will fall by the wayside by adopting some "innovative" take on social architecture that will, in the final analysis, miss the point. Like a social bookstore that forces us to stare, endlessly, at the stacks, and makes it hard to find out what your friends are reading, or to connect with other people that quote Tolstoy in their blogs. Even established realms like Amazon will have to rework their architecture to match the social architecture latent in our wiring, or they will get pushed aside by an upstart that cracks the social code.