Jul 31, 1997 - Lorraine Body, Linda Rising, and Charles Weir provided detailed comments on the conference draft ... 10] Adele Goldberg and David Robson.
They used methods derived from statistical mechanics to show that the as- sumption ..... technician, one James Watt, were laying the foundations for a proper un-.
Abstract. In this paper, we model a recent legal case as presented in a court of first instance using argument schemes and an argumentation framework, ...
will set the stage for our argument that `precedence' is essential in examining circles. ..... The other wffs take the role of ..... But let's take it a step further.
an opponent's claim, what Ghosh et al. call call- out-target ...... ever they perform what we call argument extrac- ... Proc. of the 20th Int. conference on Computational ... The main reasons I oppose the death penalty are: #1) It is permanent. S6.
In this paper we propose a framework for tailoring an evaluative argument ... can be compelling in arguing for an alternative either because of its strength or ...
realization is performed through interface with the. LOLITA system (Smith et al., 1994). The current work employs a concept of argument based upon the study of ...
... on the relation between the comparative morpheme -er and its corresponding than-phrase, the analysis proposed here i
as the Web (Pennacchiotti and Pantel, 2006; Yates ... ner (Yates et al., 2007), there is no standardized manner for ..... famous entities like Michael Jordan.
neal intraoperative chemotherapy, which is the head of the reduced relative clause, after cytore- ductive surgery. Hyperthermic intraperitoneal intraoperative.
Marina del Rey, CA 90292-6695. {kozareva,hovy}@isi.edu. Abstract. A challenging ...... Ellen Riloff and Jessica Shepherd. 1997. A Corpus-. Based Approach for ...
help illustrate our points we draw on examples of three software methods: the .... Packaged software developers often have at best a distant relationship with their user population. This separation means that intermediaries B such as help desk ...
an argumentative framework. We will present a two fold approach con- sisting of (1) an argumentation framework for learning agents, and (2) an individual policy ...
miraculously. And he made Sodom and Gomorrah perish. God had power over these matters, and took action. Bob: Yeah, but that's just what it says in the Bible.
Aug 23, 2011 - that might appeal to youth at risk of smoking ..... New Delhi: High Court of Delhi, Available: ... Mississippi State University: Social Science Re-.
Jun 11, 2012 - Goldstone Research Unit · Philosophy, Politics and ..... argument nearly all of them are convinced that Rob's idea is right. In the Diversity Low ...
based on any one-way function as well is digital-signature NY, Ro]. Furthermore these primitives (and primitives derived from them, e.g. identi cation) were ...
a conceptual problem because it has been known since Miller and Chomsky ...... Anoop Mahajan, Judith Meier, Gereon Muller, Hans-Georg Obenauer, Eric.
7 Jun 2007 - EMORY LAW JOURNAL. Volume 56. 2007. Number 4. ARTICLES. SEX EQUALITY ARGUMENTS FOR REPRODUCTIVE. RIGHTS: TH
Moderation instead of Modelling: Some Arguments against Formal Engineering Methods. Matthias Rauterberg. Center for Research on User-System Interaction ...
encumbrances in the generated code that programmers usually (or presumably) do not write. Redundant arguments are useless for computing the semantics of ...
Michael J. Almeida and Graham Oppy. Sceptical theistsâe.g., William Alston and Michael Bergmannâhave claimed that considerations concerning human ...
NÄgÄrjuna's MÅ«lamadhyamakakÄrikÄ, translated and commented by. J.L. Garfield, Oxford University Press, 1995. [12] R.E. Nisbett, The Geography of Thought: ...
正ICETRANS. INF&SYST., VOL.E88−D, NO、9 SEPTEMBER 2005
2021 PAPER Special section on software Agent and’ts Apρlica tions
Applying Logic of Multiple-Valued Argumentation to Eastern
ity a皿d adaptability of LMA to various multiple−valuedness
required in argumentatien und訓nc斑迦㎡brmation. This Traditional syrnbo韮c AI models of reasoning have been ba− sicaUy fbunded upon logic such as first order predicate ca1−
cuius,品d later have㎞ed to seek m・re expressive 1・gics since the f・rmer is inadequate for dealing with situati・ns
which are incomplete, uncertain, or dynamicaUy changi皿9. On也e other hand, studies on argumenta毛]on and infor−
mal Iogic haΨe been do鵬with也eir own purposes, some− t口工1es criticizing traditienal logic as saying that it is not re−
ally a true legic but simply one of several mathematica三the−
odes. Argumentatio皿, at丘rst glance, seems to be synony−
mous with logic in some logic textbookS. In compariso蝋o Iogic, argurnentS are not proefs in the traditional s酷se, but
proofs that can faiL They are deft∋asible or ovem11ed proofs
when faced t・・也er王㎎umen題虻are bette…s仕・ng・・ 出a皿the o亘9三nal argument・Put it differently, a proof is in− variably a static obj ect, and an argument is a proof of dy−
namlc皿at蹴Therefbre, argumentation is more suitable to describ㎞g and processi皿9廿1e dynalnic and cha皿gi皿g nature of infor nation in a networked dist】dbuted informadon envi_ ro㎜ent, or to problem solving in ill−fbmed unce血in prob− lem domains where modeling is cli伍cult to make.
