The decreasing marginal value of evaluation network size

  • Authors:
  • Zheng Dong;L. Jean Camp

  • Affiliations:
  • Indiana University, Bloomington, IN;Indiana University, Bloomington, IN

  • Venue:
  • ACM SIGCAS Computers and Society
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

The best way to protect information is never to release it. Yet even the earliest definition of security recognizes availability as a necessary quality. In this work, we seek to quantify the value of information disclosure for web resource evaluation and discovery. Communal evaluation tools help users share ratings on websites, music, and other online resources. This approach assumes that experiences are self-similar, so that a site one person visits is likely to have been evaluated and thus visited by others. Collaborative search tools aim for discovery as opposed to evaluation. Therefore, they assume participants in a collaborative network have large sets of non-overlapping sites so that an increase in network size corresponds to an increase in web coverage. We quantify the value of information sharing for these closely related but sometimes distinct functions. In this paper, we analyzed a dataset that includes eight weeks of browsing history of 1084 college students that live in the same dormitory. Our experiments showed that for discovery, more sharing monotonically improves results; for evaluation, however, there are decreasing marginal returns for each participant added to the network. The subject population was selected for its homogeneity in order to mimic a collaborative network.