Studies on trust in information sources have mostly focused on whether the source is capable of providing correct and complete information, thus overlooking another essential aspect of trust: the assessment of relevance. Information, even when true, is literally useless, unless it relates meaningfully to the informational needs and practical goals of its recipient. Moreover, relevance, exactly like truth, is not a self-evident feature of information: agents frequently realize whether or not an information was relevant for their purposes only ex post, and this is precisely why relevance, like truth, requires trust. The agent needs to be able to (i) rely on the relevance of the information provided by a source before or without being able to directly verify such relevance, and thus (ii) estimate the quality of sources also based on their ability to consistently deliver relevant (as well as correct and complete) information. In this paper we outline some desiderata for modeling relevance as one of the key features in deciding whether to trust an information source, we analyze its related role in determining belief formation and change, we detail how to assess relevance in order to avoid biases (e.g., giving systematic priority to good news over bad ones), we discuss whether relevance is a subjective or an objective feature of information (and in what sense), and we conclude by suggesting possible ways of implementing and/or formalizing trust in relevance for MAS, based on previous work on trust dynamics.
Trust in relevance
Contributo in atti di convegno
M. Jeusfeld c/o Redaktion Sun SITE, Informatik V, RWTH Aachen., Aachen, Germania
CEUR-WS.org, Aachen, DEU
First International Conference on Agreement Technologies. Proceedings, pp. 332–346, Dubrovnik, 15-16 October 2012