Skip to main content

Table 1 Overview of properties of trust described in literature [ 14 , 41 - 46 ]

From: Reusable components for online reputation systems

Dynamic Trust can increase or decrease through gathering new experiences. Moreover, trust is said to decay with time (time-based aging [45]). Because of these characteristics, trust values strongly depend on the time they are determined. The greater importance of new experiences compared to old experiences has been widely studied and considered in many trust models such as [32,47] or [30].
Context-dependent Trust is bound to a specific context. For example, Alice trusts Bob as her doctor. However, she might not trust him as a cook to prepare a delicious meal for her.
Multi-faceted Even in the same context, a trust value may not reflect all aspects of this context [43]. For example, a customer may trust a particular restaurant for its quality of food but not for its quality of service. The overall trust on this restaurant depends on the combination of the amount of trust in the specific aspects.
Propagative One property of trust made use of in several models is its propagativity. If Alice trusts Bob, who in turn trusts Claire, Alice can derive trust on Claire from the relationships between her and Bob as well as between Bob and Claire. Because of this propagative nature, it is possible to create trust chains passing trust from one agent to another agent. As clarified by Christianson and Harbison [48], trust is not automatically transitive although trust transitivity was assumed proven for a long time. If Alice trusts Bob, who in turn trusts Claire, it does not inherently mean that Alice trusts Claire. It follows from the foregoing that transitivity implies propagation. The reverse, though, is not the case.
Composable When trust is propagated, a particular agent may be connected to multiple trust chains. To come up with a final decision whether to trust or distrust this agent, the trust information received from the different chains need to be composed in order to build one aggregated picture. In this context, trust statements propagated from nodes close to oneself should have greater influence on the aggregated value than the ones from distant nodes (distance-based aging [45]). Composition is potentially difficult if the trust statements are contradictory [14].
Subjective The subjective nature of trust becomes clear if one thinks about a review on Amazon [26]. A book review that totally reflects Alice’s opinion will probably resolve in a high level of trust against the reviewer Rachel. Bob, however, who disagrees with the review, will have a lower trust in Rachel although it bases on the same evidence.
Fine-grained Although trust is sometimes modeled in a binary manner (i.e. either trust or distrust), it is possible that Alice trusts both Bob and Claire but that she trusts Bob more than Claire. Hence, there may be multiple discrete levels of trust such as high, medium and low [41]. Mapped to numbers, trust may also be a continuous variable taking values within a certain interval (e.g. between 0 and 1).
Event-sensitive It can take a long time to build trust. One negative experience, though, can destroy it [23].
Reflexive Trust in oneself is always at the maximum value.
Self-reinforcing It is human nature to preferentially interact with other agents that are trusted. Analogously, agents will avoid interacting with untrustworthy agents. Thus, the trustworthiness of other agents is inherently taken into consideration.