Boards Not Trees

To increase the value of social networks, interactions should be based not on individuals but on relevant fractions of individuals.

For example, when I see an interesting post on 3D printing from someone on Twitter and follow them, I may then see 1000 posts about political candidates, cats or other things that I don’t care about from the person I just followed.

All current social networks function in this way, filtering at the level of granularity of the individual contributor. The result is harmful on multiple levels, the first being an overall increase in ongoing filtering effort (the need to assess and reject content of those you follow manually). The second is unnecessary alienation due to personal-level incompatibilities that become apparent unnecessarily due to forced out-of-context information sharing.

Instead, a better design would create associations between publishers and subscribers which are sensitive to the context of the establishment of the relationship. Instead of “following” or “friending” a person, the act of selecting an individual post for quality establishes an interest in a subset or context of the publisher.

I call this “boards not trees”; consider the case of a carpenter building a cabinet. The carpenter could (in theory) build the cabinet by acquiring several whole trees, but instead it makes more sense for them to acquire only a subset of the tree in the form of a board that can be used directly, or perhaps broken-down with little waste into the parts that make up the cabinet.

In current social networks we take the opposite approach, and establish every relationship directly with other members of the network. Google+ comes close to getting this right, and to some degree this goal can be accomplished manually but to be truly useful the context establishment needs to be automated.

Returning to my original reference to a post about 3D printing on Twitter, it’s not hard to imagine a primitive system that would be capable of identifying the context of my interest by examining keywords (or better yet, hashtags) in the selected post and then use this context to present future posts that are likely to be interesting to me from this user (and nothing more). Taking a (small) step further, future feedback on these suggestions could provide positive or negative reenforcement and a basic neural network would be capable of automatically keeping the feed in line with my interests, even as they change, providing quality content with minimal manual filtering.

Moreover, this same system could be used to provide feedback to the author to steer them toward topics of interest (perhaps even suggest topics when input is solicited) and by analyzing the response to this input, establish better overall value rankings compared to the simplistic systems currently in place.

I’ve been noodling on what the implementation of a system of this type would look like. If you’d be interested in participating in the development and testing of such a system, leave a comment below.


evernote

491 Words

2012-10-29 00:00 +0000