Hostname: page-component-586b7cd67f-g8jcs Total loading time: 0 Render date: 2024-11-24T02:38:21.819Z Has data issue: false hasContentIssue false

Abstract Argumentation and Answer Set Programming: Two Faces of Nelson’s Logic

Published online by Cambridge University Press:  03 May 2022

JORGE FANDINNO
Affiliation:
University of Nebraska Omaha, Omaha, NE, USA (e-mail: [email protected])
LUIS FARIÑAS DEL CERRO
Affiliation:
IRIT, Universite de Toulouse, CNRS, Toulouse, France (e-mail: [email protected])
Rights & Permissions [Opens in a new window]

Abstract

In this work, we show that both logic programming and abstract argumentation frameworks can be interpreted in terms of Nelson’s constructive logic N4. We do so by formalising, in this logic, two principles that we call noncontradictory inference and strengthened closed world assumption: the first states that no belief can be held based on contradictory evidence while the latter forces both unknown and contradictory evidence to be regarded as false. Using these principles, both logic programming and abstract argumentation frameworks are translated into constructive logic in a modular way and using the object language. Logic programming implication and abstract argumentation supports become, in the translation, a new implication connective following the noncontradictory inference principle. Attacks are then represented by combining this new implication with strong negation. Under consideration in Theory and Practice of Logic Programming (TPLP).

Type
Original Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2022. Published by Cambridge University Press

1 Introduction

Logic programming (LP) and Abstract Argumentation Frameworks (AFs) are two well-established formalisms for Knowledge Representation and Reasoning (KR) whose close relation is well-known since the introduction of the latter: besides introducing AFs, Dung (Reference Dung1995) studied how logic programs under the stable models (Gelfond and Lifschitz Reference Gelfond and Lifschitz1988) and the well-founded semantics (Van Gelder et al. Reference Van Gelder, Ross and Schlipf1991) can be translated into abstract argumentation frameworks. Since then, this initial connection has been further studied and extended, providing relations between other semantics and ways to translate argumentation frameworks into logic programs (Nieves et al. Reference Nieves, CortÉs and Osorio2008; Caminada and Gabbay Reference Caminada and Gabbay2009; Wu and Caminada Reference Wu and Caminada2010; Toni and Sergot Reference Toni and Sergot2011; DvorÁk et al. Reference DvorÁk, Gaggl, Wallner and Woltran2011; Caminada et al. Reference Caminada, SÁ, AlcÂntara and DvorÁk2015).

On the other hand, Nelson’s constructive logic (Nelson Reference Nelson1949) is a conservative extension of intuitionistic logic, which introduces the notion of strong negation as a means to deal with constructive falsity, in an analogous way as intuitionism deals with constructive truth. Pearce (Reference Pearce1996, Reference Pearce2006) showed that a particular selection of models of constructive logic, called equilibrium logic, precisely characterize the stable models of a logic program. This characterization was later extended to the three-valued stable model (Przymusinski Reference Przymusinski1991) and the well-founded semantics by Cabalar et al. (Reference Cabalar, Odintsov, Pearce and Valverde2007). Versions of constructive logic without the “explosive” axiom $\varphi \to (\sim\!\varphi \to \psi)$ have been extensively studied in the literature (Nelson Reference Nelson1959; LÓpez-Escobar Reference LÓpez-Escobar1972; Thomason Reference Thomason1969; Almukdad and Nelson Reference Almukdad and Nelson1984; Odintsov Reference Odintsov2005; Odintsov and Rybakov Reference Odintsov and Rybakov2015; Kamide and Wansing Reference Kamide and Wansing2015) and can be considered a kind of paraconsistent logics, in the sense, that some formulas may be constructively true and false at the same time. The notion of equilibrium has been extended to one of these logics by Odintsov and Pearce (Reference Odintsov and Pearce2005), who also showed that this precise characterize the paraconsistent stable semantics (Sakama and Inoue Reference Sakama and Inoue1995).

In this paper, we formalize in Nelson’s constructive logic a reasoning principle, to be called non-contradictory inference (denoted 1), which states that

  • NC “no belief can be held based on contradictory evidence.”

Interestingly, though different from the logic studied by Odintsov and Pearce, the logic presented here is also a conservative extension of equilibrium logic (and, thus, also of LP under the stable models semantics) that allows us to deal with inconsistent information in LP. The interesting feature of this new logic is that, besides LP, it also captures several classes of AFs, under the stable semantics. It is worth to mention that the representation of AFs in this new logic is modular and it is done using an object language level. Recall that by object language level, we mean that AFs and its logical translation share the same language (each argument in the AF becomes an atom in its corresponding logical theory) and the relation between arguments in the AF (attacks or supports) are expressed by means of logical connectives. This contrast with meta level approaches, which talk about the AFs from “above,” using another language and relegating logic to talk about this new language. It is important to note that, as highlighted by Gabbay and Gabbay (Reference Gabbay and Gabbay2015), the object language oriented approaches have the remarkable property of providing alternative intuitive meaning to the translated concepts through their interpretation in logic. In this sense, from the viewpoint of constructive logic, AFs can be understood as a strengthened closed world assumption (Reiter Reference Reiter1980) that we denote as CW:

  • CW “everything for which we do not have evidence of being true or for which we have contradictory evidence, should be regarded as false”

The relation between AFs and logic has been extensively studied in the literature and, as mentioned above, can be divided in two categories: those that follow an object language approach (Caminada and Gabbay Reference Caminada and Gabbay2009; Gabbay and Gabbay Reference Gabbay and Gabbay2015; Gabbay and Gabbay Reference Gabbay and Gabbay2016) and those that follow a meta-level approach (Besnard and Doutre Reference Besnard and Doutre2004; Caminada and Gabbay Reference Caminada and Gabbay2009; Grossi Reference Grossi2011; DvorÁk et al. Reference DvorÁk, Szeider and Woltran2012; Arieli and Caminada Reference Arieli and Caminada2013; Doutre et al. Reference Doutre, Herzig and Perrussel2014; Besnard et al. Reference Besnard, Doutre and Herzig2014; DvorÁk et al. Reference DvorÁk, Gaggl, Linsbichler and Wallner2015). In particular, the approach we take here shares with the work by Gabbay and Gabbay (Reference Gabbay and Gabbay2015) the use of strong negation to capture attacks, but differs in the underlying logic: constructive logic in our case and classical logic in the case of Gabbay and Gabbay’s work. On the intuitive level, under the constructive logic point of view, attacks can be understood as

  • AT “means to construct a proof of the falsity of the attacked argument based on the acceptability of the attacker”

On the practical level, the use of constructive logic allows for a more compact and modular translation: each attack becomes a (rule-like) formula with the attacker – or a conjunction of attackers in the case of set attacking arguments (Nielsen and Parsons 2007) – as the antecedent and the attacked argument as the consequent. Moreover, when attacks are combined with LP implication, we show that the latter captures the notion of support in Evidential-Based Argumentation Frameworks (EBAFs; Oren and Norman Reference Oren and Norman2008): for accepting an argument, these frameworks require, not only its acceptability as in Dung’s sense, but also that it is supported by some chain of supports rooted in a kind of special arguments called prima-facie.