Recently it has been recognized by computer scien軸
6sts・and in particuIar AI researchers that argumentation has clear correlation with the interests of the newly巴加erg一
三ng丘eld of soc呈al computation such as multi−agent systems
aspect of LMA is most advantageous to other approaches to multiple一Ψalued a㎎umentation such巨s[4],[18]. Hae皿i et 品・uses the term’ argumentation’in his probabilistic argu− mentation systems[6], but it is not a dialectics・−based ftame−
work as opposed to LMA. as well as o也er argumentation
systems wi甑d坤ctical pro。f theory. lt should be noted 出at LIVLA is different f『om our fommer paper[20]in that it is
a4−valued paraconsistent argumentation system wjth differ− ent phiiosophy in its fbn皿aliza廿on, paying attention to the
notion of difference as a momentum of argum聞tation. Tbe paper is oエga」nized as follows. In Sect.2, we out−
1ine the underlying language for the legic of Inultiple−valued
argumentation, Extended Annotated Logic Programming (EALP), toge出er with its interpretation. EALP is a very expressive, in the sense that it allows to r巴present different
kinds of uncertainty such as vagueness a皿d inconsistency(or
paraconsistency)i蝋em 3 of multi−valuedness、 and in。om− pleteness with the help of de』1t nega加n・1n Sect・3, we
ouUine the logic of inultiple−valued argumentation(LMA).
in Sect.4, we illustrate two arg迦e斑example iロwhich a popular set of trut}1 values, the closed interΨal猟[0.1]of real mlmbers, and a power set of dates as a co:nplete lattice
are employed fbr argumentation under unceltain informa− tien. In Sect.5, we describe a signi{icant specialization of
L蜘o愉由e㎜awi血輪stem mind. We discuss in particular the third and fbuf曲h le】㎜as of Tetrale]㎜a which
Manusc坤t民ceived Novemもer 24,2004. Manusc『t revised May 16.2005.
are said to characte]rize Eastern thought and logic・In Sect・6,
we take up two argument examples to曲ow伽t LMA a1− 1・ws・f・r・a・kind・f pluralistic・argumentati。叫胴血si・n of
a)E・mai1:sawamura@ie.niigata−U.ac」P
Eastem and Westem a培umen血d皿. They also坦us蜘e the
b)E−mail:takeh」isa@csje、niigata.u、acjp
uses of由e敵d and fb賦h le….血the釦al sec目on we
DOI:10.1093/ietisy/e88−d9.2021
summarize the paper by d巴sc曲ing曲antages and implica一
Copyright@20e5 ’1 he lnstitute of Eleetronics, lnformation and Communicatien Engineers
IEICE TRANS.聴&SYST,VOLE89−D, NO.9 SEPTIEMBER 200S
2022
tjons of LMA,
in the ru]es can evaluate to the elements of 7’. The head of a rule is called a conclusion of a rule. An−
2・ Extended Annotated Logic Programs
notated o正)jective literals and annotated default literals in the
body of the rule are called anteceゴβπr50f the rule and as−
EALP is an underlying knowledge representation language
sumptions of the rule respectively. We identify a distributed
that we formalized for our logic of multiple−valued argu−
EALP with an a8ent, and treat a set of EALPs as a mul’」一
mentation LMA. EALP has two kinds of explici{ negation:
α88n砧yぷf醐.
Episternic Explicit Negation ‘一’and Ontological]巳xpl三cit Negati・n‘一一’, and the default negation ‘not’. lhey are sup−
2.2 1nterpretation
posed to yield a momentum or driving force for argumenta− tion or dialogue in LMA、 Here, we outline EALP.
1)efinitio皿4: (Extended annotated Herbrand base). The set of all annotated literals constructed f士om an EALP」P on
2」 Language
acomplete latticeワーof tuth values is called the extended
annotated Herbrand base璋. Definition 1:(Annotation and a皿netated atoms[9]). We assurne a complete lattice(?一,≦)of truth values, and denote
1)efinition S:(lnterpretation). Let?一 be a complete lat−
its least and greatest element by⊥and T resp㏄tiΨely. The
tice of tru血values, and P be a皿EALP. Then,也e interpre−
Ieast upper bound operator is denoted by u. An annotati皿
tation on P is the subset∬⊆・確of the extended annotated
is ei吐Ler an element of「アー(constant annotation), an annota−
tion variable on「r, or an annotation term. Annotation teml is defined recursively as」follows:an element of?−and anno− tation variable are annotation terms. In addition, if t t,...,tn
i鵬a皿notatiOn terms, then f(tl,_,tn)is an a皿otation term.
Here, f is a tota1 continuous fUnction of typeワm→7−. . If A is an atomic fbrmula andμis an annotation, then
Herbrand base畔of P such that for any annotated atom A,
1.lf A:μ∈Iandρ≦μ, then A:p{≡1(downward heredity);
2.if A:μ∈Iand A:ρ∈1, then A:(μUρ)∈…1(tolerance of difference);
3.If∼Alμ∈∫andρ≧μ, then∼A:ρ∈∬(upward heredity).
・A:μis an annotated atom. We assume an annotation func− tion r:ワ’→「1−, and define th at r(ノ1:μ):=A:(「μ). rA:μ
is called the epistemic explicit negation(e−explicit negation)
ofA:μ. 1n this paper, the e−explicit negatien ”A:μis embedded
into an annotated atom、4:「μ, and implicitly handled.
Defin」籠on 2:(A皿otated literals). Let A:μ be an anno−
The conditions l a皿d 20f De丘nition 5 reflect the defi− nition of the ideal of a co111Plete lattice of truth values. The
ideals−based semanticS was丘τst introduced for the interpre−
tadon of GAP by Kifer and Subrahmanian[9]. Our EALP for argumentation also employs this since it was shown that the general semantics with ideals’ 奄刀@more adequate輪也e resUicted one simply with a complete lattice of truth val−
tated atom・Then∼(A:μ)is the ontological explicit nega−
ues[19工.
tion(o−expliciE negation)of A :μ. An annotated objective
We de丘ne three notiens of inconsistencies correspond−
Uteral is either∼A:μ or A:μThe symbol・’・is also used
i皿gto three concepts of negation in EALP.
to denote co皿Plemel1町y an皿otated objective literals. Thus ∼∼
`:μ=A:μ
If L is an annotated objective literal, then not L is a
default negatipn of L, and called an annotated default litera!.
