Search

Service providers face the ever-increasing problem of meeting customer expectations while maximizing profits. This optimal balance is very important for delivering better service quality to users and keeping costs under control through efficient resource allocation. In this paper we suggest optimal strategies for managing system trustworthiness in two different contexts. In the first one the provider has limited information about the users' trustworthiness preferences, which have to be satisfied on every transaction. In the second context, the provider knows what the effect of possible outcomes on customer's trust level and, given that the customer will perform a certain number of transactions, would like to know whether the system trustworthiness should be managed at any point in time in order to meet customer's expectations in a cost-effective way. The optimality of the proposed strategies is demonstrated via both analytical techniques and simulations.

In this article, we describe an approach for computing the current trust level of individual users towards an online system and present initial validation results from a small-scale experiment. This trust computational model relies upon survey research for identifying the set of key trust attributes and grouping users into four segments of expected behaviors. Each user’s initial trust level is computed based on a set of assumptions tailored to the specific segment she belongs to, while the trust level evolution takes additionally into account the system outcomes she has experienced so far. More specifically, the trust update follows a machine learning approach, where during the training phase that consists of a small number of system outcomes, users are asked to report their actual trust levels. Finally, we demonstrate the trustors’ segmentation validity and trust estimation accuracy by performing a small-scale experiment in the context of a fictitious online security service.