Computational Trust to Insure the Quality of Assistance Given to Human Actors

Abstract

This paper presents an Interactive Assistance System (IAS) that we have defined to help human worker reach the “effectiveness of a machine” while keeping his qualities as a human being: flexibility, adaptability and “humanity”. Why is such a system interesting? From our point of view, because there are processes for which the collaboration of human actors is essential to reach the optimal quality of functioning. In this paper, we give two examples of such unit: one in an industrial environment and the other in a medical environment. In this context, we analyze the dangers of the bad use of such an assistance system. We apply (computed) trust measures to prevent the dysfunction of the IAS. The trust computing engine allows us to evaluate the chances that an action will be correctly executed by the human operator. Thanks to the computation of trust values, IAS can dynamically prevent dysfunction by setting its limits.