Knowledge generalization and its organization around "classes" is a foundational need of human cognition. Using inference, prediction, inheritance allows us to access a lot of knowledge about entities without any direct experience of them. This crucial power is exploited also in social life, to get information about people we never met; it is essential for trust evaluations and relations, to choose a potential partner. Society works also on the basis of trust between strangers: if I know (through signals, marks, declaration ...) the class of a given agent I can have a reliable opinion of its trustworthiness derived from its class membership; then if I trust (or not) that class of individual, as Y belongs to that class, I can trust (or not) Y. In this study we intend to explain and experimentally show the advantage of using categorical recommendations in trust evaluation with respect to recommendations on single agents. In an open world or in a broad population how can we have sufficient direct or reported experience on everybody? This inferential device has to be strongly present in WEB societies supported by MAS as here, according to us, it will have a strong impact and its utility will be even greater than in human societies. In order to show this, we propose two kind of simulation, representing the two way of interaction in human (localized knowledge) and digital societies (non-localized knowledge).
Exploring Categories Recommendations within human and digital societies
Alessandro Sapienza;Cristiano Castelfranchi
2015
Abstract
Knowledge generalization and its organization around "classes" is a foundational need of human cognition. Using inference, prediction, inheritance allows us to access a lot of knowledge about entities without any direct experience of them. This crucial power is exploited also in social life, to get information about people we never met; it is essential for trust evaluations and relations, to choose a potential partner. Society works also on the basis of trust between strangers: if I know (through signals, marks, declaration ...) the class of a given agent I can have a reliable opinion of its trustworthiness derived from its class membership; then if I trust (or not) that class of individual, as Y belongs to that class, I can trust (or not) Y. In this study we intend to explain and experimentally show the advantage of using categorical recommendations in trust evaluation with respect to recommendations on single agents. In an open world or in a broad population how can we have sufficient direct or reported experience on everybody? This inferential device has to be strongly present in WEB societies supported by MAS as here, according to us, it will have a strong impact and its utility will be even greater than in human societies. In order to show this, we propose two kind of simulation, representing the two way of interaction in human (localized knowledge) and digital societies (non-localized knowledge).I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.