AIC Seminar Series
There Aren't Any Real BDI Theories/Models Out There
|David J Israel||Artificial Intelligence Center, SRI International||[Home Page]|
Notice: Talk and discussion
Date: Thursday, May 29th 2008 at 3:30pm
Location: Enterprise Room (SRI E building) (Directions)
"Plans and Resource-Bounded Practical Reasoning," by Martha Pollack, Michael Bratman and myself was recently honored as one of two influential papers in the field of Autonomous Agents and Multi-Agent Systems by the International Foundation for AAMAS. This is a deeply gratifying award, .... on the other hand.... and here's where I look this particular gift-horse in the mouth.
I assume that the paper is considered influential for being one of the first fairly clear arguments for and statements of the BDI approach to autonomous agents. And certainly there has been a large body of work advertising itself as in the BDI line, much of it citing our paper. But, I think much of that is false (though not dishonest!) advertising. In my mind, at least, "BDI" was and still is to be parsed as "(BD)I"; that is, though we argued for the insufficiency of Decision Theory (BD-theory), we did not at all argue for its non-necessity. My starting point then and now was (some form or other of) Decision Theory, a theory of action choice under uncertainty with outcomes at the very least ordered by desirability. But much of the work in the so-called BDI tradition deals only flat-out beliefs and flat-out goals. If we call flat-out beliefs "acceptances", then we can say that much of this work is really in a AGI tradition. Maybe that is the tradition to be working in, but AGI ain't BDI.
The talk will even include a sketch of a (Heaven Help Us!) BADI account.
N.B. Just as my co-authors were largely responsible for the clarity and felicity of the prose of the original paper and most of the good ideas, as well, I am wholly responsible for this take on our work and, as the French would say, on its Re-ception.