"Social networks do best when they tap into one of the seven deadly sins. Facebook is ego, Zynga is sloth, LinkedIn is greed." [R. Hoffman]

In our IntelligentAdvice.org group we focus on expanding practical analytical methods based on subjective logic for preference fusion, trust and reputation reasoning and multi-criteria decision-making in online environments. Specific themes of our research are:

  • Multi-criteria and multi-agent preference fusion in decision-making. This theme focuses on developing practical models for decision making in multi-agent scenarios, where several agents express their preferences and the goal to find a socially optimal outcome, and in scenarios where it is necessary to take into account multiple criteria to optimize. To be considered are the combination of hard and soft constraints, which are typical formalisms to represent in a compact way numerical preferences, and the subjective weighting of criteria. Application domains are e.g. in resources management and economics. Another application is to bring reputation systems and recommender systems together, by exploiting multi-criteria decision algorithms, in order to take advantage of the benefits of both and to mitigate their respective drawbacks. Along this line, users’ tastes and reputation of services/goods need to be considered at the same time.
  • Methods for assessing the trustworthiness of online information sources. This theme focuses on computational trust reasoning to assess the trustworthiness, and thereby the reliability of online information sources. To be considered are trust scopes, reputation, trust transitivity, extrinsic and intrinsic trust indicators, as well as fraud and deception detection. Application domains are e.g. the semantic web and online social networks. We can consider a reputation system as a kind of voting system, where the voted candidate corresponds to the entity with the highest reputation score. Some users can manipulate, that is, they can misreport their preferences to make a more preferred candidate the winner of the “election”. Moreover, considering it as a multi-criteria problem, it is interesting to detect the most convenient parameter to influence the winner.
  • Stability of online communities and social networks. Trust and reputation systems are increasingly being used to leverage collaborative moderation of online communities and markets. While it is commonly recognized that online reputation systems play a crucial role for the stability of online communities, there is little research into defining models and criteria for social network and online community stability. This theme focuses on identifying central factors and criteria for stability and how they are related.
  • Recommendation systems. Recommender systems, as an essential component of e-commerce and online service applications, provide users with personalized high-quality recommendations to overcome the well-known information overload problem. Collaborative filtering (CF) is a widely adopted technique to generate recommendations using the ratings of like-minded users. However, CF inherently suffers from two severe issues: data sparsity and cold start, due to the fact that users usually rated only a few resources, especially for the new users. Moreover, these systems are usually vulnerable to shilling attacks, where bogus rating profiles are injected to promote or degrade some resources.
  • Integration of reputation and recommendation systems. Although recommender systems and reputation systems have quite different theoretical and technical bases, both types of systems roughly serve the same purpose which is to provide advice for decision making in e-commerce and online service environments. The purpose of recommender systems is mainly to generate suggestions about resources that a user a priori is not aware of but possibly interest in. The main purpose of reputation systems is to provide advice about resources that the user already is aware of and interest in. The similarity in purpose makes it natural to integrate both types of systems in order to produce better online advice, but their difference in theory and implementation makes the integration challenging.
  Home Site Map