COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00004 00003	.bb Introduction
C00010 00004	.skip 1
C00011 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source;
.GROUP SKIP 20

.cb MINIMAL INFERENCE - A WAY FOR PROGRAMS TO JUMP TO CONCLUSIONS

Abstract: We present a new mode of conjectural reasoning in first order
logic.  A minimality axiom schema expresses the conjecture that
the individuals forced to exist by a certain set of axioms are all
the individuals there are that also satisfy a given predicate.  
We believe that minimal reasoning corresponds to a common human
mode of conjectural reasoning and that intelligent computer programs
will also need to use minimal reasoning.
Yet other modes of minimal reasoning may be important.
.skip to column 1
.bb Introduction

	
	We will argue that people often legitimately draw conclusions from
a set of assertions that go beyond the logical consequences of these
assertions.  We formalize two such methods of conjectural reasoning, but
there may be others that will be useful in artificial intelligence.

	The general phenomenon is that people often draw conclusions
from a set ⊗A of sentences that they will not draw from a more
inclusive set ⊗B.  However, all mathematical logical systems
have the property that any consequence of ⊗A is also
a consequence of any ⊗B such that %2A_⊂_B%1.  This holds for both the
semantic notion of entailment and the syntactic notion of deduction.
.skip 1
.begin verbatim
John McCarthy
Artificial Intelligence Laboratory
Computer Science Department
Stanford University
Stanford, California 94305

ARPANET: MCCARTHY@SU-AI
.end

.turn on "{"
%7This draft of

PUBbed at {time} on {date}.%1