Dear UAI Community,

our discussion during the last 2-3 weeks about "Bayesian Networks and
Belief Functions" was very interesting and exciting. However, at the end,
there were so many comments and remarks from different people, it took so
much time to read and understand all of them, and finally, it was no longer
possible for me to answer as I would have liked.

Now, as our heads have cooled down a little, let me add one more but
general remark. What we did during the last 2-3 weeks is REASONING UNDER
UNCERTAINTY. There are different open questions like "Is total ignorance
represented correctly by prior probabilities", "Is Bel(R)=0.72 and
Pl(R)=0.98 reasonable", etc. The true answers for these question are unkown
and therefore uncertain (although everybody has a personal opinion).

The point to realize is that the formalisms we were defending are quite
different from our way of REASONING UNDER UNCERTAINTY during our
discussion. What happened is that everybody tried to defend her/his opinion
by (1) giving reason supporting her/his approach and (2) giving reasons
refuting other approaches. In other words, everybody tried to find
ARGUMENTS and COUNTER-ARGUMENTS. Finally, our opinions (or beliefs) result
from attaching (subjective) weights to the arguments. For me, for example,
an argument given by Philippe Smets is more weighty than an argument given
by Kevin Van Horn.

In my eyes, the idea of finding and weighting arguments and
counter-arguments  reflects properly our way of judging open questions and
is therefore an important element of common sense reasoning. In constrast,
we are not judging open questions in terms of "conditional probabilities"
or in terms of "mass functions".

The consequence of this should be that we, the experts of "reasoning under
uncertainy", were trying to formalize this fundamental idea of finding and
weighting arguments and counter-arguments. A first step in this direction
is done by our "PROBABILISTIC ARGUMENTATION SYSTEMS" (PAS) formalism. The
general idea behind the PAS model is exactly as I described above the
nature of common sense reasoning.

My feeling therefore is that the PAS formalism is much closer to the real
nature of reasoning under uncertainty than most formalisms defended
intensely during our discussion. Unfortunately, only small attention has
been paid to the PAS formalism until today.

Best wishes,


Rolf Haenni



PS1: Note that the contents of this message is nothing else than an
additional argument supporting the PAS formalism :-)

PS2: Interested people can download the slides of my recent talk about the
PAS formalism from: http://www2-iiuf.unifr.ch/tcs/rolf.haenni/eth.pdf.gz.
The corresponding technical report is available at:
http://www2-iiuf.unifr.ch/tcs/publications/article/HKL99.htm. Furthermore,
note that we are giving a tutorial at the IJCAI'99 conference in Stockholm.
More information is available at http://www.dsv.su.se/ijcai-99/.



************************************************************************
*                                                                      *
*  Dr. Rolf Haenni                        __/  __/  __/ __/  _______/  *
*  Institute of Informatics (IIUF)       __/  __/  __/ __/  __/        *
*  University of Fribourg, Switzerland  __/  __/  __/ __/  _____/      *
*  Phone: ++41 26 300 83 31            __/  __/  __/ __/  __/          *
*  Email: [EMAIL PROTECTED]        __/  __/  ______/  __/           *
*                                                                      *
************************************************************************
*  World Wide Web: http://www2-iiuf.unifr.ch/tcs/rolf.haenni           *
************************************************************************

Reply via email to