Breese, John S. & Horvitz, Eric J. & Henrion, Max - Decision Theory in Expert Systems and Artificial Intelligence.pdf

文档介绍：
Decision Theory in Exp ert Systems and Arti
cial In telligence
Eric J
Horvitz John S
Breese
puter Science Group Ro c kw ell In ternational Science Cen ter
Kno wledge Systems Lab oratory P alo Alto Lab oratory
Stanford Univ ersit y
High Street
Stanford
California
P alo Alto
CA
Max Henrion
Departmen t of Engineering and Public P olicy
Carnegie Mellon Univ ersit y
Pittsburgh
P A
July
Abstract
Despite their di
eren t p ersp ectiv es
arti
cial in telligence
AI
and the disciplines of decision
science ha v mon ro ots and striv e for similar goals
This pap er surv eys the p oten tial for
addressing problems in represen tation
inference
kno wledge engineering
and explanation within
the decision
theoretic framew ork
Recen t analyses of the restrictions of sev eral traditional AI
reasoning tec hniques
coupled with the dev elopmen t of more tractable and expressiv e decision
theoretic represen tation and inference strategies
ha v e stim ulated renew ed in terest in decision
theory and decision analysis
W e describ e early exp erience with simple probabilistic sc hemes
for automated reasoning
review the dominan t exp ert
system paradigm
and surv ey some recen t
researc h at the crossroads of AI and decision science
In particular
w e presen t the b w ork
and in
uence diagram represen tations
Finally
w e discuss issues that ha v e not b een studied in
detail within the exp ert
systems setting
y et are crucial for dev eloping theoretical metho ds and
computational arc hitectures for automated reasoners

This is the tec hnical rep ort v ersion of a pap er app earing in the Internationa l Journal of Appr oximate R e asoning
Sp ecial Issue on Uncertain t y in Arti
cial In telligence
This w ork w as supp orted b y a NASA
F ello wship to Eric Horvitz under Gran t NCC
to Stanford Univ ersit y
b y the National Science F oundation
under Gran t IRI