L・A:μ∈Iand rA:μ∈1⇔Iis epistemologically
An annotated literal is either of the form not L or L.
inCOnSiSte11t(e−inCOnSiStent).
The tems, epistemic n’egation and ontological negation, originate from Kifer and Lozinskii[8]. Note, however, that
the rneaning of our ontological explicit negation is different from their ontological negation, being properly adj usted to argumentation as can be seen in the succeeding sections.
Defi皿ition 3:(Extended Annotated Loglc Progralns (EALP)). An extended a皿notated logic prograrn(EALP)is aset Qf annotated rules of the fbrm:H← Ll&...&」Ln, where」H is an annotated obj ective litera1, and Li(1≦i≦n) are annotated literals in which the annotation is either a con−
2・If P」ayeri=Pla:ソerj’=二P(i≠」)then Argi≠Argj. 3.If、Playeri=・」P(匡≧ 3)then(Argi,Ar8i_1) ∈y;and if
P栂εrド0(f≧2)血en(Arg輌,Argi.1)∈エ.
involvement with negation. In this section, we wi110utline
1}efinition 12: (x/y−dialogue tree[13]). An xly−dialogue
our argumentation f『amework.「lb begin with, we describe
tree is a tree of moves such that every branch is an x/y−
abstract argumentation framework.
dialogue, and for al1 meves mov召i=(P, Argi), the children
of movei are all血ose moves(0,Ar8輌+lj)(」≧1)such也at
3.1 Abstract Argum」entation Framework
(A㎎’・1.」、Argi)∈x・
1〕面nition 13:(Provably x/yrjustified[13]). An x tY− 1}e血ni目on 8言(Abstract attack relation[5ユ). Let Ar8s be aset of abstract arguments. An attack relation x on Args is abinary relation on Args, i. e., x⊆Args× Args・. We define the argunientation se皿antics as the least fix−
dialogue D is a winning x∫y−dialogue⇔Ule terlnination of Dis a move of proponellt. An x/y−dialogUe tree T is a win−
ning x/y−dialogue tree⇔eΨery branch of T is a winning エ/y−dialogue. An argument A r8 is a provablyエ/y−j ustified
poiqt of the fUnction which coHects all acceptable argu−
argument⇔吐1ere exists a winningエ/y−dialogue tree with
ments.
Ar8 as its root.
De触ion 9:(鞠一acceptable and j副ied a㎎㎜ent[5】)・
Theorem 2:Let Ar85 be an abstract argument seしThen
Let x and y be attack relations on Ar8∫・ SupPose Argl−E
A㎎∈Ar8s is proΨably x/yづustified⇔Ar8 isエ/yrjusdfied・
Args and S⊆Ar8s. Then」Argl is x/y−acceptable wrt・∫ if fbr every Ar82 ∈Ar8s such that(Arg2,Argi) ∈xthere
exists、Ar83∈Ssuch吐1at(A rg3,Ar92)∈y・ The fuuction }「A rgs,xノァ InapPi皿g from P(A rgs) to
P(Args)is defined by FA ,g醐(∫)={A・8∈A・8・1A・8 isエ/y−acceptable wrt・ S}・ We denote a least fixpoint of 1『Args,xly by JA rgs.x∫γ ’An argtl−
ment A rg is x∫y−justified if A rg∈Jxty;an argument iSエ/y−
overruled if it is attacked by a x/yづustified argument;and an argument is xノンーdefensible if it is neither x/y−j usti fi ed
’nor x/y−ovemlled,
From here on, we concredze the abstract argumentati加 framework to introduce various notions. proper to.a logic of
multiple−valued argum号ntation LMA.
33 Annotated Arguments Defi皿ition 14:(Reductant and Minimal祀du幽nt). Sup− pose P is an EALE and Ci(1 ≦ 匡 ≦ R)are annO− tated−rules in P of the’ ?b窒香F」4:ρ輌←Ll&...&Lh∫, in which A is an atom. Letρ=・u{ρ1,∴,Pk}・Then the fbllowing annotated nlle is a reducta祉of P・ A:
’W・・writ・曲ply・F.ノ, and J.ly f・・馬綱1y蝋』.x∫,
ρ←Ll&_叫&…&Lf&一・&Lい・edU・t・nti・
when A r8ぷis b恒ious・SinCe Fエiy is monotonic・1it has a least
called a hUn㎞al reducta皿t when there d㏄s not exlst non−
fixpoint, and can be eonstructed by tbe iterative method[5}
empty proper subset S⊂{ρ1,_,Pk}such that P=〕S・
Definition 10:(Conflict・free[5]). Let Argsbe an abst]ract
Definition 15:(An皿otated a㎎凹皿lents)・ Let P be an
狂…1CE TRANS.!NIF.&S yS’r., VOL.E8S−D,NO.9 SEPTEMBER 2005
2024
EALP. An a皿otated argument in P三s a丘nite sequence Arg = [rl,...,rn]of rules i!】Psuch that曇)r every輌(1 ≦
3.5 Argumentation Semantics for EALP
f≦n),
By Theorem 1, we㎞ow JA r8s,aly(a=attack, a⊇y)is 1.ri輌s either a rule in」P or a min三ma五reductant in P.
confiict−free wrt. attack. Funhermore, we have,
2.For every annotated atom A:μin由e body of ri, there exists a rk(n≧k>のsuch that A:ρ(ρ≧μ)is head of
Theorem 3:工f a set of arguments S is con伍ct−free wrt. the defeat relat▲onship, then s is conHict−free wrt. the attack.
