Archive for the ‘social search’ Category

Growth and decay of social networks

Sunday, December 17th, 2006

The growth of online computer-mediated social networks and the shrinking of real-world social networks nearly coincide in time. Both are examples of social networks, but nevertheless they are quite independent. We can easily import our real-life friendships into online social networks, but the opposite direction—exporting online friendships into real-life friendships—is much harder. “Americans’ circle of confidants has shrunk dramatically in the past two decades and the number of people who say they have no one with whom to discuss important matters has more than doubled, according to a new study by sociologists at The University of Arizona and Duke University” (source).

An interesting question is whether online services are causing decay of real-life social networks or just filling the emptiness caused by extinction of their real-life counterparts due to other reasons. I tend to think in terms of the latter.

Real-life social networks were vital before markets became efficient as people had to form reliable social connections for their survival. In the absense of money and related infrastructure you have to rely on your family and friends. Your wealth is literally determined by the number of friends you have. In countries with weak market economy, like Russia, social networks had no substitute in serving this function until recently. Another important function served by real-life social networks is that they were the main source of news and useful information for people. The advent of mass media weakened this function as well: many people now trust radio and TV more than what their friends say. Finally, many people now turn to Google as a replacement for real-life social search (asking a friend) to solve their real-life problems. This further weakened real-life connections as they don’t serve this function anymore and there is less opportunity to interact with friends.

What is social search?

Thursday, December 7th, 2006

A panel on social search at SES Chicago tried to define social search more precisely yesterday. Chris Sherman suggested the following definition: social search are “wayfinding tools informed by human judgment.” Further discussion of this definition can be found here and here.

I am an evolutionary computation researcher and see the striking resemblance of this new definition to the definition of an interactive genetic algorithm. An interactive genetic algorithm (IGA) is defined as a genetic algorithm informed by human judgement. Genetic algorithm itself is a search procedure inspired by the Darwinian model of evolution. So you can see from here the tight connection between social search as defined above and IGA. However, similarities don’t end here. The situation with defining social search mirrors the earlier one with defining IGA. The definition of IGA was too narrow to encompass other kinds interaction in addition to human judgement. One important part that was missing is the use of human creativity in addition to human judgement or without it. The new class of algorithms therefore was called human-based genetic algorithms (HBGA), though they are even more heavily interactive than IGA. If the definition of IGA did not limit human input to judgement there would be no need to create a new term.

Similar things are now happening with social search. In the case of social search, some examples discussed are not well captured by the proposed definition. For example, Yahoo Answers already uses both human judgement and human creativity and is a typical HBGA. This is why I would like to propose a different definition for social search:

Social search is a search algorithm where some functions are outsourced to humans.

This definition positions social search in abstraction hierarchy between human-based computation and human-based evolutionary computation. On one hand, human-based computation (”algorithms outsourcing some functions to humans”) may be used for other purposes than search. On the other hand, there may be other ways of doing social search than those using evolutionary models.

Google says adieu to Google Answers

Thursday, November 30th, 2006

Google recently announced that it closes its Google Answers service that was created in 2002-2003.

Google Answers is a second attempt of Google at Q&A service where live researchers answer questions on any topic for a fee. I am not sure if the Google Answers service had any novelty when it was launched, but it evolved over time. Initially it was very similar to the first paid online Q&A service Experts exchange launched in 1996. In this service, experts bid to answer questions and the winning expert provided the answer. Later Google Answers added elements from Free Knowledge Exchange service launched earlier by 3form.com (1998-1999). 3form service explores collective intelligence and collaboration among ordinary people to find solutions. There is no fee for posting a question and there is no experts: everyone can provide a possibly partial answer and improve answers provided by others, such that good answers can eventually evolve from partial ones.

In my opinion, late Google Answers was a very interesting experiment combining fee-based and free Q&A services together. To my knowledge it was the first attempt to combine them. However, paid researchers had to compete with free answerers in answering questions. This created unsatisfaction among researchers and they start to publicly criticize the service. I believe this criticism could be one of the reasons for the Google’s decision to close the service. The unsatisfaction of paid researchers in Google combined service might have something in common with what experts feel towards Wikipedia, where they have to compete on equal foot with non-experts.

So far Google attempts to explore social search technologies had only a limited success. In Korea, Google struggles to increase its market share, unable to compete effectively with Naver (NHN) exploring Knowledge-IN service (Korean equivalent of Free Knowledge Exchange). Yahoo adopted such service from Naver a year ago and successfully deployed it worldwide under the name Yahoo! Knowledge or Yahoo! Answers.

Despite the limited success so far, I believe Google has pretty good chances to improve its performance in this area. I expect that Google will continue its experiments with social search and eventually we will see a good replacement for Google Answers.