A Multi-Criteria Decision Making approach based on the Choquet integral for assessing the credibility of User-Generated Content

The Social Web promotes social interactions among people through Web 2.0 technologies. In this context, User-Generated Content (UGC) spreads across social media platforms in the absence of traditional intermediaries that can verify both the believability of the content and the reliability of the sources that generated it. For this reason, the problem of how to assess the credibility of UGC is receiving nowadays increasing attention. In the literature, several approaches have tackled this issue mainly as a classification problem, by categorizing information into genuine and fake. The majority of the proposed solutions follows a data-driven approach, by employing supervised or semi-supervised machine learning techniques that act on multiple features related to credibility. Despite its effectiveness, however, machine learning may present some possible drawbacks, including data-dependency and the possible inscrutability of the contribution that single or interacting features have in the final classification process. In this paper, a Multi-Criteria Decision Making approach is proposed, aimed to assess the credibility of UGC. A given information item (alternative) is evaluated with respect to the considered credibility features (criteria) based on prior domain knowledge, where an overall credibility estimate is obtained by means of a suitable model-driven approach based on aggregation operators. The credibility estimate allows to classify credible UGC with respect to non-credible one, and can also be used to provide a ranking of the alternatives with respect to credibility. To consider interactions among features, the Choquet integral is employed.