2 Background

In this section, we recall the needed background regarding Nelson’s constructive logic, logic programming, and argumentation frameworks.

2.1 Nelson’s constructive logic

The concept of constructive falsity was introduced into logic by Nelson (Reference Nelson1949) and it is often denoted as $\textbf{N3}$ . It was first axiomatized by Vorob’ev (Reference Vorob’ev1952), and later studied by Markov (Reference Markov1953), who related intuitionistic and strong negation, and by Rasiowa (Reference Rasiowa1969), who provided an algebraic characterization. Versions of constructive logic without the “explosive” axiom $\varphi \to (\sim\!\varphi \to \psi)$ are usually denoted as $\textbf{N4}$ and they are based on a four valued assignment for each world corresponding to the values unknown, (constructively) true, (constructively) false and inconsistent (or overdetermined). The logic $\textbf{N3}$ can be obtained by adding back the “explosive” axiom. We describe next a Kripke semantics for a version of $\textbf{N4}$ (Thomason Reference Thomason1969; Gurevich Reference Gurevich1977) with the falsity constant $\bot$ , which is denoted as $\textbf{N4}^\bot$ by Odintsov and Rybakov (Reference Odintsov and Rybakov2015). We follow here an approach with two forcing relations in the style of the work by Akama (Reference Akama1987). An alternative characterization using 2-valued assignments plus an involution has been described by Routley (Reference Routley1974).

Syntactically, we assume a logical language with a strong negation connective “ $\sim\!$ ”. That is, given some (possibly infinite) set of atoms At, a formula ${\varphi}$ is defined using the grammar:

\begin{equation*}{\varphi} \quad ::= \quad \bot \ \mid \ a \ \mid \ \sim\! {\varphi} \ \mid \ {\varphi} \wedge {\varphi} \ \mid \ {\varphi} \vee {\varphi} \ \mid \ {\varphi} \to {\varphi},\end{equation*}

with $a \in At$ . We use Greek letters ${\varphi}$ and ${\psi}$ and their variants to stand for propositional formulas. Intuitionistic negation is defined as ${\neg {\varphi} \underline{\underline{\mathrm{def}}} ({\varphi} \to \bot)}$ . We also define the derived operators ${{\varphi} \leftrightarrow {\psi} \underline{\underline{\mathrm{def}}} ({\varphi} \to {\psi}) \wedge ({\psi} \to {\varphi})}$ and ${\top \underline{\underline{\mathrm{def}}} \sim\! \bot}$ .

A Kripke frame ${\mathcal{F}} = {\langle W,\leq \rangle}$ is a pair where W is a nonempty set of worlds and $\leq$ is a partial order on W. A valuation ${V : W \longrightarrow 2^{At}}$ is a function mapping each world to a subset of atoms. A Nelson’s interpretation (N-interpretation) is a 3-tuple ${{\mathcal{I}} = {\langle {\mathcal{F}},V^+,V^- \rangle}}$ where ${{\mathcal{F}} = {\langle W,\leq \rangle}}$ is a Kripke frame and where both $V^+$ and $V^-$ are valuations satisfying, for every pair of worlds $w,w' \in W$ with $w \leq w'$ and every atom $a \in At$ , the following preservation properties:

  1. (i) $V^+(w) \subseteq V^+(w')$ , and

  2. (ii) $V^-(w) \subseteq V^-(w')$ .

Intuitively, $V^+$ represents our knowledge about constructive truth while $V^-$ represents our knowledge about constructive falsity. We say that ${\mathcal{I}}$ is consistent if, in addition, it satisfies:

  1. (iii) $V^+(w) \cap V^-(w) = \varnothing$ for every world $w \in W$ .

Two forcing relations $\models^+$ and $\models^-$ are defined with respect to any N-interpretation ${{\mathcal{I}} = {\langle {\mathcal{F}},V^+,V^- \rangle}}$ , world $w \in W$ and atom $a \in At$ as follows:

\begin{eqnarray*}{\mathcal{I}}, w \models^+ a &\text{ iff }& a \in V^+(w),\\{\mathcal{I}}, w \models^- a &\text{ iff }& a \in V^-(w).\end{eqnarray*}

These two relations are extended to compounded formulas as follows:

\begin{eqnarray*}{\mathcal{I}}, w \not\models^+ \bot\\{\mathcal{I}}, w \models^+ \varphi_1 \wedge \varphi_2 &\text{ iff }& {\mathcal{I}}, w \models^+ \varphi_1 \text{ and } {\mathcal{I}}, w \models^+ \varphi_2\\{\mathcal{I}}, w \models^+ \varphi_1 \vee \varphi_2 &\text{ iff }& {\mathcal{I}}, w \models^+ \varphi_1 \text{ or } {\mathcal{I}}, w \models^+ \varphi_2\\{\mathcal{I}}, w \models^+ \varphi_1 \!\to\! \varphi_2 &\text{ iff }& \forall w'\!\geq\! w \ {\mathcal{I}}, w' \!\not\models^+\! \varphi_1 \text{ or } {\mathcal{I}}, w' \!\models^+\! \varphi_2\\{\mathcal{I}}, w \models^+ \sim\!\varphi &\text{ iff }& {\mathcal{I}}, w \models^-\varphi\\{\mathcal{I}}, w \models^- \bot\\{\mathcal{I}}, w \models^- \varphi_1 \wedge \varphi_2 &\text{ iff }& {\mathcal{I}}, w \models^- \varphi_1 \text{ or } {\mathcal{I}}, w \models^- \varphi_2\\{\mathcal{I}}, w \models^- \varphi_1 \vee \varphi_2 &\text{ iff }& {\mathcal{I}}, w \models^- \varphi_1 \text{ and } I, w \models^- \varphi_2\\{\mathcal{I}}, w \models^- \varphi_1 \!\to\! \varphi_2 &\text{ iff }& {\mathcal{I}}, w \models^+ \varphi_1 \text{ and } {\mathcal{I}}, w \models^- \varphi_2\\{\mathcal{I}}, w \models^- \sim\!\varphi &\text{ iff }& {\mathcal{I}}, w \models^+\varphi.\end{eqnarray*}

An N-interpretation is said to be an N-model of a formula $\varphi$ , in symbols ${\mathcal{I}} \models^+ \varphi$ , iff ${\mathcal{I}} , w \models^+ \varphi$ for every $w \in W$ . It is said to be N-model of a theory $\Gamma$ , in symbols also ${{\mathcal{I}} \models^+ \Gamma}$ , iff it is an N-model of all its formulas ${{\mathcal{I}} \models^+ \varphi}$ . A formula $\varphi$ is said to be a consequence of a theory $\Gamma$ iff every model of $\Gamma$ is also a model of $\varphi$ , that is ${\mathcal{I}} \models^+ \varphi$ for every ${\mathcal{I}} \models^+ \Gamma$ . This formalization characterizes $\textbf{N4}$ while a restriction to consistent N-interpretation s would characterize $\textbf{N3}$ . As mentioned above, $\textbf{N4}$ is “somehow” paraconsistent in the sense that a formula $\varphi$ and its strongly negated counterpart $\sim\!\varphi$ may simultaneously be consequences of some theory: for instance, we have that ${\{a, \sim\! a\}} \models^+ a$ and ${\{a, \sim\! a\}} \models^+ \sim\! a$ . Intuitively, these two forcing relations determine the four values above mentioned: a formula $\varphi$ satisfying ${\mathcal{I}} \not\models^+ \varphi$ and ${\mathcal{I}} \not\models^- \varphi$ is understood as unknown. If it satisfies ${\mathcal{I}} \models^+ \varphi$ and ${\mathcal{I}} \not\models^- \varphi$ , is understood as true. False if ${\mathcal{I}} \not\models^+ \varphi$ and ${\mathcal{I}} \models^- \varphi$ , and inconsistent if ${\mathcal{I}} \models^+ \varphi$ and ${\mathcal{I}} \models^- \varphi$ .

