Re: is communication a requirement to collaborate?

> According to my definition (see other response), agent1's activity
> could change if agent2 is removed - the activity of agent1 may not be
> observable after removal of agent2, but it could affect the mental
> state of agent1.

OK, I suppose I didn't see where you were aiming.  But there is still 
communication going on, isn't there?
 
> I am just not sure if communication (through whatever means) is the
> only ability to influence each other. Martijn has mentioned
> reinforcement learning as beeing a way to influence the overall agent
> activity without explicit communication among them.

Likewise, the "overall agent activity" cannot be modified in a non-trivial
way [1] by the reinforcement signal unless the agents are sensitive to one
another's presence.

What I'm driving at is, there's often an artificial separation between
communication as explicit exchange of bits, and communication as observing
someone else.  In a computer, what difference does it make?  Bits travel
from one agent to the other and information is conveyed thereby.  It's
just our human-level storytelling about the system that distinguishes
these two cases.  I don't think the storytelling is very useful.


Anthony

[1]  Meaning, not just modified independently by the reinforcement signal, 
as if they were both alone, but rather their joint behavior is modified.