The Glue of Society - Trust Theory: A Socio-Cognitive and Computational Model - page 265

his disappointments and negative expectations to the whole class; thus, also X 's trust in Z
collapses. And so on.
9.4.2 Trans-Local Repercussions
Also non-personal and non-local changes of the trust relationships can affect X 's views and
the entire network. Y can, for example, just observe J's delegation to K and on such a basis
change her own trust disposition (and decision) towards Z . In fact, there might be analogies
between J 's delegation and X 's potential delegation, or between K and Z .Or X can just imitate
J , use his example as a model or cue, not only in a pseudo-transitive way ( X 's attitude towards
K is derived from J 's attitude), but in an analogical way: if J trusts K , X can trust Z (because
X and J have similar needs, and K and Z similar properties) (see again Section 6.6).
Notice that the observation of the others' behavior and their evaluation (in relation to social
conventions and norms, to fairness, honesty, etc.) is a basic fundamental social disposition and
ability in human beings. We do not just observe the behaviors of agents that directly concern
us (exchanging or cooperating with us, competing with us, etc.); we observe agents interacting
with other agents not related with us. And we in a sense provide an altruistic 'surveillance'
of each other (Gintis, 1957); we evaluate them, we spread around our blame or admiration,
we spread the circulating reputation. This is the most powerful instrument for social control
and social order in human societies, but, it will be equally important in virtual and in artificial
societies. This is exactly why 'identification' of the agents matters in the first place.
Other trans-local mechanisms for trust repercussion apart from observation of distal events,
are referrals (other agents report to me their evaluation or the success/failure of their delegation
(Yu and Singh, 2003), and reputation: spreading around opinions and gossip about agents
(Conte and Paolucci, 2002).
Particularly delicate conditions for web effects are default and generalized forms of trust .
Some of them could collapse in an impressive way (like our trust in money). If, for example,
the rule of my generalized trust is:
'Since and until everybody trusts everybody (or everybody is not diffident and suspicious towards
the others) == > I will trust anybody (I will not be diffident)'.
This rule, given one single case of personal bad experience, or of bad observation, or referral,
or reputation, can invert its valence: I become suspicious in a generalized way. And if this rule
is diffused (all the agents or many of them use it) the impact will be a generalized collapse of
the general trust capital and atmosphere.
Analogously, if I follow an optimistic (non prudent) default rule: 'Except I have a negative
example I will assume that agents in this community are trustworthy'.
The same can hold for affective trust disposition (Chapter 5) and a generalized mood in
a given context that can be wasted by just one very negative personal experience or by the
contagion of different moods of others.
Of course the network is not necessarily uniform or equally connected; it might be an
archipelago of non well-connected 'islands' of very connected sub-communities. Thus the
propagation might just have local effects.