η ・
3.F()r every o−explicit negation∼A:μin the body of ri,
Proposition 3: ・IA rgs,ば15。=JArgs.d∫凹=JIArgs,d/d・
there exists aηヒ(n≧た〉輌)such that∼A:ρ(ρ≦μ)is head of rk,
4.There exists no proper subsequence of[r1,_,ηJ
In this paper, we employ/Args,dlsu to specifシthe set of
justified.arguments since the stricdy undercut is smallest or
which meets from the first to the th三rd condi目ons, and
simplest in these attack relations, and straightfbrward.
三ncludes r1.
Let P be an EALR andルtAS be a set of EALPs. Then we de丘ne the sernantics of argumentation on P by JArgsp,dtエu
Asubargumelnt of an argument Ar8 is a subsequence ofArg.
(Jp fbr short), and defi ne也e semantics of argumentation
The conclusions of n11es in Ar8 are called conclusions of
on MAS by JA rgs”As ,dtsu(JMAs fbr short). When we do not
Ar8, and the assump加ns of rules in Ar8 are ca11ed assulnp−
need to teU apart P a皿dルtAS,we simply denote Ar8ぷp and
tions ofA㎎. We wri te concl(Ar8)for 1山e set of conchlsions
Arg5、MAs by Ar8s, and denote Jp and JMAs by J. 」♪and
andαぷ∫η1(Arg)fbr the set of assumpt三〇ns of」4 rg. We denote
JMAs can be dialectically determined by也e dialecdcal proof
the set of a1!arguments in P by Argsp, and de五ne the set of
theory specified in Sect.3.2.・
all arguments in a set of EALPs MAS={KBi,…,KBn}by Ar8工〃AS =ArgSKBs U−・UAr8SKe,(⊆Ar8s」K、BtU…vrrBn)・
3.6 (Snictly)Well−B ehaved EALP
Anon−minimal reductant usually results in having a longer anteceden杜han a m三nimal reductant. if an argument
We scrutinized the argumentation semantics fbr EALP in
is made by non−minimal reductants, the argument may be
various directions, and found some logica1 anomalies(inter−
redundant in its content. Fur也ermore, a non−minimal re−
ested readers should refer to our foπner paper[19]). For
ductant occasionally tends to be irrelevant rule(refer to[19]
ex ample, even if the j us面ed arguments are conflict一丘ee,
fbr details).
the interpretation satisfying血em is not always consistent・ Tb avoid them, we introduced subset3 of EALP, called we皿一
3.4 Attack RelatiOn
behaved and strictly weU−behaved EALP. 1皿what follows, we assu皿e that agents’㎞owledge bases are sUictly weU−behaved EALPs fbr argu皿entation.
Definition 16: (Rebut). Arg l rebuts Arg2⇔tbere exists
Alμ1∈cohct(Argi)and・’一 A:μ2∈concl(Arg2)such・that
Definition 20: ((Strictiy)Well■behaved EA]LP).
μ1≧μ2,0r exists∼〆いμ1∈concl(Arg1)a皿d A:μ2∈i
Let P be an EALP白n a complete latticeワー of truth values,
concl(Arg2)such tha[μ1≦μ2.
∼A:μbe a conclusion of a rule in、P, and S⊆「アーbe a fi皿ite
subset such that S ≠φ. Then, if P satisfies the condition,
Definition 17: (Undercut). 」4r81 undercuts Ar82⇔
μ≦〕S⇒]ρ∈Sμ≦ρ,Pis called well−behαved EALP.
吐1ere ex三sts A :μ1 ∈ coπd(Argi)and not A :μ2 ∈
haddition, if an assumption not A:μof a rule in P
α∬m(Arg2)such血atμ1≧μ2,0r exists∼A:μ1∈
also satisfies the above condition, P is called stri‘tly well−
concl(Argi)and not ∼ A :μ2 ∈ assm(Ar82)such that
behaved EALP.
μ1≦μ2.
Proposition 1: For any Ar8i and Arg2 in Ar8s「, if Argi
4.lllustratiΨe Argument Examples
rebutS Arg2,1 such that 1トArgs is o・inconsistent. And if Argi undercuts Arg2, I such that lトArgs is d−inconsistent.
In order to see potentials of LMA for extensive』applic ations,
we、illustrate two argument examples that exploit usual and Defi皿ition 18: (Strictly undercut〕. A rgi strictly under−
cuts Ar82⇔Argi undercuts Arg2 and Arg2 does not under− cut Argi.
1)efinition 19: (Defeat). Argi defeats・4r82⇔Ar8 i un− dercuts Arg2,0r Argi rebutS Ar82 and/lr82 does not undeエー cut Aア8い
unロsual uses of complete lattices respectively・
Example 1:Let us consid¢r an argumentation about the pros and cons of the death penalty of Murderers・SupPose acomplete latticeワー=恨[O,1]2 ・where(#1,ρ1)≦(μ2,ρ2)
⇔μ1≦μ2andρ1≦m,…md in(μ,P)∈∼「,μ and P represent the degrees of an af丘rInation岨d a negation re− ■ コ
spectively. This tru.th va Iue can represe皿t e−1皿consistent o「
Propositien 2: For any 4 r81 and Arg2 in Args, if/Arg i at−
unknown state and degree of tru廿1 togethe: Let MAS =
tacks or defea臨Ar82, l such that 1トAr85 is o−inconsistent
{KB,,KB2,、KB3}be a set of EALPs, a皿d each、1(’Bi be a
or d−inconsistent.
