Trustworthiness and the importance of graph structure

Abstract

We begin by giving a comprehensive literature review that ties together many fields which have heretofore remained separate. We comment on the approaches from each field and show which algorithms are similar and which are different. Then, starting from a concrete task, we extend traditional trustworthiness algorithms to deal with the more complex situation of multiclass list-valued trustworthiness. In addition, we introduce a learned predictive method based on standard classification algorithms. In the last section, we explore the theory of trustworthiness and begin to make progress towards charting the space of all trustworthiness graphs. We address the commonly underestimated importance of the structure of a trust- worthiness graph, and define a space in which to work as well as defining the solvability of a trustworthiness graph. Finally, we provide recommendations for future work.