2.2 Logic programming, equilibrium logic, and here-and-there Nelson’s models

In order to accommodate logic programming conventions, we will indistinctly write ${\varphi \leftarrow \psi}$ instead of ${\psi \to \varphi}$ when describing logic programs. An explicit literal is either an atom $a \in At$ or an atom preceded by strong negation $\sim\! a$ . A literal is either an explicit literal l or an explicit literal preceded by intuitionistic negation $\neg l$ . A literal that contains intuitionistic negation is called negative. Otherwise, it is called positive. A rule is a formula of the form $H \leftarrow B$ where H is a disjunction of atoms and B is a conjunction of literals. A logic program $\Pi$ is a set of rules.

Given some set of explicit literals ${\mathbf{T}}$ and some formula $\varphi$ , we write ${{\mathbf{T}} \models^+ \varphi}$ when ${\langle {\mathcal{F}},V^+,V^- \rangle} \models^+ \varphi$ holds for the Kripke frame ${\mathcal{F}}$ with a unique world w and valuations: $V^+(w) = {\mathbf{T}} \cap At$ and $V^-(w) = {\{\ { a }\ \big|\ { \sim\! a \in {\mathbf{T}}}\ \}}$ . A set of explicit literals ${\mathbf{T}}$ is said to be closed under $\Pi$ if ${\mathbf{T}} \models^+ H \leftarrow B$ for every rule $H \leftarrow B$ in $\Pi$ .

Next, we recall the notions of reduct and answer set (Gelfond and Lifschitz Reference Gelfond and Lifschitz1991):

Definition 1 (Reduct and Answer Set) The reduct of program $\Pi$ w.r.t. some set of explicit literals ${\mathbf{T}}$ is defined as follows

  1. i. Remove all rules with $\neg l$ in the body s.t. $l \in {\mathbf{T}}$ ,

  2. ii. Remove all negative literals for the remaining rules.

Set ${\mathbf{T}}$ is a stable model of $\Pi$ if ${\mathbf{T}}$ is a $\subseteq$ -minimal closed set under $\Pi$ .

For characterizing logic programs in constructive logic, we are only interested in a particular kind of N-interpretation s over Here-and-There (HT) frames. These frames are of the form ${\mathcal{F}_{HT}} = {\langle {\{h,t\}}, \leq \rangle}$ where $\leq$ is a partial order satisfying $h \leq t$ . We refer to N-interpretation s with an HT-frame as HT-interpretation s. A HT-model is an N-model which is also a HT-interpretation. We use the generic terms interpretation (resp. model) for both HT and N-interpretations (resp. models) when it is clear by the context. At first sight, it may look that restricting ourselves to HT frames is an oversimplification. However, once the closed world assumption is added to intuitionistic logic, this logic can be replaced without loss of generality by any proper intermediate logic (Osorio et al. Reference Osorio, PÉrez and Arrazola2005; Cabalar et al. Reference Cabalar, Fandinno, FariÑas del Cerro, Pearce and Valverde2017).

Given any HT-interpretation, ${\mathcal{I}} = {\langle {\mathcal{F}_{HT}},V^+,V^- \rangle}$ , we define four sets of atoms as follows:

\begin{eqnarray*}H_{\mathcal{I}}^+ &\underline{\underline{\text{def}}}& V^+(h)\\H_{\mathcal{I}}^- &\underline{\underline{\text{def}}}& V^-(h)T_{\mathcal{I}}^+ &\underline{\underline{\text{def}}}& V^+(t)\\T_{\mathcal{I}}^- &\underline{\underline{\text{def}}}& V^-(t)\end{eqnarray*}

These sets of atoms correspond to the atoms verified at each corresponding world and valuation. Every HT-interpretation ${\mathcal{I}}$ is fully determined by these four sets. We will omit the subscript and write, for instance, $H^+$ instead of $H^+_{\mathcal{I}}$ when ${\mathcal{I}}$ is clear from the context. Furthermore, any HT-interpretations can be succinctly rewritten as a pair ${\mathcal{I}} = {\langle {\mathbf{H}},{\mathbf{T}} \rangle}$ where ${\mathbf{H}} = H^+ \cup \sim\! H^-$ and ${\mathbf{T}} = T^+ \cup \sim\! T^-$ are sets of literals. Footnote 1 Note that, by the preservation properties of N-interpretations, we have that ${{\mathbf{H}} \subseteq {\mathbf{T}}}$ . We say that an HT-interpretation ${\mathcal{I}} = {\langle {\mathbf{H}},{\mathbf{T}} \rangle}$ is total iff ${\mathbf{H}} = {\mathbf{T}}$ . Given HT-interpretations ${\mathcal{I}} = {\langle {\mathbf{H}},{\mathbf{T}} \rangle}$ and ${\mathcal{I}}' = {\langle {\mathbf{H}}',{\mathbf{T}}' \rangle}$ , we write ${\mathcal{I}} \leq {\mathcal{I}}'$ iff ${\mathbf{H}} \subseteq {\mathbf{H}}'$ and ${\mathbf{T}} = {\mathbf{T}}'$ . As usual, we write ${\mathcal{I}} < {\mathcal{I}}'$ iff ${\mathcal{I}} \leq {\mathcal{I}}'$ and ${\mathcal{I}} \neq {\mathcal{I}}'$ .

Next, we introduce the definition of equilibrium model (Pearce Reference Pearce1996).

Definition 2 (Equilibrium model) A HT-model ${\mathcal{I}}$ of a theory $\Gamma$ is said to be an equilibrium model iff it is total and there is no other HT-model ${\mathcal{I}}'$ of $\Gamma$ s.t. ${\mathcal{I}}' < {\mathcal{I}}$ .

Interestingly, consistent equilibrium models precisely capture the answer set of a logic program. The following is a rephrase of Proposition 2 by Pearce (Reference Pearce1996) using our notation.

Proposition 1. Let $\Pi$ be a logic program. A consistent set ${\mathbf{T}}$ of explicit literals is a stable model of $\Pi$ if and only if ${\mathbf{T}}$ is the set of explicit literals true in some consistent equilibrium model of $\Pi$ .

More in general, it has been shown by Odintsov and Pearce (Reference Odintsov and Pearce2005) that the (possible nonconsistent) equilibrium models of a logic program capture its paraconsistent answer sets (Sakama and Inoue Reference Sakama and Inoue1995).

