Social networks and search
Two years ago I did a four-city speaking tour of New Zealand under the auspices of SmartNet. Before my lunch presentation in Wellington, I sat out on one of the tables, and was astounded to find that the person I was chatting to was an executive of Eurekster, which was at the time a hot new player in applying social networks to search. I’ve never come across much public mention of this, as they present themselves as a US company, but much of Eurekster’s development has been done in New Zealand. The news today is that Microsoft, in endeavoring to integrate social network functionality into its own search offering, will either buy or partner with Eurekster, according to BusinessWeek.
While there are a number of approaches to what is being called “social search”, the heart of it is drawing on the experiences and search results of people with similar interests. Rather than using pure algorithms to rank relevance, it makes a lot of sense to use as inputs what people have found to be useful. This can be done in bounded groups, so for example racing car enthusiasts could form a social group where all the members can draw on the search processes or interesting results others are finding. A search for “fiat” would give very different results than it would in a generic research, or even for a car buyer’s interest group. However I think that forming specific search groups is only a preliminary step down this path. Everyone has many interests and roles, and it is not easy to find and join relevant search groups for each of these areas. In the long-term, collaborative search must automatically draw on people’s most relevant peers and their search results. This relates to how reputation networks will develop, where you have an implicit trust rating for each person’s input into the system. This may be through personally knowing that person, or it may be by how they – or the information they uncover – are viewed by your peers. There is no question that social search will over time give far better results than pure algorithmic search. But what Eurekster, Microsoft, and Yahoo! are doing now in this field are very early steps.
