Cosley et al 2010

From Cohen Courses
Jump to navigationJump to search

Online Version

An electronic version of this paper can be downloaded here: [1]

Summary

This paper investigates the topic of sequential influence modeling in social networks. The modeling is based on two types of temporal data: 1) ordinal-time observations and 2) snapshot observations. The former is concerned with the fraction of the k-exposed set that becomes adopters before its (k+1)^st neighbor becomes an adopter, and the latter also investigates the evolution of the k-exposed set but within the interval of two consecutive snapshots. One of the major contributions of this paper is providing a method for simulating ordinal time from snapshots, which works by sampling the number of neighbors each node had when it became an adopter according to a hypothesis set, so that it's consistent with the snapshot observations.

Results

The authors applied their analysis to the Wikipedia dataset that consists of the page edit history and user-talk pages (defining the social network) in multiple languages from January 15, 2001 to April 2, 2007. Here are some of their findings on this dataset:

1. For ordinal-time observations, the curve that shows the probability of editing and article on Wikipedia as a function of the number of interactions in the user-talk page has increasing effect for the first five links, and becomes saturate with more subsequent links.

2. For snapshot observations, the same curve steadily increases with more links but with diminishing marginal influence.

Fig1.png

3. A power law is a good approximation for estimating the number of user/community pairs with respect to number of neighbors in community.

Fig2.png

4. The simulation of ordinal time observations from snapshots becomes increasingly accurate with more snapshots, which is not very surprising since in the limit snapshot measurements converge to the ordinal-time measurements.

Related Papers

Study Plan

  • The power law distribution and its unique properties.
  • More details on the sampling approach adopted for simulating ordinal time from snapshots.