[Modeling] Summary

James Odell email@jamesodell.com
Sun, 08 Jun 2003 14:57:27 -0400

On 6/4/03 4:10 AM, Marc-Philippe Huget scribed:

> 3. Generic agent description" is trashed in my opinion, I find
> ridiculous to write a generic agent where the only attribute is an ID.

I am not so sure.  We need to think about what does every agent have in
common.  Unless they are deaf, dumb, and blind, every agent must be able to
perceive, determine what to do, and act.  In other words, do we need the
function analogous to the human senses -- as well as the ability to effect
state changes in the environment?  Even something as primitive as
receiveMessage and sendMessage.  Also, maybe something analogous to the
psychological notion of "id."

After reading more of my email, it seems that Stephen has said something
similar: on 6/5/03 6:56 PM:
> I have only had time to look at your document very quickly, but my first
> impression is that I like your approach.  However, I think the metamodel
> or profile does somehow need to represent the idea that agents can receive
> messages.  One way to do this would be to specify that agents have exactly
> one operation: receiveMessage.  A more elegant (but more difficult) approach
> would be to design the metamodel/profile to take into account agents' social
> abilities, such as the ability to send and receive messages and the
> maintenance of conversation and social agreement states.
> - Stephen