Page 105 - Socially Intelligent Agents Creating Relationships with Computers and Robots

88 Socially Intelligent Agents

stimuli evolves in relation to the resolution of some of the problems. As long
as a task hasn’t been fulfilled, it can still be perceived as a task to be done by an
agent passing by; if another agent does it, the stimulus disappears or decreases.
Agents learn, for example by the use re-enforcement according to tasks they
have done already, and thus according to what was not done by other entities.
The system is completely socially regulated, although not a single agent has the
perception of the existence of other agents.
This is often considered as a radically opposed point of view to the preceding
one. It has a great flexibility in terms of the openness in the system. On the
other hand it is not easy to know how this approach can help to coordinate
agents that need to exchange complex information or confidential information
that cannot be abandoned in an open environment.
None of these approaches exhibits agents which combine the two different
abilities: to be able to meet new entities with potentially different ways of com-
municating (openness) and integrate with them into a normal communication
network so that they can exchange important information or consciously or-
ganise common work. This double social competence could be defined as the
ability to build trust (a "coherent" trust: which doesn’t affect the survival of
the individual agents or the system) and is clearly hard to create with artificial
entities.

3. Social Intelligence And Creativity
It could be argued that because the kind of inputs that humans get from their
interactions are of diverse forms, are more complex and carry more information
than written messages, that this explains why they are more able to make infer-
ences about unknown people. In the description of human interactions, not only
body movements and positions are studied but also geographical relative posi-
tion and the use of time in relations - as can be seen with [15] and [16]. I claim
that this argument is not relevant, since the characteristics of human interaction
can be recognised in very artificial settings. Here the example of interactions
among humans who use computer networks to communicate is relevant. A
human looking at a screen has necessarily less data coming from the interaction
channel than the computer itself, but he or she seems to be able do much more
about it, and be able to turn this data into information. Two examples illustrate
the social complexity that can emerge from exchanges over open networks:
academic discussion lists, and communities of teenagers playing games on the
Internet.
Academics frequently exchange points of view via the Internet and do so
publicly in discussion lists. Watching the traffic on these lists, one can identify
unofficial reasons that lead people to participate. These include: the creation
of their own reputation and the discovery of allies. Sometimes, considering the