Information Technology Reference
In-Depth Information
(i.e., they act according to the contrast concept), that specific interaction has a
probability
p(a2)
of providing observers evidence of a shared concept.
We make some very simple and quite generally accepted assumptions about our
agents' psychology. If the observer witnesses evidence (blind to whether it is true or
illusory agreement), then his own conceptual state increases its relevance in his mind
(i.e., our cognitive assumption; c.f., [15], [16], [17]), will be more likely to guide his
behaviour in the future (i.e., our motivational assumption; c.f., [18]), and the observer
will want to interact with that particular agent again in the future (i.e., our social
assumption; c.f., [19]).
3
ABM Implementation
The theory presented in section 2 is implemented in an agent-based model (ABM). In
summary, the ABM represents how concepts spread and get stronger (or weaker) in a
social group, by observing the behaviour of other members. In the ABM, each
individual is an agent (actor, A), which acts according to its concept with probability
equal to the strength of the concept. That behaviour is observed by another agent
(observer, O), and that changes the strength of its concept. In general, if the observed
behaviour agrees with the behaviour expected from O's concept, then O's concept
strengthens. Conversely, if the observed behaviour differs from what is expected from
that concept, then O's concept weakens. Concurrently, the agents begin to interact
more frequently with those that have strengthened their concepts. In the following
paragraphs we describe the details of the ABM.
In the social group that the ABM represents, one can set the number of members
that belong to the group. Each agent can have one of five different related concepts or
versions of the same concept and each of the concepts or versions is represented by a
number in the [0, 1] interval, labelled the
coefficient
of the concept. This coefficient
determines the probability that an agent behaves according to the given concept. The
initial values of the coefficients are sampled from a normal distribution with a mean
and standard deviation, which can be set. The model checks that the assigned
coefficients will always remain in the [0, 1] interval.
Agents modify the strength of their concept's coefficients by observing the
behaviour of other agents. Every time they see behaviour consistent with their
concepts, the corresponding coefficients are incremented by 0.02. On the other hand,
if the observed behaviour is not consistent with their concepts, the corresponding
coefficients are decremented by 0.02. The model makes sure that the coefficients
always remain inside the [0, 1] interval. Thus, when an agent sees that another agent
acts according to its concept, it is more probable that the agent will act according to
its own concept in the future. These actions spread concepts throughout the group.
Agents develop interaction preferences as they observe each other. Specifically,
agents will tend to interact more frequently with agents who have confirmed their
concepts in previous interactions, and indirectly, they will be less likely to interact
with those that have not confirmed their concepts. This aspect of the ABM limits the
diffusion of concepts, given that it imposes certain heterogeneity to the diffusion