The following propositions characterizes some interesting properties of HT and strong negation that will be useful through the paper Footnote 2 :

Proposition 2. (Persistence) Any HT-interpretation ${\mathcal{I}}$ , formula $\varphi$ and world $w \in {\{h,t\}}$ satisfy:

  1. 1. $I, w \models^+ \varphi$ implies $I, t \models^+ \varphi$ , and

  2. 2. $I, w \models^- \varphi$ implies $I, t \models^- \varphi$ .

Proposition 3. (HT-negation) Any HT-interpretation ${\mathcal{I}}$ , formula $\varphi$ and world $w \in {\{h,t\}}$ satisfy:

  1. i. ${\mathcal{I}}, w \models^+ \neg\varphi$ iff ${\mathcal{I}}, t \not\models^+ \varphi$ , and

  2. ii. ${\mathcal{I}}, w \models^+ \neg\neg\varphi$ iff ${\mathcal{I}}, t \models^+ \varphi$ , and

  3. iii. ${\mathcal{I}}, w \models^+ \neg\neg\neg\varphi$ iff ${\mathcal{I}}, w \models^+ \neg\varphi$ , and

  4. iv. ${\mathcal{I}}, w \models^- \neg\varphi$ iff ${\mathcal{I}}, w \models^- \sim\!\varphi$ .

2.3 Abstract argumentation frameworks

Since their introduction, the syntax of AFs have been extended in different ways. One of these extensions, usually called SETAFs, consists in generalizing the notion of binary attacks to collective attacks such that a set of arguments B attacks some argument a (Nielsen and Parsons 2007). Another such extension, usually called Bipolar AFs (BAFs), consists in frameworks with a second positive relation called support (Karacapilidis and Papadias Reference Karacapilidis and Papadias2001; Verheij Reference Verheij2003a; Amgoud et al. Reference Amgoud, Cayrol and Lagasquie-Schiex2004). In particular, Verheij (Reference Verheij2003b) introduced the idea that, in AFs, arguments are considered as prima-facie justified statements, which can be considered true until proved otherwise, that is, until they are defeated. This allows introducing a second class of ordinary arguments, which cannot be considered true unless get supported by the prima-facie ones. Later, Polberg and Oren (Reference Polberg and Oren2014) developed this idea by introducing Evidence-Based AFs (EBAFs), an extension of SETAFs (and, this, of AFs) which incorporates the notions of support and prima-facie arguments. Next we introduce an equivalent definition by Cayrol et al. (Reference Cayrol, Fandinno, FariÑas del Cerro and Lagasquie-Schiex2018), which is closer to the logic formulation we pursue here.

Definition 3. (Evidence-Based Argumentation framework) An Evidence-Based Argumentation framework $\mathbf{EF}={\langle \mathbf{A},\mathbf{R}_a,\mathbf{R}_s,\mathbf{P} \rangle}$ is a 4-tuple where $\mathbf{A}$ represents a (possibly infinite) set of arguments, $\mathbf{R}_a \subseteq 2^{\mathbf{A}} \times {\mathbf{A}}$ is an attack relation, $\mathbf{R}_s \subseteq 2^{\mathbf{A}} \times \mathbf{A}$ is a support relation and $\mathbf{P} \subseteq \mathbf{A}$ is a set of distinguished prima-facie arguments. We say that an $\mathbf{EF}$ is finitary iff B is finite for every attack or support $(B,a) \in \mathbf{R}_a \cup \mathbf{R}_s$ .

The notion of acceptability is extended by requiring not only defense against all attacking arguments, but also support from some prima-facie arguments. Furthermore, the defense can be provided not only by defeating all attacking sets of arguments, but also by denying the necessary support for some of the non-prima-facie arguments of these attacks.

Definition 4. (Defeat/Acceptability) Given some argument $a \in \mathbf{A}$ and set of arguments $E \subseteq \mathbf{A}$ , we say

  1. 1. a is defeated w.r.t. E iff there is some $B \subseteq E$ s.t. $(B,a) \in \mathbf{R}_a$ , Def(E) will denote the set of arguments that are defeated w.r.t. E.

  2. 2. a is supported w.r.t. E iff either $a \in \mathbf{P}$ or there is some $B \subseteq E \setminus{\{a\}}$ whose elements are supported w.r.t. $E \setminus{\{a\}}$ and such that $(B,a) \in \mathbf{R}_s$ ,

  3. 3. a is supportable w.r.t. E iff it is supported w.r.t. $\mathbf{A} \setminus {Def}({E})$ ,

  4. 4. a is unacceptable w.r.t. E iff it is either defeated or not supportable,

  5. 5. a is acceptable w.r.t. E iff it is supported and, for every $(B,a) \in \mathbf{R}_a$ , there is $b \in B$ such that b is unacceptable w.r.t. E

Sup(E) (resp. UnAcc(E) and Acc(E)) will denote the set of arguments that are supported (resp. unacceptable and acceptable) w.r.t. E.

Then, semantics are defined as follows:

Definition 5. A set of arguments $E \subseteq \mathbf{A}$ is said to be:

  1. 1. self-supporting iff $E \subseteq {Sup}({E})$ ,

  2. 2. conflict-free iff $E \!\cap\! {Def}({E}) \!=\! \varnothing$ ,

  3. 3. admissible iff it is conflict-free and $E \subseteq {Acc}({E})$ ,

  4. 4. complete iff it is conflict-free and $E = {Acc}({E})$ ,

  5. 5. preferred iff it is a $\subseteq$ -maximal admissible set,

  6. 6. stable iff $E = \mathbf{A} \setminus {Un}{Acc}({E})$ .

SETAFs can be seen as special cases where the set of supports is empty and all arguments are prima-facie. In this sense, we write $\mathbf{SF}= {\langle \mathbf{A},\mathbf{R}_a \rangle}$ instead $\mathbf{EF} = {\langle \mathbf{A},\mathbf{R}_a,\varnothing,\mathbf{A} \rangle}$ . Furthermore, in their turn, AFs can be seen as a special case of SETAFs where all attacks have singleton sources. In such case, we just write ${\mathbf{AF}= {\langle \mathbf{A},\mathbf{R} \rangle}}$ instead $\mathbf{SF}= {\langle \mathbf{A},\mathbf{R}_a \rangle}$ , where $\mathbf{R} = {\{\ { (b,a) }\ \big|\ { ({\{b\}},a) \in \mathbf{R}_a }\ \}}$ For this kind of frameworks, the respective notions of conflict-free (resp. admissible, complete, preferred or stable) coincide with those being defined by Nielsen and Parsons (2007) and Dung (Reference Dung1995), respectively.

To illustrate the notions of support and prima-facie arguments, consider the well-known Tweety example:

Example 1 Suppose we have the knowledge base that includes the following statements:

  1. 1. birds (normally) can fly,

  2. 2. penguins are birds,

  3. 3. penguins cannot fly and

  4. 4. Tweety is a penguin.

