Information Technology Reference
In-Depth Information
11.1 Using a Fuzzy Approach
Given our purpose of modelling a graded phenomenon like trust (that is difficult to estimate
experimentally) we have chosen a Fuzzy Logic Approach (FLA). A clear advantage with FLA
is the possibility of using natural language labels (like: ' this doctor is very skilled ') to represent
a specific real situation. In this way, it is more direct and simple to use intervals rather than
exact values.
In addition, the behavior of these systems (e.g. their combinatorial properties) seems to
be good at modeling several cognitive dynamics (Dubois and Prade, 1980), even if finding
'the real function' for a mental operation and estimating the contribution of convergent and
divergent belief sources remain ongoing problems.
We have used an implementation based on a special kind of fuzzy system called Fuzzy
Cognitive Maps (FCM); they allow the value of the trustfulness to be computed, starting from
belief sources that refer to trust features. The values of those features are also computed,
allowing us to perform some cognitive operations that lead to the effective decision to trust
or not to trust (e.g. impose an additional threshold on a factor, for example risks). Using this
approach we describe beliefs and trust features as approximate (mental) objects with a strength
and a causal power over one another.
11.2 Scenarios
The scenario we are going to study is medical house assistance and we will look at it in two
particular instances:
a) A doctor (a human operator) visiting a patient at home, and
b) A medical automatic system used to support the patient (without direct human intervention).
The case studies under analysis are:
An emergency situation , in which there is a need to identify what is happening (for example,
a heart attack) as soon as possible, to cope with it; we consider in this case the fact that the
(first) therapy to be applied is quite simple (perhaps just an injection).
A routine situation , in which there is a systematic and specialist therapy which needs to be
applied (using quite a complex procedure) but in which there is no immediate danger to
cope with.
We will show how the following factors can produce the final trust for each possible trustee
who is dependent on it:
The initial strength of the different beliefs (on which trust is based); but also
How much a specific belief impacts on the final trust (the causality power of a belief).
It is through this second kind of factor that we are able to characterize some personality traits
of the agents (Castelfranchi et al. , 1998).
Search WWH ::




Custom Search