Interactional Trust and Trustworthiness in On-Line Social Networks

Select |




Print


Rogers, Glynn


2014-12-12


Report


Sydney, Australia


Trust plays a critical role in traditional social structures and substantial research has been devoted to its emerging role in on-line social networks including multi-agent societies. However trust is a complex, multifaceted concept and care is needed in translating it into the on-line context to avoid confusion and inconsistency. This report begins with a brief examination of how the trust concept is treated in philosophy and sociology focusing on the notion of interpersonal trust which, it becomes clear, is a state of mind and hence not directly observable. However it is argued that the interaction graph of the network enables an objective notion, interactional trust, to be defined which captures the level of interpersonal trust invested in maintaining the network of relationships. This leads to an objective measure of trustworthiness on which trust decisions can be based. Trust decisions involve risk - otherwise there would be no need to trust - and Decision Theory is based on subjective probability. After carefully relating the subjective probability involved in single trust decisions to objective probability, a simple statistical model of the distribution of interactional trustworthiness in an on-line social network is developed. A trust decision process utilising this is then sketched followed by some speculative suggestions on the application to multi-agent societies. The approach employed is quite distinct from previous research based on credentialing or specific reputation mechanisms. Empirical support for the model is provided using data from a location service.


Interactional trust, graph theory, subjective probability, social networks, decision theory, multi-agent societies.


1833-9646-8448


nicta:8448


Rogers, Glynn. Interactional Trust and Trustworthiness in On-Line Social Networks. 2014-12-12. http://hdl.handle.net/102.100.100/93474?index=1



Loading citation data...

Citation counts
(Requires subscription to view)