Section: New Results

Computational Trust based on User Behavior

Participants : Quang Vinh Dang, Claudia-Lavinia Ignat.

We continued our investigation on computing a trust score for each user according to their behaviour during a collaborative task. Previously we proposed a contract-based collaboration model [31] where trust in users is established and adjusted based on their compliance to the contracts specified by the data owners when they share the data.

We continued this work by proposing an experimental design for testing the proposed trust-based collaboration model. We studied the trust game, a money exchange game that has been widely used in behavioural economics for studying trust and collaboration between humans. In this game, exchange of money is entirely attributable to the existence of trust between users. In the context of the trust game we proposed a trust metric that reflects user behaviours during the collaboration [10]. This metric is robust against fluctuating user behaviour. Our trust metric is the first one that was proposed in the context of the trust game in order to predict user behaviour.

In order to compute the trust score of users according to their contributions during a collaborative editing task, we need to evaluate the quality of the document content. As an initial work in this direction we investigated how to automatically assess the quality of Wikipedia articles in order to guide readers towards high quality articles and to suggest to authors which articles need to be improved. In this context we proposed two automatic assessment methods of the quality of Wikipedia articles. In the first approach we introduced readability features for a better prediction of quality [11]. The second approach is based on a deep-learning mechanism that automatically learns features from document contents rather than manually defining them [13], [4].