Reasoning With Incomplete Knowledge in a Resource-Limited Environment: Integrating Reasoning and Knowledge Acquisition

Abstract

This pdper describes a n a pproa c h t o rea~o~in g with incomplete in f o rmati on in a r eeo ur ce -l1m1t ed env ir o nm e nt. Approaches to dat e either as s ume infini t e r esou r ce s and pn,ceed t o e num er:.l t e a _Jar ge infere n ce space , o r as s um e few r eso ur ces and 1gno re th e missing info rm atio n. Th ey do n ot reason abou t resource co nstraint s and th e inferen c e methods amis sab l e under them. A hear say -II-lik e system 1S desc ri bed wher e each knowledg e source i s _a sepa rat e pr od uc tion sys t em. During rule evalua t1 on, a rule anteceda nt i s evalua t e d usi ng minimal - resour ce methods . A rule antecedant i s eval ua t ed t o. true, false , o r an expected re so ur ce co s t to acq ulr: the info r ma t io n ne cessa ry to complete its eva luat1on. r" r conf li c t resolution chooses a parti a ll y evaluat ed rule, it po s t s a goal asking other . kn o wled ge sou r ces t o provide th e mi ss ing informat1on, s u spen ds th e knowl e d ge sou r ce, and informs t~e knowledge source ' s manager ab o ut th e s uspenslon and l·ng goa l The ma na ger d ec id es wh et heraccompa ny · f th e goa l is worth pu r suing now, the amou nt o resou rc es to apply t o th e task, what knowledge source t o app l y , and when to g ive up . Th e knowledge so ur c;s that a tt em pt th e goa l ca n i mpl eme nt a varie~y o infe re nti al and knowledg e acquisitio n techn1ques.