Transactions on Computational Collective Intelligence II

Volume 6450 of the series Lecture Notes in Computer Science pp 46-69

Enhancing Social Search: A Computational Collective Intelligence Model of Behavioural Traits, Trust and Time

  • Luca LongoAffiliated withDepartment of Computer Science and Statistics, Trinity College
  • , Pierpaolo DondioAffiliated withDepartment of Computer Science and Statistics, Trinity College
  • , Stephen BarrettAffiliated withDepartment of Computer Science and Statistics, Trinity College


The Web has been growing in size and with the proliferation of large-scale collaborative computing environments, Social Search has become increasingly important. This recent field focuses on assigning relevance to Web-pages by considering the reader’s perspective rather than Web-masters’ point of view. Current searching technologies of this form tend to rely on explicit human recommendations. In part because it is hard to obtain user feedback, these methods are hard to scale. The challenge is in producing implicit rankings, by reasoning over users’ Web-search activity, without recourse to explicit human intervention. This paper focuses on a novel Social Search model based on Information Foraging Theory, Effort and Computational Trust, showing a different way to implicitly judge Web-entities. The formalism has been divided in two sub-models. The first considers the effort expended by users, in viewing Web-sites, to assess their relevance to a given searching problem. The second enhances the first sub-model by considering only the most trustworthy users’ opinions, identified by Computational Trust techniques.

100 university students explicitly evaluated the usefulness of 12 thematic Web-sites and their browsing activity were gathered. Correlation indexes suggests the existence of a considerable relationship between explicit feedback and implicit computed judgements that were further compared to the ones, provided by the Google search engine. The proposed nature-inspired approach suggests that, by considering the same searching query, Social Search to be more effective than the Google Page-Rank Algorithm. The consideration of the only identified trustworthy students’ implicit feedback provides a way to increase the accuracy of the effort-based approach. This evidence supports the presentation of a schema for a Social Search engine that generates implicit rankings by considering the collective intelligence emerged from users on the Web.