We can formalize this by the following graph:

where pT, bT and fT respectively stand for “Tweety is a penguin”, “Tweety is a bird” and “Tweety can fly.” Double arrows represent support while simple ones represent attacks. Furthermore, circles with solid border represent prima-facie arguments while dashed border ones represent ordinary ones. That is, “Tweety is a penguin” is considered a prima-facie argument that supports that “Tweety is a bird” which, in its turn, supports that “Tweety can fly.” The latter is then considered also prima-facie, that is, true unless proven otherwise. Note that “Tweety is a penguin” also attacks that “Tweety can fly”, so the latter cannot be accepted as true. Formally, this corresponds to the framework $\mathbf{EF}_1 ={\langle \mathbf{A},\mathbf{R}_a,\mathbf{R}_s,\mathbf{P} \rangle}$ with $\mathbf{R}_a = {\{ ({\{pT\}},fT) \}}$ and $\mathbf{R}_s = {\{ ({\{pT\}},bT), \, ({\{bT\}},fT) \}}$ and $\mathbf{P} = {\{ pT \}}$ whose unique admissible, complete, preferred and stable extension is ${\{pT , bT\}}$ . In other words, we conclude that “Tweety cannot fly.” Note that “Tweety is a penguin” provides conflicting evidence for whether it can fly or not. In EBAFs, this is solved by giving priority to the attack relation, so “Tweety cannot fly” is inferred.

3 Reasoning with contradictory evidence in equilibrium logic

In this section, we formalize principles NC and CW in constructive logic, obtaining as a result a formalism which is a conservative extension of logic programming under the answer set semantics (see Theorem 1 and Corollary 1 below) and which is capable of reasoning with contradictory evidence. We start by defining a new implication connective that captures 1 in terms of intuitionistic implication and strong negation:

(1) \begin{gather}\varphi_1 {\Rightarrow} \varphi_2 \ \ \underline{\underline{\mathrm{def}}} \ \ (\neg\!\sim\!\varphi_1 \wedge \varphi_1) \to \varphi_2.\end{gather}

Recall that intuitionistic implication $\varphi_1 \to \varphi_2$ can be informally understood as a means to construct a proof of the truth of the consequent $\varphi_2$ in terms of a proof of truth of the antecedent $\varphi_1$ . In this sense, (1) can be understood as a means to construct a proof of the truth of the consequent $\varphi_2$ in terms of proof of the truth of the antecedent $\varphi_1$ and the absence of a proof of its falsity, or in other words, in terms of a consistent proof of the antecedent $\varphi_1$ . It is easy to see that (1) is weaker than intuitionistic implication, that is, that

\begin{gather*}\varphi_1 \to \varphi_2 \ \models^+ \ \varphi_1 {\Rightarrow} \varphi_2,\end{gather*}

holds for every pair of formulas $\varphi_1$ and $\varphi_2$ . We can use the following simple example to illustrate the difference between intuitionistic implication and (1).

Example 2 Let $\Gamma_2$ be the following set of formulas:

\begin{align*}&a&&b&\sim\!&b&a &{\Rightarrow} c&b &{\Rightarrow} d,\end{align*}

and let $\Gamma_2'$ be the theory obtained by replacing each occurrence of implication ${\Rightarrow}$ by intuitionistic implication $\to$ . On the one hand, we have that both, $\Gamma_2$ and $\Gamma_2'$ , entail atoms a and c. On the other hand, we have: $\Gamma_2' \models^+ d$ but $\Gamma_2 \not\models^+ d$ . This is in accordance with NC, since the only way to obtain a proof of d is in terms of b, for which we have contradictory evidence. Note also that an alternative proof of d could be obtained if new consistent evidence becomes available: for the theory $\Gamma_3 = \Gamma_2 \cup {\{ a {\Rightarrow} d\}}$ we obtain $\Gamma_3 \models^+ d$ . It is also worth highlighting that, in contrast with intuitionistic implication, this new connective (1) is not monotonic: for $\Gamma_4 = {\{ b ,\ b {\Rightarrow} d\}}$ we have $\Gamma_4 \models^+ d$ and $\Gamma_4x \cup {\{ \sim\! b\}} \not\models^+ d$ . Obviously, it is not antimonotonic either: $\Gamma_4 \setminus {\{ b\}} \not\models^+ d$ .

The following result shows that, when dealing with consistent evidence, these differences disappear and (1) collapses into intuitionistic implication:

Proposition 4 Let ${\mathcal{I}}$ be a consistent N-interpretation and let $\varphi_1$ and $\varphi_2$ be any pair of formulas. Then, ${\mathcal{I}} \models^+ \varphi_1 {\Rightarrow} \varphi_2$ iff ${\mathcal{I}} \models^+ \varphi_1 \to \varphi_2$ .

Let us now formalize the CW assumption. As usual nonmonotonicity is obtained by considering equilibrium models (Definition 2). However, to capture CW, we need to restrict the consequences of these models to those that are consistent. We do so by introducing a new cw-inference relation which, precisely, restricts the consequences of $\models^+$ to those which are consistent:

(2) \begin{gather}{\mathcal{I}}, w \models \varphi \quad\text{iff}\quad{\mathcal{I}}, w \models^+ \neg\!\sim\! \varphi \wedge \varphi.\end{gather}

Furthermore, as usual, we write ${\mathcal{I}} \models \varphi$ iff ${\mathcal{I}}, w \models \varphi$ for all $w \in W$ . We also write $\Gamma \models \varphi$ iff ${\mathcal{I}} \models \varphi$ holds for every equilibrium model ${\mathcal{I}}$ of $\Gamma$ . For instance, in Example 3, it is easy to see that $\Gamma_2 \models^+ b$ and $\Gamma_2 \models^+ \sim\! b$ , but $\Gamma_2 \not\models b$ and $\Gamma_2 \not\models \sim\! b$ because the unique equilibrium model of $\Gamma_2$ contains contradictory evidence for b. On the other hand, as may be expected, when we deal with noncontradictory evidence cw-inference $\models$ just collapses to the regular inference relation $\models^+$ (see Proposition 5 below).

To finalize the formalization of CW, we also need to define default negation. This is accomplished by introducing a new connective ${not}\, $ and adding the following two items to the Nelson’s forcing relations:

\begin{eqnarray*}{l ,C, l}{\mathcal{I}}, w \models^+ {not}\, \varphi &\text{ iff }& {\mathcal{I}}, w \models^+ \neg \varphi \vee (\varphi \wedge \sim\!\varphi),\\{\mathcal{I}}, w \models^- {not}\, \varphi &\text{ iff }& {\mathcal{I}}, w \models^+ \varphi \text{ and } {\mathcal{I}}, w \not\models^- \varphi.\end{eqnarray*}

Then, an extended formula ${\varphi}$ is defined using the following grammar:

\begin{equation*}{\varphi} \quad ::= \quad \bot \ \mid \ a \ \mid \ \sim\! {\varphi} \ \mid \ {not}\, \varphi \ \mid \ {\varphi} \wedge {\varphi} \ \mid \ {\varphi} \vee {\varphi} \ \mid \ {\varphi} \to {\varphi},\end{equation*}

with $a \in At$ an atom. The following result shows that cw-inference and default negation are conservative extensions of the satisfaction relation $\models^+$ and HT-negation $\neg$ when restricted to consistent knowledge.

Proposition 5 Let ${\mathcal{I}}$ be a consistent N-interpretation and $\varphi$ be any extended formula. Then, the following conditions hold:

  1. i. ${\mathcal{I}} \models \varphi$ iff ${\mathcal{I}} \models^+ \varphi$

  2. ii. ${\mathcal{I}} \models {not}\, \varphi$ iff ${\mathcal{I}} \models \neg \varphi$ .

