Agents, Roles and Systems

[Jochen Fromm]

Recently we had an interesting talk about ecological
inheritance, niches and their evolution. I think classes,
roles, functions, niches and jobs are all similar: they
describe the border or interface between the internal
and the external region of a complex system, between the
object/agent on the one hand and the system/organization
on the other hand: CAS || Role || Agent. Different roles are
for example group-member/-leader, student/instructor,
customer/vendor, friend/enemy, employee/employer.
Different systems are for instance

 object-oriented Systems
 Existence (computational/cognitive context): to be - class - objects

 Multi-Agent or social Systems
 Behavior (social context): to play - role - agents

 Biological Systems
 Purpose (biological context): to fulfil - function - organisms

 Ecosystems
 Occupation (ecological context): to occupy - niche - species

 Economic Systems
 Work (economic context): to work - job - employees

In social psychology and Multi-Agent Systems, the notion
"role" is very important. A role is a socially defined pattern
of behavior. The concept of a role originates in the theatrical
notion of roles as parts that actors play in a dramatic presentation.
(see for example "Social Psychology in the 90s", K. Deaux et al.,
6th Edition, Brooks/Cole Publishing Company, 1993).
Classes, roles, functions, niches and jobs can be defined in
a similar way:

a class is defined by the functional aspects associated with a
specific position in a computational or cognitive context

a role is defined by functional aspects associated with a
specific position in a social context

a niche is defined by functional aspects associated with a
specific position in a ecological context

etc.

If their is a basic calculus for Multi-Agent Systems (MAS) and
Complex Adaptive Systems (CAS) similar to the differential calculus
in Mathematics, it should describe changes in the boundary between
Agent and CAS, "topological" changes in classes, roles, functions,
niches and jobs. A natural duality used in Object-Oriented Programming
(OOP) is Aggregation/Inheritance. Because agents are only objects
with complete sovereignty or autonomy, this duality can be transfered
to MAS in form of Aggregation/Role-Differentiation.

A group is the fundamental social unit. There are two basic
group processes which affect the agent-system boundary: on the
one hand we have unifying group-formation/integration processes,
and on the other hand separating specialization/role-differentiation
processes. In social psychology it is known that roles can emerge
in the course of group interaction, a process named role-differentiation.

What do you think, is it possible to talk of a "calculus for CAS"
based on Aggregation/Role-Differentiation, or is this too far-fetched ?

Regards,

Jochen