knowledge base such that:
SAWAMURA and TAICAHASHI:APPLYING LOG rC OF MULTrPLE・VALUED ARGUME㎜工ON T[)EASTERN ARGUMENTS
2025
Fig・1 R・画。・・血・ng.arg・r・entS, wh・fピu‘・・ta・d・・f・・u・d・ユ・nポピ鋼d・f・・民b・tta1・th・ broken lines sta皿d fbr由e op已ra面on of the reducta皿t、 and a:gume耐s framer]in a吐Uck line are justifi巴d
argumellts・
(Arg21 with a c。nclusion with annotaUon(02,0.8)), an oP・
κB1ニ{ ∼agree(叔の:(α0、0・8)←
position of the death penalry is acknowledged to some de・ gree at the same time(A rg22 with a conclusion with anno−
声1ロre(∫amily, murderer):(0.8,0.0) &
tation(02,0.4))」’Westerners generally might prefer a defi−
ば已∫’re(∫antily, death):(0.7,0.0),
nite cenclusion of either true or false. However, as a whole,
』’ε(『醐ity, murderer):(1.0.0.0)←
we would say that吐1e agent society of血鶴agents三s in−
一一 aUowげamily、 murderer):(0.5、0・0),
clined to institute the death penalty in this argument among
由ose three a、gents. The exalmple would suggest a possibility
and feasibility of argurnent systems 1ike LMA as a function fc)r E−government in the future.
In Example 1, assume 一・(μ,ρ)=(LO一μ, LO一ρ),
and replace every o−expUcit negation‘∼’in each KBi by e−expUcit negation‘「’. Then, Arg is will nor more re− but Arg33, resulting in、Arg34 being juStified, and thus the
conclusion of Arg34 mea皿s”a bereaΨed f㎞且y desires and does not desire廿1e death penalty with amotations O.6碑一 spectively’, The degree凸f opPosition of the death p¢nalty
Figure l shows eΨery possible argument and every poss1−
ble attack relation among them. The just通ed arguments are JKBs={Ar8n,A㎎12,Ar813,Ar814, Ar821,Ar8ユ2・Ar831・
increases to O.6(Arg34)in comparison with the use of th巳 か ontologic…il negation. From dle viewpoint of Westem lo91c・
廿ris is a contradiction that Westerners hate most. However,
thanks to the paraconsistency LMA has, it avoids ex falso quodlibet. T is not a truth value that sirnply turns knowl− edge base into a meaningless one. Instead, It represeptS an ・pi・t・mi・・t・t・・n血・p・・P・・iti・n”血・d・a也P・nalty・i・・titu− tion,’agellts bear i皿L]ULA・
Thus agents can incorporate what也ey intend to do in
argumentati。n int・血・廿㎞・wledge・bas・by・・1㏄血g ap− propriate explicit negation accordmg to a topic of argumen−
A・83、}.rn a literal sense, the justifi・d argumentS・w・uld b・
able to read as]bHows:‘‘lt is sure吐lat a bereaved fam− ily hates a mロrderer(Ar831 with a conclusion with an、no−
ta口on.
Ex・mpl・2:L・t・us・c・n・id・・紐卿m・ntati・n・b・ut th・
tation(0.9,0.2)), and so agreement to the death penalty
mont hity schedule management・Here we use an unconven‘
is pardy possible(Arg32 with a conclusion with allnotatiol1
tional complete lattice of truthΨalues which is the power set
(0.6,0.0)).In addition, because a bereaved faエnily ca皿no匡
aCCept・ll・Wing・mU・d・・e・(A・gll Wi血aC・nCIU・i・n Wi也 a㎜o城on(0.5,0.0)), and desires the death penalty(Ar8日 with a conclusion with annotation(0.7,0.0)), a complete op−
position of tiユe death penalty is never acknowledged(A rg i4 with a conclusion with annotation(0.0,.0.8)). However, be−
canse there is no evidence to show that a dead ma皿ls m remorse fbr the crime, and the death can not atone a 9uilt
P({1,...,31D of the set of the monthly dates, with the or−
也e comp}ete lattice FO〔頂wiぬ恥ale㎜a wi也an Eastem mind, so that LMA allows for argumentadon of the kind
KB占={
seen in廿1e Eastem tradition of culture.
rwork{占):112,19,261←, 蘂’ゴ直y:{5.6、12,13}←1,
5.l Tetralemma KB。={ 一・ arrive(companent):{5) 一一 not paXup‘harge):ip}.
In the early philosophical literature and text of 13uddhism,
也e notion of four altemative positions(ca噂ko亘in Sans㎞t; Where KB.,κβロ,KBb a皿d KBσ stand for knowledge bases of a manager agent m, employee agents a, b and a sub−
shikttfunbetsu in Japanese)appears vely o丘en in argume皿ts on met且physical questions such as whether Nirviin4 is an ex−
contractor agent o respectiΨely. Agent m’s argument which
istent, whether Ta由互gata existS after death, in such a way
has the conclusi皿finish(proノ松ロ):{6}(the project will fin−
that all conceptually imaginable pos垣ons are exhaustive
ish on也e 6th)is justヨfied by the dialectical proof theory as
and exclusive such as”Nirv亘皿a is an existent, Nirvana is a ロ ’
shown in Fig.2.
non:existent, Nirvfina is both an eXistent a皿d a皿on−eXistent,
In the winning dialogue tree, initially Agent m says‘‘廿
Or Nirv亘na iS neither an eXiSt¢nt nOr a IIOn−eXiStent.”
