Information Technology Reference
In-Depth Information
because the elements which should be combined could produce an incoherent picture: if
someone says that Mary dresses a hat and another one says that she does not dress a hat, I
cannot infer that Mary dresses half a hat; or again if there are two persons that both say that
Dr Smith is not too good a doctor and also not too bad a doctor while two other persons give
us two diverging evaluations on Dr White (one says that he is an excellent doctor and another
says that he is a really bad doctor) we would not have an equivalent evaluation of Dr White
and Dr Smith, and our decision would be guided by other criteria.
These criteria are linked
with context, emotions, personality factors
. We could have people who, in the presence of
diverging opinions, decide to suspend judgment (they become unable to decide), or people
who take into consideration the best opinion (
optimistic personality
), or, on the contrary,
people who take into consideration the worst opinion (
pessimistic personality
). And so on. A
good model should be able to implement different heuristics. For the moment, in our model,
we simply sum up all the contributions and we squash the result with a threshold function. In
fact, the exact heuristics that humans choose depend on the situation and eventually the exact
threshold functions can be the object of empirical analysis or simulations. The model itself is
independent to those heuristics, that is they can be easily substituted.
11.7 Homogeneous and Heterogeneous Sources
We have the problem of summing up the contribution of many different sources. We have
already discussed the case of homogenous sources (e.g. different opinions about a fea-
ture/person/thing, etc.), when an heuristic has to be chosen.
The same problem occurs when we want to sum up the contribution of heterogeneous fonts
(e.g. direct experience and reputation about the ability of a doctor). Even in this case, many
heuristics are possible. For example, which is more relevant, our own personal experience or
the reputation about a specific ability of a person? There is not a definitive answer to this
question: are we able to evaluate that ability in a good way? Or is it better to rely on the
evaluation of others? And vice-versa. Our analysis is limited to a plain estimation of all the
relevant factors, but many other strategies are possible, as in the case of homogenous sources.
Also, in this case, some strategies depend on personality factors.
We have described how it is possible to model belief sources starting from the single beliefs;
now we describe how trust is computed starting from the belief sources.
11.8 Modeling Beliefs and Sources
Following a belief-based model of trust we can distinguish between trust
in the trustee
(be
it either someone, e.g. the doctor, or something, e.g. the automated medical system) which
has to act and produce a given performance thanks to its internal characteristics, and trust
in
the
(positive and/or negative)
environmental conditions
(like opportunities and interferences)
affecting the trustee's performance, which we call 'external factors'. In this work we take into
account:
three main beliefs regarding the trustee:
an ability/competence belief
;a
disposition/
availability belief
, and
an unharmfulness belief;
two main beliefs regarding the contextual factors:
opportunity beliefs
and
danger beliefs
.
Search WWH ::
Custom Search