Despite the relation between default negation ${not}\, $ and HT-negation $\neg$ on consistent interpretations, in general, they no not coincide. The following example illustrates the difference between these two kinds of negations:

Example 3 Let $\Gamma_5$ be the following theory:

\begin{gather*}a \sim\! a {not}\, \sim\! a {\Rightarrow} b.\end{gather*}

This theory has a unique equilibrium model ${\mathcal{I}} = {\langle {\mathbf{T}},{\mathbf{T}} \rangle}$ with ${\mathbf{T}} = {\{ a , \sim\! a, b \}}$ . Note that, every model ${\mathcal{J}}$ of $\Gamma_5$ must satisfy ${\mathcal{J}} \models^+ a \wedge \sim\! a$ and, thus, it must also satisfy ${\mathcal{J}} \models {not}\, \sim\! a$ and ${\mathcal{J}} \models^+ b$ follows (Proposition 6). Hence, ${\mathcal{I}}$ is a $\leq$ -minimal model and, thus, an equilibrium model. On the other hand, let $\Gamma_6$ be the theory:

\begin{gather*}a \sim\! a \neg\!\sim\! a {\Rightarrow} b.\end{gather*}

In this case, we can check that ${\mathcal{J}} = {\langle {\mathbf{H}},{\mathbf{T}} \rangle}$ with ${\mathbf{H}} = {\{ a, \sim\! a\}}$ is a model of $\Gamma_6$ because ${\mathcal{J}} \not\models \neg\!\sim\! a$ and, thus, now ${\mathcal{I}}$ is not an equilibrium model. In fact, ${\langle {\mathbf{H}},{\mathbf{H}} \rangle}$ is the unique equilibrium model of $\Gamma_6$ .

The following result shows the relation between default negation, implication, and cw-inference.

Proposition 6 Let ${\mathcal{I}}$ be any N-interpretation and $\varphi$ be any formula. Then,

  1. i. ${\mathcal{I}} \models \varphi$ and ${\mathcal{I}} \models^+ \varphi {\Rightarrow} \psi$ implies ${\mathcal{I}} \models^+ \psi$ ,

  2. ii. ${\mathcal{I}} \models {not}\, \varphi$ implies ${\mathcal{I}} \not\models \varphi$ .

Furthermore, if ${\mathcal{I}}$ is a total HT-interpretation, then

  1. (iii) ${\mathcal{I}} \models {not}\, \varphi$ iff ${\mathcal{I}} \not\models \varphi$ .

Condition (i) formalizes a kind of modus ponens for ${\Rightarrow}$ in the sense that, if the we have a consistent proof of the antecedent, then we have a (possibly inconsistent) proof of the consequent. It is clear that this statement cannot be strengthened to provide a consistent proof of the consequent because any other formula could provide the contradictory evidence to make it inconsistent. Note also that this relation is nonmonotonic as adding new information may result in a contradictory antecedent. Condition (iii) formalizes the CW assumption, that is, ${not}\, \varphi$ holds whenever $\varphi$ is not known to be true or we have contradictory evidence for it. Note that, according to this, the default negation of an inconsistent formula is true and, therefore, the evaluation of default negation itself is always consistent (even if the formula is inconsistent): that is, ${\mathcal{I}}, w \not\models^+ {not}\, \varphi$ or ${\mathcal{I}}, w \not\models^- {not}\, \varphi$ holds for any extended formula.

On the contrary that implication ${\Rightarrow}$ , default negation ${not}\, $ cannot be straightforwardly defined Footnote 3 in terms of Nelson’s connectives.

Another alternative, we have investigated was defining ${not}\, \varphi$ as and $\neg \varphi \vee (\varphi \wedge \sim\!\varphi)$ . in terms of cw-inference. The following result shades light on this attempt.

Proposition 7 Let ${\mathcal{I}}$ be any N-interpretation and $\varphi$ be any formula. Then, ${\mathcal{I}} \models \neg \varphi \vee (\varphi \wedge \sim\!\varphi)$ iff ${\mathcal{I}} \models \neg \varphi$ .

That is, in terms of cw-inference, $\neg \varphi \vee (\varphi \wedge \sim\!\varphi)$ is equivalent to HT-negation. As illustrated by Example 3, default negation and HT-negation do not behave in the same way.

The following example illustrates that, though default negation allows to derive new knowledge from contradictory information, it does not allow to self justify a contradiction.

Example 4 Let $\Gamma_7$ be a logic program containing the following single rule:

(3) \begin{gather}{not}\, \sim\! a {\Rightarrow} a, \end{gather}

stating, as usual, that a holds by default. As expected this theory has a unique equilibrium model ${\mathcal{I}}$ which satisfies ${\mathcal{I}} \models a$ and ${\mathcal{I}} \not\models \sim\! a$ . Let now $\Gamma_8 = \Gamma_7 \cup {\{ \sim\! a\}}$ . This second theory also has a unique equilibrium model ${\mathcal{I}}$ which now satisfies ${\mathcal{I}} \models \sim\! a$ and ${\mathcal{I}} \not\models a$ . To see that ${\mathcal{J}} = {\langle {\mathbf{T}},{\mathbf{T}} \rangle}$ with ${\mathbf{T}} = {\{ a , \sim\! a \}}$ is not an equilibrium model of $\Gamma_8$ , let ${\mathcal{J}}' = {\langle {\mathbf{H}},{\mathbf{T}} \rangle}$ with ${\mathbf{H}} = {\{ \sim\! a \}}$ be an interpretation. Since ${\mathcal{J}}'$ satisfies ${\mathcal{J}}' < {\mathcal{J}}$ and it is a model of $\sim\! a$ , it only remains to be shown that ${\mathcal{J}}'$ is a model of (3). For that, just note ${\mathcal{J}} \models \sim\! a$ and, thus, ${\mathcal{J}} \not\models {not}\, \sim\! a$ follows by Proposition 6. This implies that ${\mathcal{J}}'$ satisfies (3) and, consequently, that ${\mathcal{J}}$ is not an equilibrium model. In fact, ${\langle {\mathbf{H}},{\mathbf{H}} \rangle}$ is the unique equilibrium model of $\Gamma_8$ .

3.1 A conservative extension of logic programming

Let us now consider the language formed with the set of logical connectives

\begin{equation*}\mathcal{C}_{LP} \underline{\underline{\mathrm{def}}} {\{ \bot, \sim\!, \wedge, \vee, {\Rightarrow}, {not}\, \}}.\end{equation*}

In other words, a $\mathcal{C}_{LP}$ -formula ${\varphi}$ is defined using the following grammar:

\begin{equation*}{\varphi} \quad ::= \quad \bot \ \mid \ a \ \mid \ \sim\! {\varphi} \ \mid \ {not}\, \varphi \ \mid \ {\varphi} \wedge {\varphi} \ \mid \ {\varphi} \vee {\varphi} \ \mid \ {\varphi} {\Rightarrow} {\varphi},\end{equation*}

with $a \in At$ being an atom. A $\mathcal{C}_{LP}$ -literal is either an explicit literal l or is default negation ${not}\, l$ . A $\mathcal{C}_{LP}$ -rule is a formula of the form $H {\Leftarrow} B$ where H is a disjunction of atoms and B is a conjunction of $\mathcal{C}_{LP}$ -literals. $\mathcal{C}_{LP}$ -theories and $\mathcal{C}_{LP}$ -programs are respectively defined as sets of $\mathcal{C}_{LP}$ -formulas and $\mathcal{C}_{LP}$ -rules. The definition of an answer set is applied straightforwardly as in Definition 1. Given any theory $\mathcal{C}_{LP}$ -theory $\Gamma$ , by $\mathcal{C}_{N}({\Gamma})$ we denote the result of

  1. 1. replacing every occurrence of ${\Rightarrow}$ by $\to$ and

  2. 2. and every occurrence of ${not}\, $ by $\neg$ .

