A Fuzzy Implementation for the Socio-Cognitive Approach to Trust - Trust Theory: A Socio-Cognitive and Computational Model - page 309

+ 0.4 of ability either means that the trustor is pretty sure that

the trustee is rather good or that he/she is rather sure that the trustee is really excellent ,etc.
In this implementation we do not address how the degree of credibility/certainty of the belief
combines with the degree of the content dimension (even if this analysis is quite relevant); we
just use a single resulting measure.
We address the third question above designing the graph. Some nodes receive input values
from other nodes; these links represent the reasons on which their values are grounded. Direct
edges stand for fuzzy rules or the partial causal flow between the concepts. The sign (

0.4). For example, the value +

)
of an edge stands for causal increase or decrease. For example, the Ability value of a doctor
influences positively (e.g. with weight

+ or −

+

0.6) his Trustfulness: if ability has a positive value,
Trustfulness increases; otherwise it decreases.
We address the fourth question above by assigning values to the edges: they represent the
impact that a concept has over another concept. The various features of the trustee, the various
components of trust evolution do not have the same impact, and importance. Perhaps, for a
specific trustee in a specific context, ability is more important than disposition. We represent
the different quantitative contributions to the global value of trust through these weights on the
edges. The possibility of introducing different impacts for different beliefs surely represents
an improvement with respect to the basic trust model.
FCMs allow causal inference to be quantified in a simple way; they model both the strength of
the concepts and their relevance for the overall analysis. For example, the statement: 'Doctors
are not very accessible and this is an important factor (for determining their trustfulness)
in an emergency situation' is easily modeled as a (strong) positive causal inference between
the two concepts of accessibility and trustfulness. FCMs also allow the influence of different
causal relations to be summed up. For example, adding another statement: 'Doctors are very
good in their ability, but this is a minor factor in an emergency situation' means adding a new
input about ability, with a (weak) positive causal influence over trustfulness. Both accessibility
and ability, each with its strength and its causal power, contribute to establish the value of
trustfulness.

11.9.1 A Note on Fuzzy Values
Normally in fuzzy logic some labels (mainly adjectives) from natural language are used for
assigning values; each label represents a range of possible values. There is not a single universal
translation between adjectives and the exact numerical values in the range.
FCM is different from standard fuzzy techniques, in that it requires the use of crisp input val-
ues; we have used the average of the usual ranges, obtaining the following labels, both for posi-
tive and negative values: quite ; middle ; good; etc. However, as our experiments show, even with
little variation of these values in the same range, the FCMs are stable and give similar results.
As Figure 11.4 shows, the ranges we have used do not divide the whole range

into
equal intervals; in particular, near the center (value zero) the ranges are larger, while near the
two extremities they are smaller. This implies that a little change of a value near the center
normally does not lead to a 'range jump' (e.g. from some to quite ), while the same little change
near the extremities can (e.g. from very to really ).
This topology is modeled in the FCM choosing the threshold function; in fact, it is possible
to choose different kinds of functions, the only constraint is that this choice must be coherent
with the final convergence of the algorithm. With the function chosen in our implementation,

{ −

1,1

}