Title :
Social learning with decentralized choice of private signals
Author :
Chamley, Christophe ; Scaglione, Anna
Author_Institution :
Boston Univ., Boston, MA, USA
Abstract :
Social learning is analyzed when agents choose the type of information in a set of signals to identify the state of a fundamental variable. It is shown that agents may herd on the type of signal that is chosen and herding is socially inefficient. The public reports of agents may have to be restricted to improve the efficiency of social learning. As an example of application, the mechanism is relevant for warnings against a particular threat.
Keywords :
learning (artificial intelligence); multi-agent systems; agents; decentralized choice; private signals; social learning; Biological system modeling; Educational institutions; Entropy; Information rates; Search problems; Standards; Vectors; Social networks; decentralized learning;
Conference_Titel :
Global Conference on Signal and Information Processing (GlobalSIP), 2013 IEEE
Conference_Location :
Austin, TX
DOI :
10.1109/GlobalSIP.2013.6736936