Then, the following results follow directly from Propositions 4 and 5:

Theorem 1 Let $\Gamma$ be any $\mathcal{C}_{LP}$ -theory and ${\mathcal{I}}$ be any consistent interpretation. Then, ${\mathcal{I}}$ is an equilibrium model of $\Gamma$ iff ${\mathcal{I}}$ is an equilibrium model of $\mathcal{C}_{N}({\Gamma})$ .

Corollary 1 Let P be a $\mathcal{C}_{LP}$ -program and ${\mathbf{T}}$ be any consistent set of explicit literals. Then, ${\mathcal{I}} = {\langle {\mathbf{T}},{\mathbf{T}} \rangle}$ is an equilibrium model of P iff ${\mathbf{T}}$ is an answer set of P.

In other words, the equilibrium models semantics are a conservative extension of the answer set semantics. The following example shows the usual representation of the Tweety scenario in this logic (an alternative representation using contradictory evidence will be discussed in the Discussion section).

Example 5 (Ex. 1 continued) Consider again the Tweety scenario. The following logic program $P_9$ is a usual way of representing this scenario in LP:

(4) \begin{eqnarray}&{flyTweety} &{\Leftarrow}& {birdTweety} \wedge {not}\, \sim\!{flyTweety,} \end{eqnarray}
(5) \begin{eqnarray}&{birdTweety} &{\Leftarrow}& {penguinTweety,}\end{eqnarray}
(6) \begin{eqnarray}\sim\!\,&{flyTweety} &{\Leftarrow}& {penguinTweety}\end{eqnarray}
\begin{eqnarray}&{{penguinTweety,}} \notag\end{eqnarray}

where rule (4) formalizes the statement “birds normally can fly.” This is achieved by considering $\sim\!{flyTweety}$ as an exception to this rule. It can be checked that $P_{9}$ has a unique equilibrium model ${\mathcal{I}}_{9}$ , which is consistent, and which satisfies ${\mathcal{I}}_{9} \not\models {flyTweety}$ and ${\mathcal{I}}_{9} \models {not}\, {flyTweety}$ . In other words, Tweety cannot fly.

Example 6 (Ex. 3 continued) Consider now the theory obtained by replacing formulas $a {\Rightarrow} c$ and $b {\Rightarrow} d$ in $\Gamma_2$ by the following two formulas:

\begin{eqnarray*}{not}\, e \wedge a &{\Rightarrow}& c&{not}\, e \wedge b &{\Rightarrow}& d.\end{eqnarray*}

Let $\Gamma_10$ be such theory. It is easy to see that neither $\Gamma_10$ nor $\mathcal{C}_{N}({\Gamma_10})$ monotonically entail c nor d. This is due to the fact that the negation of e is not monotonically entailed: $\Gamma_10 \not\models^+ {not}\, e$ and $\mathcal{C}_{N}({\Gamma_10}) \not\models^+ \neg e$ . On the other hand, the negation of e is non-monotonically entailed in both cases: $\Gamma_10 \models {not}\, e$ and $\mathcal{C}_{N}({\Gamma_10}) \models \neg e$ . Note that both $\Gamma_10$ and $\mathcal{C}_{N}({\Gamma_10})$ have a unique equilibrium model, ${\mathcal{I}}_{10} = {\langle {\mathbf{T}},{\mathbf{T}} \rangle}$ and ${\mathcal{I}}_{10}' = {\langle {\mathbf{T}}',{\mathbf{T}}' \rangle}$ with ${\mathbf{T}} = {\{a,b ,\sim\! b, c\}}$ and ${\mathbf{T}}' = {\{a,b, \sim\! b, c, d\}}$ , respectively, and in both cases we have ${\mathcal{I}}_{10} \models {not}\, e$ and ${\mathcal{I}}_{10}' \models \neg e$ . As a result, we get that both theories cautiously entail c. However, as happened in Example 3, only $\mathcal{C}_{N}({\Gamma_10})$ cautiously entails d, because the unique evidence for d comes from b for which we have inconsistent evidence. This behavior is different from paraconsistent answer sets (Sakama and Inoue Reference Sakama and Inoue1995; Odintsov and Pearce Reference Odintsov and Pearce2005). As pointed out by Sakama and Inoue (Reference Sakama and Inoue1995), the truth of d is less credible than the truth of c, since d is derived through the contradictory fact b. In order to distinguish such two facts Sakama and Inoue (Reference Sakama and Inoue1995), also define suspicious answer sets which do not consider d as true. Footnote 4

This example also helps us to illustrate the strengthened closed world assumption principle CW. On the one hand, we have that $\Gamma_10 \models {not}\, e$ holds because there is no evidence for e. On the other hand, we have that $\Gamma_10 \models {not}\, b$ holds because we have contradictory evidence for b. Moreover, we have that $\Gamma_10 \models {not}\, d$ holds because the only evidence we have for d is based on the contradictory evidence for b.

4 Argumentation frameworks in equilibrium logic

In this section, we show how AFs, SETAFs, and EBAFs can be translated in this logic in a modular way and using only the object language. This translation is a formalization of the intuition of an attack stated in AT. Theorems 2, 3, and 4 show that the equilibrium models of this translation precisely characterize the stable extension of the corresponding framework.

4.1 Dung’s argumentation frameworks

Now, let us formalize the notion of attack introduced in AT, by defining the following connective:

(7) \begin{gather}\varphi_1 {\leadsto} \varphi_2 \ \ \underline{\underline{\mathrm{def}}} \ \ \varphi_1 {\Rightarrow} \sim\!\varphi_2.\end{gather}

Here, we identify the acceptability of $\varphi_1$ with having a consistent proof of it, or in other words, as having a proof of the truth of $\varphi_1$ and not having a proof of its falsity. Then, (7) states that the acceptability of $\varphi_1$ allows to construct a proof of the falsity of $\varphi_2$ . In this sense, we identify a proof of the falsity of $\varphi_2$ with $\varphi_2$ being defeated.