acomponent wiII arrive on the 5th, and Agent a works on
These represent four logical possibilities of the form:
the 3th, the 4th and the 5th, th印the pr〔)j ect wiU finish on
(1)athrmation,(2)negation,(3)both affillnation and nega−
the 6th.”Then Agent o defeats it as follows‘‘I will be not
tion, and(4)neither aMrmation nor negation. The ancient
able to bring a component on the 5也ifthe additional charge
Indians believed that the truth with regard to any]matter Iay
is not paid.”But Agent m sUictly undercutS o’s argument by
in one of these alte而atives. Over the last few decades, the
saying 4‘I will pay it匡o you on the 8th.” For the first argument
Iogical structure of the four altema且ve positions has been
of Agent m, Agentロalso defeats by saying‘‘the 5th is a
the subject of a considerable amount of discussion and con−
holiday, and if the coworker b does not work, I do not want
troversy, and also of some speculatio4 in 吐te context of
to work on the 5th.’, However Agent b sUictly undercuts it
what is sornedmes called East−West philosophical’ Dcompar− ison. Nowadays”the four alternatives(positions)”is also
by saying‘‘I wi皿work on days except the 12th,191血and
termed”tetraiemma”or”tetrachotomジin the literature[1], [7],[14],[16],[21].
PAg己川m
Tetralemma is an epistemic state or way of recogniz− ing things, beings, objects, propositions, etc. We use similar patterns of expressions very often in our daily life, fbr exaln−
ple,‘‘Animals have moral rightS, Animals do not have moral
dghts, Animals have moral rights and do not, or A垣mals .neither have moral rights nor do not.”
T snictly
sajctly
undercut
undercut
f
t
⊥ Fig 2 The vvinning dialegue tree in Exarnple 2・
ITl 1f1
ltl
」
φ
Fig.3 0『O伽R and」r(チーon?R).
SAWAMURA:and TAKAHASHI:APPLY ING LOGIC OF MVLTPLE−VALUED ARGUMENT灯10N TO EASTERN ARGUMENTS
2027
In his book, Logos and i尤㎜a[21].緬uchi
dialectics.1t is a dialecUcal contradiction but not a contra臼
TokUryu contrasted Wes〔em and Eastern thought by char−
diction in formal logic, being a fbrm of the un並y of opPo.
acterizing the first as being determined by logos and the laト
sites as a diaiectical law in Hegelian and Marxist dialectics.
ter as being structured by the principles of the tetraleユ
The eastem style of民asoning is basicaUy dialectical from
thereby conceiving of logos as a method of exclusion and
scratch in the sense that it focuses on contradictions and how
of le㎜a as a way of撫dly inclusion(eΨen of the血d−
to reso1Ψe them or虹anscend也em or伽d the tru由in both.
dle), The tetralemma, in fact, is closely related to由e view of
In either world, we need to invent concepts to bo吐l discard
e皿ptiness in Buddhism and the culturahdiosyncrasy such as
and absorb those contradictory propositions. 「rurning our
holisticΨiew on world, society and ecosystem in nature[12].
eyes to the development in other traditional sciences, we can
It would be helpful to spe田late about the meanlngs
see that dialectical thought unites various opPosite concepts,
of the third and fourth lernmas in more details since they
principles, and theories through mediating logical links in
are beyond simply saying either true or false, so that we
could exploit them in applications to knowledge representa−
tion and argumentatlve reasonlng・
higher syn吐1etic constructions.
We use」the top symbol T for such an epistemic state or way of recognizing things, beings, objectS, propositions, etc., and annotate propositiens with the symbol T as in[2].
5.2The Third’Lemma:‘it is and it is not’ 5.3The Fourth Lemma: ‘it neither is nor is n。t’
The出ird lemma obviously seems to violate the law of non− contradiction by Aristotle. It. however, often appears in口le
Like血e third lemma, the fourth lemma al so appears very
sutras−of B uddhism and the eastern tradition of thought and
o丘en in the sutras of Buddhism and the eastern tradition of
culture. Daoists and Zen Buddhists, for example, see the
thought and culture, such as in Nag萌una’s
two sides of any apParent contradiction existing in a har−
Malamadhyamakakarik5, which is one of the Inost inHuen−
m皿y,opposed but{nterconnected, interpenetrating, and in−
tial work in the history of Indian philosophy[11], and in the
terdependent. in everyday situ ations where experienc己or
Heart Sutra.
des辻e is dorninant, easterners are tolerant of contradictions・
Murt三interprets it 1ike thi s. It represents the fUll con−
They haΨe thought that 1由e law of noncontradiction appUes
sciousness that no correspQnding aff…rl皿ation is aΨailable・
only to the realm of concepts and abstractions, which are
This is an、 extreme fbrm of non−committal. It is not an at−
merely re且ections of thiIlgs. In his recent book[12]、 a cul−
titude of decision, but of doubt.The competence of thought
1皿ral psychologist, Nisbett has delnonstrated tendency aLnd
is not questioned and reason is not缶anscended. It may cor−
evidence which indicate various di{丑≧rences of cognition and
respond to the agnosdc position[10]・According to廿1e in−
reasoning between Easterners andi Westerners, including the preferen直al exaエninalion of accepta皿ce or avoidance of con− tradictions.
