Copyright © IFA C Artificial Inte lli gence. Leningrad. US SR 1 9R3 ! ON A COMPUTER-ORIENTED FORMALIZATION OF PLAUSIBLE REASONING IN F. BACON-J. S. MILL'S STYLE (MAIN PRINCIPLES AND COMPUTER EXPERIMENTS) V. K. Finn, M. I. Zabezhailo and O. M. Anshakov A ll Ullio ll Ill stitllle of Scientific alld Technical Inf ormation, VINIT!, M OSCO II ', USS R Abstract. The paper generalizes the problems of creation the solvers of some special types. The computer-aided formaliza- tion of inductive approaches of F.Bacon-J.S.Mill is proposed. The experiments with computer-orientation are discussed • .,. Keywords. Inductive inference; inductive reasoning; correlati- .,,:·on theory; empirical dependence; plausible reasoning. 1. Artifitial Intelligence may be considered to be a discipline,which creates and studies computer-orient- ed formal means of simulating man's cognitive activity. naturally, it would be desirable for this simula- tion to contain both a sensor model of the external world's perception and a model of a reasoning individu- al. This individual sould be able to combine sensor data and stored know- ledge in order to apply logical pro- cedures of data processing for making appropriate decisions. It follows from the above, that an investigation of different types of reasonings (both in theoretical, and in applied aspects - from the view- point of an effectivity of their si- mulation on a computer is a fundamen- tal problem of Artifitial Intelligen- ce. This problem is included in the variety of problems, which may be called "computer epistemology". The term "reasoning" may be understood as: reasoning = heuretic + deduction. In this sense reasoning is connected with the simulation of mentality: heuretic (formally described) is si- mulation of fragments of intuition while deduction is a systematical formulation of heuretic results (con- clusions from hypotheses, obtaining of consequences including those,that are in contradiction with given data or, conversely those that confirm automatically created hypotheses etc.). Successes and difficulties of deduction's realization on computers 3 51 are well kno\vn. The simulation purely heuristic procedures, including em- pirical induction, is less advanced (both in theoretical, and practical respects). The concept of "heuretic" may be approximated by a system of plausible inference rules with dif- ferent organizations (strategies). !n this paper an approach to the for- malization of heuristic reasoning of the type is examined, i.e. that with data containing incomplete informa- tion, hence, requiring the use of ma- ny-valued (non-classical) logics as a means of formalization. Let Vdes be a set of designated truth-values, V undes - a set of undesignated truth- -values, Val-function of valuation; cp,. .,., <Pit, 1.p,j( - formulas of the lo- gics involved. Inference rule R will be called "plausible inference rule" (p.i.r.) if there are such CPf, · .· CPIt, If/" that 'P{, ... rpll f-;[ VI', VQl(Y',)E V<::Je$ and Va.{ C 1.p) E V undes' but the num- ber of such cases is sufficiently small. Let r be some set of formulas in the language of first-order predicate logic, CPCa,)is CP(X./Qi)' i=1, ••• ,n, then: