/LMAR=0/XLINE=3/FONT#0=BASL30/FONT#1=BASI30/FONT#2=BASB30/FONT#10=BAXM30/FONT#11=ZERO30/FONT#5=MICR25/FONT#6=NGB30
␈↓ ↓H␈↓¬␈↓ α≡OFFICE MEMORANDUM  *  STANFORD UNIVERSITY  *  OFFICE MEMORANDUM

␈↓ ↓H␈↓ε␈↓ βBSTANFORD ARTIFICIAL INTELLIGENCE LABORATORY

␈↓ ↓H␈↓α␈↓ ¬]January 17, 1979




␈↓ ↓H␈↓αTo:␈↓ αH␈↓Anne Gardner

␈↓ ↓H␈↓␈↓αFrom:␈↓ αH␈↓John McCarthy, Director, Arti≡cial Intelligence Lab.

␈↓ ↓H␈↓αSubject:␈↓ αH␈↓Your ␈↓↓Notes Toward a Legal Reasoning Program␈↓

␈↓ ↓H␈↓        I␈α⊃like␈α∩your␈α⊃␈↓↓Notes␈α⊃Toward␈α∩a␈α⊃Legal␈α⊃Reasoning␈α∩Program␈↓.␈α⊃ A␈α⊃program␈α∩that␈α⊃could␈α∩do␈α⊃the
␈↓ ↓H␈↓example␈αon␈αthe␈α≡rst␈αpage␈α
in␈αa␈αreasonable␈αway␈αwould␈α
represent␈αa␈αconsiderable␈αadvance␈αin␈α
arti≡cial
␈↓ ↓H␈↓intelligence.

␈↓ ↓H␈↓        I␈α
agree␈αwith␈α
your␈α
solution␈αto␈α
the␈αproblem␈α
as␈α
a␈αdecision␈α
tree␈αbranching␈α
on␈α
the␈α"signi≡cant"
␈↓ ↓H␈↓interpretations of the law and the facts.

␈↓ ↓H␈↓        I␈α∩think␈α∩you␈α∩will␈α∩have␈α∩to␈α∩use␈α∩circumscription␈α∩or␈α∩some␈α∩other␈α∩method␈α∩of␈α⊃non-monotonic
␈↓ ↓H␈↓reasoning␈α
in␈αa␈α
central␈α
way␈αin␈α
the␈α
program.␈α The␈α
essence␈α
of␈αlegal␈α
rules␈α
is␈αthat␈α
they␈αhave␈α
exceptions
␈↓ ↓H␈↓that can't be incorporated in the text.

␈↓ ↓H␈↓        Perhaps␈α
you␈α
should␈α
do␈α
an␈α
axiomatic␈α
formalization␈α
before␈α
you␈α
attempt␈α
a␈α
program.␈α It␈α
should
␈↓ ↓H␈↓be␈αeasier␈αand␈αwill␈αclarify␈α
the␈αreasoning.␈α You␈αwill␈αhave␈αto␈α
decide␈αon␈αthe␈αmost␈αconvenient␈α
way␈αof
␈↓ ↓H␈↓handling␈α≠non-monotonicity.␈α≤ Apropos␈α≠of␈α≤axiomatization,␈α≠you␈α≤should␈α≠read␈α≤Bob␈α≠Moore's
␈↓ ↓H␈↓forthcoming␈α⊂M.I.T.␈α⊂Phd␈α⊃thesis.␈α⊂ I␈α⊂can␈α⊃show␈α⊂you␈α⊂a␈α⊃copy,␈α⊂and␈α⊂perhaps␈α⊃he␈α⊂will␈α⊂make␈α⊃you␈α⊂one,
␈↓ ↓H␈↓though it's rather long.

␈↓ ↓H␈↓        The␈αbiggest␈αproblem␈αwill␈αcome␈αin␈αputting␈αin␈αsu≠cient␈αreal␈αworld␈αknowledge.␈α I␈αam␈αnot␈αsure
␈↓ ↓H␈↓that␈α∂Moore's␈α∞axioms␈α∂for␈α∂knowledge␈α∞and␈α∂action␈α∂are␈α∞usable,␈α∂but␈α∞they␈α∂are␈α∂certainly␈α∞a␈α∂step␈α∂in␈α∞the
␈↓ ↓H␈↓direction you need.

␈↓ ↓H␈↓p3␈α∞-␈α∞I␈α∞don't␈α∞agree␈α∞with␈α∞your␈α
reasons␈α∞for␈α∞rejecting␈α∞MYCIN's␈α∞probabilistic␈α∞reasoning.␈α∞ While␈α
the
␈↓ ↓H␈↓judge␈α∀cannot␈α∀be␈α∀frank␈α∀if␈α∀he␈α∀has␈α∀acted␈α∃in␈α∀the␈α∀way␈α∀you␈α∀describe,␈α∀he␈α∀may␈α∀actually␈α∃do␈α∀so.
␈↓ ↓H␈↓Incidentally,␈αthis␈αraises␈αa␈αbig␈αproblem.␈α Suppose␈αthat␈αthe␈αreasoning␈αa␈αlawyer␈αor␈αjudge␈αcan␈αput␈αon
␈↓ ↓H␈↓paper␈αis␈αgoverned␈αby␈αcertain␈αconventions,␈αand␈αdoes␈αnot␈αparallel␈αthe␈αreasoning␈αthat␈αgoes␈αon␈αin␈αhis
␈↓ ↓H␈↓head.␈α∞ Would␈α
you␈α∞follow␈α
one␈α∞of␈α
them␈α∞or␈α
would␈α∞your␈α
program␈α∞have␈α
both␈α∞an␈α∞internal␈α
reasoning
␈↓ ↓H␈↓process␈α∞and␈α
what␈α∞it␈α∞"put␈α
on␈α∞paper"?␈α
 The␈α∞same␈α∞comment␈α
applies␈α∞to␈α
the␈α∞separation␈α∞of␈α
diagnosis
␈↓ ↓H␈↓and␈αremedy.␈α
 Of␈αcourse,␈αthere␈α
may␈αbe␈αother␈α
reasons␈αfor␈α
rejecting␈αthe␈αMYCIN␈α
model,␈αand␈αmy␈α
own
␈↓ ↓H␈↓opinion, and even hope, is that legal reasoning will require new methods.

␈↓ ↓H␈↓p3 - I agree with the limitations you have chosen.

␈↓ ↓H␈↓p4␈α∂-␈α∂I␈α∂agree␈α∂that␈α∂you␈α∂should␈α∂choose␈α∂a␈α∂subdomain␈α∂of␈α∂law␈α∂in␈α∂which␈α∂there␈α∂isn't␈α∂major␈α∞doctrinal
␈↓ ↓H␈↓controversy, and I suppose this contracts problem is in such a subdomain.

␈↓ ↓H␈↓p6␈α⊂-␈α⊃It␈α⊂would␈α⊃very␈α⊂good␈α⊂if␈α⊃the␈α⊂program␈α⊃would␈α⊂state␈α⊂its␈α⊃assumptions,␈α⊂e.g.␈α⊃ the␈α⊂one␈α⊃about␈α⊂the
␈↓ ↓H␈↓Uniform Commercial Code.
␈↓ ↓H␈↓αAnne Gardner␈↓ ¬]January 17, 1979␈↓ 
nPage 2␈↓ 


␈↓ ↓H␈↓p10␈α∪-␈α∪Formulation␈α∪or␈α∪use␈α∪of␈α∀rules␈α∪concerning␈α∪burden␈α∪of␈α∪proof␈α∪will␈α∀involve␈α∪non-monotonic
␈↓ ↓H␈↓reasoning.␈α It␈αseems␈α
to␈αme␈αthat␈αyour␈α
proposal␈αto␈αuse␈αthem␈α
to␈αsolve␈αthe␈αmissionaries␈α
and␈αcannibal
␈↓ ↓H␈↓problem merely displaces the problem.

␈↓ ↓H␈↓p11␈α
-␈αMaybe␈α
the␈α
program␈αneed␈α
not␈α
say␈α"probably␈α
 the␈α
July␈α1␈α
o≥er␈α
expired"␈αbut␈α
rather␈α
"if␈αthe␈α
July
␈↓ ↓H␈↓1 o≥er is considered to have expired, then ...".

␈↓ ↓H␈↓p12 - The decision tree seems to be the right idea.