terPretation of Jayatilleke, there is left a part of the deter−
minable constituting the un、iverse of discourse which is re−
』ed to by血e飴u曲le⊇since the second altern ative is
Let us see solne multi−faceted皿eanings of the third
the contrary and not吐1e contradictory, and the曲d asserts
lemma since we think they are he1pful to understand and
that也e subject has a combination of some of the contrary
use it加actual knowledge representadon as weU. Mur丘[10]
characteristics[7]. Here is a good exalnple illustrated by
says出at血e醐le㎜a represen馳at we ha⊇e con− sciousness of the olle−sidedness of Inere ls or mere ls not・ f・・examp1・, G・d exi・t and d・e・n・t・J・y・till・k・[7]s4Ys
1由at the historical examples show that the second alternatlve is the contrary and not the contTadictory of the五rst・for ex一
迦P1・, w・・tv・、 e田t,㎞・w1・dgev・…nd・ct Th・n・也・ t1丘rd asserts that the subject has a combina直on of the con一 剛charact・ristics. F・r・exampl・, when the st・tem・nt’u㎡一. verse is both finite and infinite’is made, it is intellded that
・the universe is丘nite in one dimension(in some respects) and infinite in another. Here is a typical locution fbr‘it is and is not, in our daily life(or rather, polltlclans may use 廿ris):When they arei asked a hard question to dea1 witll・the
answer would be‘yes and no’. Also in our daily life, we
s・me直mes rhet・亘cally use由・・xym…n・It is a phrase that
combines two words that seem to be the oPPosite of each other. For ex泣皿Ple, an open secret, too much of a good thing(bear’s service), a hapPy scream, a deafeni皿g silence
and so on. Oxymoron unit問two opPosipg words into an expression with a single meaning, without contradiction・
The third lemna also can be found eve蜘Western
Jayatilleke{刀:Aperson is happy, A person is not happy, A・Pers・n i・b・螂・pPy and u血・pPy…A・Pe・s・n i・n・ithr「 hapPy nor unhapPy. HapPiness in this context is a det官rm』■
n・t・q・ality・h砲・t・ri・ing・p・rs・n’・hed・垣・t・n・・Wh・n we remove the qualities of‘‘happiness∼’‘Mnhappiness,” or a mixture of the two, we are left with‘‘neutral hedonic tonel”
So a person who is‘‘neither happy nor unhappジcompnses th・・clas・・f卿le exp・essing・a・neutral・h・d・皿i・t加e・Such aclass need not necessadly l]e a null class、 although it could
be so sometimes. Hele is anether similar type of 10cutions・ whlch can be seen in our dai正y life as we111 His act is neither
軸t…w・。ng,・・w・ar・・ei血・・fo・n…g瓜nst』ct・ The founh iemma has had a wide spectrurn Of皿6an−
i。g…epi・t・㎡・・t・t…堀・g・, being・,・切ec底興 P・・P・・iti・n・. It w・uld・b・w・血y鋤d u・・fU1 t・1ist up’t
as fbllows fb抽e knowledge representatiop and argumenta’ tion described later:no com㎡血ent, no commenL no con° cem, no infonnation, an attitude of doubL a neutral or t「an−
、ce。d,血1・pi・t・血・・田t・, and・vi・w・f E岬血・ss□e Buddhist technical terms fbr the lack of independent exls° t,n、,, i。h,,ent・xi・t・n・e,・・es・en。・i噛・g・・W・d・gene「一
1EICE TRANS. INF&SYST, VOしE呂8−D, NO.9 S EFrEMBER 2005
2028
飢e these i・t・由・b・仕・m・ymb・1⊥・吾d血e c・nt・・t i・s・p−
Aristotie
posed to speci fy a meaning of it.
It migh由e instructiv飢o take up contemporary expres− sions that capture o酷aspect of the meaning of⊥. For ex−
ample’we have three ways to pronounce‘‘s垣p」medar・in
that Galileo made his argument by reductio ad absurdum
(Frαηcε醐Ger加卿’s・sta刀cの
f(〕rwhich吐1e default negation‘not‘has a crucial role in
(2)一・匡’酩加91f←has−w・αρ・ns.ma∬」ie蹴‘ti。n:t
the rule representation. Furthelτnore, we note that the head
(USA・and・UK’王ぷr伽cの
∼aristo’lehyp:tin the first nlle of Ga田eo’s argument does not undercut the assumption∼aristbt正ε一hyp:fof the
6.PEuraliStic er MulticU ltura 1 Argument Examples
second rule, that is, Galileピs aエgument is coherent or not self−defea1]ng, and Eastem agent does not undercut the as_
Wb i11ustrate two pluralistic or multicultural arguments based on the speciaIization of LMA in the previ皿s sec廿on. Exampl e 3:Let us consider the Wes tern and Eastern arg u−
ments against Aristotle. Ari stotle believed that吐1e heavier
・b・凪yi・・也e趣te・it飽ll・t・出e g・・und. We simply・ep−
resent this as aristot∼ehyp:tGalileo’s logical argument against吐Us pr㏄eeds as]f()110ws:‘‘Suppose that we have two bodies, a heavy one ca皿ed H and a light one called L. Under
Aristodピs assumpUon, H will fh11 faster than L. Now 8up−
pose that H a皿d L are joined togetheL Now what happens? Wel1, L plus}l is heavier that H so by the initial assu皿ptio皿
it should飼l faster that H alone. But in the joined body, L is lighter and will act as a‘brake’on H, and L plus H wilI
sumpロon∼αr輌∫ro’」ぱyp:fof the second rule in Gahleo・
argument.(lnterested readers should refer to[19エfbr the technical terms used.) In血i s example, all the argumen ts by Aristotle, Galileo
and Eastem agent become defensible. IncidentaUy, let us consider a Uttle modi血ed version of the example. We first
change Aiistotle’s belief as fbllowsl aristotleky、p:t←
not∼em」pirically.factual l t. AIld we make one more agent apPear on the stage, who is a mOdern scientist having
a丘rm belief on verificationism.
