the social horizon

Does the inherent limit on human interaction group size apply to online social networks?. That limit is called “Dunbar’s Number” and is estimated to be ~150, based on observations of social networks among primates and then extrapolating to humans taking increased brainpower into consideration. An intriguing piece in the WSJ asks whether online social networks are still bound by Dunbar’s number or whether technological innovation might permit us to exceed it:

But there is reason to believe that the social-networking sites will enable their users to burst past Dunbar’s number for friends, just as humans have developed and harnessed technology to surpass their physical limits on speed, strength and the ability to process information.
Robin Dunbar, an Oxford anthropologist whose 1993 research gave rise to the magical count of 150, doesn’t use social-networking sites himself. But he says they could “in principle” allow users to push past the limit. “It’s perfectly possible that the technology will increase your memory capacity,” he says.

The question is whether those who keep ties to hundreds of people do so to the detriment of their closest relationships — defined by Prof. Dunbar as those formed with people you turn to when in severe distress.

The problem here is the definition of the word “relationship”. Dunbar’s definition of “closest” is just one of many possible ones, and the various definitions might well overlap. But does that mean that business relationships are excluded from Dunbar’s limit? If so, then you might expect to see many more contacts on LinkedIn, which caters to a business networking model, than on Facebook which is primarily stalker heaven. LinkedIn is approaching critical mass in terms of network effect; RWW found over 80% of their business contacts already using it, for example.

There are surely other models one could employ to map relationships: blogrolls, chat client lists, twitter fans/friends, etc. I think any one of these – or a weighted combination of all of them – would be good data sets to see whether Dunbar’s number truly holds online or not.

The reason why it is important to consider is because if it does hold (or if indeed there is any limit at all) then that substantially undermines the argument that the social graph is a construct of unlimited utility for search personalization or the semantic web. If anything, the social graph could well become an obstacle to finding information rather than an asset. Everyone keeps talking about search “personalization” but that’s a synonym for search filtering; filtering is a lossy process, you are discarding data. Optimal search wouldn’t define the best result as the most “personal” but rather the most “relevant” – and often that ight well be data lying far eyond the cozy confines of your social graph. In fact, assuming that you are searching for something you don’t know, it’s more likely to be outside than inside.

Human nature eing what it is, people might not even realize that their newly personalized search results are less relevant!

social search skepticism

Last summer, there was a dust-up between several high-profile “web 2.0” personalities that made for interesting reading. It started with Robert Scoble, who created a three-part video essay provacatively titled “Why Mahalo, TechMeme, and Facebook are going to kick Google’s butt in four years“. Scoble is obsessed with the idea that search engine optimization (SEO) is poisoning the well of search and that adding the “social” element will magically improve relevance. Dave Winer had a fairly succinct rebuttal, Danny Sullivan took issue with Scoble’s explicit equating of SEO with spam, and Rand Fishkin steps through Scoble’s arguments and fact-checks it to oblivion.

Overall, I came away from the fracas convinced that social networking is not some magic bullet and the problems of “how do I find information” and “who do I want to interact with” to be wholly separate ones. I like a walled garden for my identity-driven personal and professional interactions, but I also want to wander in the wild when need be. It’s the same reason I am skeptical of “personalized search” services like Google’s own “Web History” initiative. It’s not the privacy issues that worry me, but rather the imposed limitation on what my search results are based on what my search results were in the past. Why should I assume that for a given search, the most relevant results will necessarily be related to the searches I previously made? Presumably I search for something based on a need for new information I do not currently possess.

The same argument applies to “social search” initiatives like Delver, recently praised by RWW as “more personal and meaningful to users than a generic search using ‘normal’ search engine.” Why is a filter derived from my social graph any guarantee of more relevance to my query?

If anything, the onus on the user is to craft a better query; to that end Google offers an advanced set of search operators that provide tremendous power and flexibility. Overall, every search is unique, and no amount of personalization or social networking is going to change that fact. If anything, the right approach is to allow a search to stimulate new searches; ie ask new questions rather than spoonfeed me old answers.