Predicting Tie Strength With Social Media – Best Paper CHI’09

Eric Gilbert and Karrie Karahalios from http://social.cs.uiuc.edu/people/gilbert/30

Social media is a category of online media where people are talking, participating, sharing, networking, and bookmarking online such as digg, del.icio.us, facebook, flikr, youtube, and twitter and so on. This paper tried to bridge the gap of the theory in social science, called Tie strengh and the real world and came up with the predictive model.

From one of the classic papers in the social science, Granovetter, M. S. 1973. The Strength of Weak Ties. The American Journal of Sociology, 78(6), 1360-1380.

“The Strength of a tie is a (probably linear) combination of the amount of time, the emotional intensity,  the intimacy (mutual confiding), and the reciprocal services which characterize the tie.”

There are two type of ties, which are Weak Ties (Loose acquaintances) and Strong Ties (Trusted friends or Family).

Weak ties can help a friend

1. generate creative idea
2. find a job

Strong ties (People whose social circles tightly overlap with your own) can affect

1. emotional health
2. often join together to lead organizations thru times of crisis

The Dimensions of Tie Strength

1. Granovetter, 1973: amount of time, intimacy, intensity, and reciprocal services
2. Ronald Burt, 1995: structural factors: network topology and informal social circles
3. Wellman and Wortley, 1990: emotional support: offering advice on family problems
4. Nan Lin, et al., 1981: social distance: socioeconomic status, education level, political affiliation, race and gender

Research Questions

1. The existing literature suggests seven dimensions of tie strength: Intimacy, Duration, Reciprocal Services, Structural, Emotional Support and Social Distance. As manifested in social media, can these dimensions predict tie strength? In what combination?
2. What are the limitations of a tie strength model based solely on social media?

Methodology

• 35 participants
• Task: answer 5 questions of the strength of Facebook friendships (Randomly select friends)
• One 30-minute session

Statistical Methods

The predictive model = predictive variables + dimension interactions + network structure

The model performed quite well, distinguishing between strong and weak ties with over 85% accuracy.

Follow-up Interviews

We had interviews subjects to reveal the relationships we could not predict and they found out certain issues such as:

1. Barely knew friend
2. Asymmetric friendships: Teacher-students
3. Educational difference
4. Confounding the medium
5. Unexpected ways of friendships

Conclusions

• Reveal a specific mechanism by which tie strength manifests itself in social media
• Follow-up interviews suggest profitable lines of future work
• Opportunities to use tie strength to make new conclusions about large-scale social phenomena

Advertisement

Mining User Preference Using Spy Voting for Search Engine Personalization

Wilfred Ng, Lin Deng, and Dik Lun Lee

The Hong Kong University of Science and Technology

ACM Transactions on Internet Technology, Vol. 7, No. 4, Article 19, October 2007

This meta-search is using mining a user’s preferences on the search results from clickthrough data and re-ranking the search results.

Personalization  techniques can be classified into three categories:

1. content-based personalization: deals with the “relevance” measure of Web pages and the user’s queries. The user’s query is modified to adapt the search results for the specific user.
2. link-based personalization:performs personalization based on link analysis techniques.
3. function-based personalization.

This approach falls in the function-based personalization.