MOdern scientist’s knowledge: ∼empi・ically−factua∼:t←π・t・scientげically.verified:t. The叫it is obvious that Aristotle’s argument is overruled, Galileo ’s and a modern scientist’s ones are justified, and
制slowe抽at H alone. Hence it follows from the initiai
Eastem agent’s one is sti皿defbnsible.
assumption that L plus H w沮both faster and slower that H
In today’s globalized world, such a pluralistic or mul−
alone. Since血is is absurd, the mitial assumpdon must be false.”On the other hand, Eastemers prefer a more hもlis− tic or dialectical argument like this:“Aristode is based.on
dcultur司卿ment of co㎜on interest is getdng mo祀血一 portant to us than ever・This is no exception even in agents’ ら
SOClety・
abelief也at the physical obj ect is free from any influences of other contextual factors, which is impossibIe in reality、”
Example 4(Nobunaga’dying words:Zehinioyobazu)l This
[12]
is a familia・hist・rical event, typically reflec直ng也e meaning
These are well translated into EALP as fbUows:
of⊥, and is the Iord Nobunaga’s dying words when he was
GaIileo’s knowledge:
’attacked and assassinated by his vassal Ak㏄hi in the histor−
∼aristetie−hyp:t←faster(L+H, H):T
icaI event at TempIe Honnoliji in the Warring State Period
faste・(L+H, H):t←n・t・arist・tl・」ryP:f
of the Japa皿ese history. It is‘Zbhinioyobazu’in Japanese
fdi1ぷ∫εr(L+」「f,晶「):f←5∬Ol伊er(L+」Y,、晶0:t
which mea皿s that he is in such a cognitive or epistemic state
sl・wer(L+・H, H):t←brake(L, H):t
that Akechi’s rebellious act is neither right nor wrong, sim−
占嬬ε(L,胡:t
Ply and best expressed as Akechi」rebellious act:⊥,
systems、”In Handbook of Dcfeasible Reasoning and Unce宜ainty Managemen【Systems, voL5:Algorithms for Uneertainty and De−
Welling【on、 New ZeaIand. His research intercstS include computational
feaslble Reasoning, cd. J. Kohlas and S.Mora1、 pp.221−2B8. Kluwer.
10gic, logical fo岨Clation of computer sottware and artificial intelligence.
2000.
正]eis a member of rPSJ, JSAI JSSST, and Philosophy of Science Society
【刀 K.N. Jayat川eke,‘亡The logic of fbur altematives、”Pllilosophy Eas【
ofJapan.
and W,巳st, voLl7, pp,69−83,1967.
[81M,焔fer and EL. LoZinskli,“A logic for reasoning w三【h inconsis− tcncy、”」. Au【om, Reasoning、 voL9. pp.179−215,1992,
[9]M.Ki歓and VS. Subra㎞manian,‘Theory ofgeneralized annotated l。gic programm lng nnd itS apPlicati・ns、“J・Lo9・Progra皿・. v。L12, pp.335−397,1992,
[10]T.R. Murd, The Central Phnosophy of Buddhism−A Study of the Madhyamika System, George Allen and Unwin l.960.
[11]N5g珂una, The Fundamental Wisdom of血e Middle Way, Nfigdrjuna’s MO]amadhyamakakdrikE, translated and commented by J.L, Garfield, Oxfbrd University Press,1995.
[12]R.E. Nisbett, The Geography ofThought:How Asians andWestern− ers Think Diffe間tly_and Why、 The Free Press,2003.
[13]H,Prakken and G. Sartor,“Argumen【・based extended logic pro− gramming with defeasible pdorities.’l J. Appl. Non・Classical Logics, vol.7, no.1,pp.25〔75,1997.
[14]P.T. Raju,“Th巳principle of four−cDrnered negation in indian phi}os・ ophy,”Rev, Mctaphysics, vo1,7, pp.694−713,1954.
[15】C.Reed and T,J. Nom)an, eds.. Argumentation Machines, Kluwer
Academic Publishers,2004. [16]D,S. Ruegg,‘The uses of【he four pesitions of the catulko巨al】d 由eproblem of the.desc可tion ofreali【y in mahayana buddhism,‘’」. Indian Phjlosophy, voL5, pp.1−71.1977.
〔17j H, Sawamura and E. M訂es,“How agents should exploit tetralemma with an eas【enl mind in argumen【adon,”in IntelligenヒAgents and Multi−Agent Sys¢ms V皿. ed. M Barrey and N. Kasabov. LNAI 3371,pp.259−278, Springer,2004,
[18]M、Schroeder and R. Schweimeier,‘’Fuzzy argumentation and ex− tended logic p《ograrnming,’, Proc. ECSQARU WOrkshop oo Ad− vances in Argumentation,pp.1−13,2001. [19] T.Takaliashi and H. Sawamura,”A logic of multiple−valued ar−
gumentation、”Proc. Third Intemational Joint conference om Au・
tonOmous Agen6 and Muhi Agent Systems(AAMAS12004), pp.80〔}−807.ACM,2004.
[20]Y.Umeda, T.τ歓ahashi, and H. Sawamura,”An argumentanon frarnewo【k based on paracensistent logic,”工Artificia1血telligence ofJapan, voL19レno.2, pp.83−94,2004.
[21】T. Yamauchi, Log。s and Le㎜a, Iwariami.1974.
Takehisa]P】kahashi was a master course studentofGraduate School ofScience andTech− nOlogy, Niigata UniverSity、 He now wOrks fOr Niiga【a CannOl】Co.