Socio-Cognitive Model of Trust: Quantitative Aspects - Trust Theory: A Socio-Cognitive and Computational Model - page 108

3.8 Conclusions
What is to be noticed in this chapter is how one can derive a precise model of the 'degree'
of trust simply from independently postulated beliefs, expectations, evaluations, and their
properties (like the 'certainty' of the belief, or the quantity of the quality or virtue). In general,
in our model of cognition, the pursued goals, the intentions, are based on beliefs, on 'reasons'.
This is also why trust and trust decision can be the object of argumentation and persuasion: I
can provide you with reasons for trusting or not trusting Y ; I can convince you. Of course, trust
can also be the result of mere suggestion, of manipulation, of attraction, and other affective
maneuvers (see Chapter 5); but here we were modeling explicit and arguable trust.
It is also important to notice that the impact of such a trust degree in decision making is not
just due to the 'expected utility'; the process is more complex: there are specific thresholds,
there are differences between high probability and low value versus low probability and high
value.
It is also important not to have simplistic models of trust degree in terms of mere statistics or
reinforcement learning; or of trust decision in terms of delegating to the most trustful guy. An
important additional sophistication we should have introduced - at least for modeling human
trust - would be the asymmetric evaluation of gains (and missed gains) and of losses (and
avoided losses), as explained by 'Prospect Theory': the same amount of money (for example)
does not have a comparable impact on our decision when considered as acquisition and when
considered as loss; and, as for losses, we are risk prone (we prefer uncertain losses to certain
losses), while for winnings we are risk averse (prefer certain winnings to uncertain ones)
(Allais, 1953).
In sum, trust (as attitude and disposition )is graded for seven different reasons:
1. Because it is based on explicit beliefs (like 'evaluations') with their degree of subjective
certainty , recursively due to trust in evidences and sources: on such a basis, X is more or
less sure, convinced that, and so on.
2. Because it is based on implicit, felt 'beliefs': sensations, somatic markers, emotional
activations, with their intensity and affective qualities (safety, worry, etc.); the functional
equivalents of 'beliefs' and explicit evaluations.
3. Because those judgments are about Y 's qualities , virtues, and they can be gradable: Y can
be more or less skilled, or competent, or persistent, etc. In other words, trust is graded
because trustworthiness is graded.
4. Because it is multi-dimensional (and trustworthiness too); and the global judgment or
feeling is the combination of those dimensions.
5. Because it is relative to some Goal of X 's, and goals have a ' value' : they are more or less
important.
6. Because it is a prediction about a future event, and thus about a subjective probability of
such an event.
7. Because it presupposes some risks (both failure, costs, and possible dangers), that might
be perceived with some tangible amount and threshold.
As decision and act , trust can be more or less convinced and sure, but cannot really be graded,
since X has to decide or not, given some threshold of risk acceptance and convenience.