Proposition 8 Given any N-interpretation ${\mathcal{I}}$ and any pair of formulas $\varphi_1,\varphi_2$ , the following conditions hold:

  1. i. ${\mathcal{I}} \models \varphi_1$ and ${\mathcal{I}} \models^+ \varphi_1 {\leadsto} \varphi_2$ imply ${\mathcal{I}} \models^- \varphi_2$

Using the language $\mathcal{C}_{AF} = {\{{\leadsto}\}}$ , we can translate any AF as follows:

Definition 6 Given some framework $\mathbf{AF}= {\langle \mathbf{A},\mathbf{R} \rangle}$ , we define the theory:

(8) \begin{eqnarray}{ l ,C, l}\mathcal{C}_{AF}({\mathbf{AF}}) &\underline{\underline{\mathrm{def}}}& \mathbf{A} \cup {\{\ { a {\leadsto} b }\ \big|\ { (a,b) \in \mathbf{R} }\ \}}. \end{eqnarray}

In addition, we assign a corresponding set of arguments ${E_{\mathcal{I}}} \underline{\underline{\mathrm{def}}} {\{\ {\! a \in \mathbf{A} \!}\ \big|\ {\! {\mathcal{I}} \models a \!}\ \}}$ to every interpretation ${\mathcal{I}}$ .

Translation $\mathcal{C}_{AF}({\cdot})$ applies the notion of attack introduced in AT to translate an AF into a logical theory. The strengthened close world assumption CW is used to retrieve the arguments ${E_{\mathcal{I}}}$ corresponding to each stable model ${\mathcal{I}}$ of the logical theory obtained from this translation.

Example 7 To illustrate this translation, let $\mathbf{AF}_11$ be the framework corresponding to the following graph:

Then, we have that $\mathcal{C}_{AF}({\mathbf{AF}_11})$ is the theory containing the following two attacks:

\begin{gather*}a {\leadsto} b b {\leadsto} c,\end{gather*}

plus the facts ${\{a,b,c\}}$ .

Proposition 9 Let $\mathbf{AF}$ be some framework and ${\mathcal{I}}$ be some HT-model of $\mathcal{C}_{AF}({\mathbf{AF}})$ . Then, the following hold:

  1. i. f a is defeated w.r.t. ${E_{\mathcal{I}}}$ , then ${\mathcal{I}} \models^+ \sim\! a$

  2. ii. ${E_{\mathcal{I}}}$ is conflict-free.

If, in addition, ${\mathcal{I}}$ is an $\leq$ -minimal model, then

  • (iii) a is defeated w.r.t. ${E_{\mathcal{I}}}$ iff ${\mathcal{I}} \models^+ \sim\! a$ .

Example 8 (Ex. 7 continued) Continuing with our running example, let ${\mathcal{I}}_{11} = {\langle {\mathbf{T}}_{11},{\mathbf{T}}_{11} \rangle}$ and ${\mathcal{J}}_{11} = {\langle {\mathbf{T}}_{11}',{\mathbf{T}}_{11}' \rangle}$ be two total models of $\Gamma_{\textbf{AF}_{11}}$ with ${\mathbf{T}}_{11} = {\{a,b,c,\sim\! b\}}$ and ${\mathbf{T}}_{11}' = {\{a,b,c,\sim\! a,\sim\! c\}}$ . Then, we have that both $S_{{\mathcal{I}}_{11}} = {\{a,c\}}$ and $S_{{\mathcal{J}}_{11}} = {\{b\}}$ are conflict-free (though only $S_{{\mathcal{I}}_{11}}$ is stable). Furthermore, we also can see that argument b is the unique defeated argument w.r.t. $S_{{\mathcal{I}}_{11}}$ and the unique atom for which ${\mathcal{I}}_{11} \models^+ \sim\! b$ holds. On the other hand, we get that argument c is the unique defeated argument w.r.t. $S_{{\mathcal{J}}_{11}}$ and also both ${\mathcal{J}}_{11} \models^+ \sim\! a$ and ${\mathcal{J}}_{11} \models^+ \sim\! c$ hold. Note that, as stated by (iii) in Proposition 9, this implies that only $S_{{\mathcal{I}}_{11}}$ can be an equilibrium model. Let us show that it is indeed the case that ${\mathcal{J}}_{11}$ is not an equilibrium model and let us define, for that purpose, an interpretation ${\mathcal{J}}_{11}' = {\langle {\mathbf{H}}_{11}',{\mathbf{T}}_{11}' \rangle}$ with ${\mathbf{H}}_{11}' = {\mathbf{T}}_{11}' \setminus {\{ \sim\! a\}} = {\{a,b,c,\sim\! c\}}$ . In other words, interpretation ${\mathcal{J}}_{11}'$ is as ${\mathcal{J}}_{11}$ , but removing the non-defeated argument a as a negated conclusion $\sim\! a$ . It is easy to check that ${\mathcal{J}}_{11}' \models b {\leadsto} c$ because $\sim\! c \in {\mathbf{H}}_{11}'$ holds. Besides, since $\sim\! a \in {\mathbf{T}}_{11}'$ , we have that ${\mathcal{J}}_{11}' \not\models a$ and, therefore, that ${\mathcal{J}}_{11}' \models a {\leadsto} b$ . This implies that ${\mathcal{J}}_{11}'$ is a model of $\Gamma_{\textbf{AF}_{11}}$ . Since ${\mathcal{J}}_{11}' < {\mathcal{J}}_{11}$ , we get that ${\mathcal{J}}_{11}$ is not an equilibrium model.

In fact, we can generalize this correspondence between the stable extensions and the equilibrium models to any argumentation framework as stated by the following theorem:

Theorem 2 Given some $\mathbf{AF}= {\langle \mathbf{A},\mathbf{R} \rangle}$ , there is a one-to-one correspondence between its stable extensions and the equilibrium models of $\mathcal{C}_{AF}({\mathbf{AF}})$ such that

  1. i. if ${\mathcal{I}}$ is an equilibrium model of $\mathcal{C}_{AF}({\mathbf{AF}})$ , then ${E_{\mathcal{I}}}$ is a stable extension of $\mathbf{AF}$ ,

  2. ii. if E is a stable extension of $\mathbf{AF}$ and ${\mathcal{I}}$ is a total interpretation such that $T_{\mathcal{I}}^+ \!=\! \mathbf{A}$ and $T_{\mathcal{I}}^- \!=\! {Def}({E})$ , then ${\mathcal{I}}$ is an equilibrium model of $\mathcal{C}_{AF}({\mathbf{AF}})$ .

Proof sketch. Footnote 5 First, note that condition (i) follows directly from (iii) in Proposition 9 and the facts that (a) equilibrium models are $\leq$ -minimal models and (b) ${E_{\mathcal{I}}}$ is a stable extension iff ${E_{\mathcal{I}}}$ are exactly the non-defeated arguments w.r.t. ${E_{\mathcal{I}}}$ . To show (ii), it is easy to see that ${E_{\mathcal{I}}}$ being a stable extension implies that ${\mathcal{I}}$ is a model of $\mathcal{C}_{AF}({\mathbf{AF}})$ . Hence, to show that ${\mathcal{I}}$ is an equilibrium model what remains is to prove that any ${\mathcal{J}} < {\mathcal{I}}$ is not a model of $\mathcal{C}_{AF}({\mathbf{AF}})$ . Any such ${\mathcal{J}}$ must satisfy $H_{\mathcal{J}}^+ = H_{\mathcal{I}}^+ = \mathbf{A}$ and $H_{\mathcal{J}}^- \subset H_{\mathcal{I}}^- = T_{\mathcal{I}}^- = {Def}({E})$ . Therefore, there is some defeated argument such that $a \notin H_{\mathcal{J}}^-$ and some defeating attack $(b,a) \in \mathbf{R}_a$ such that $b \in E = H_{\mathcal{I}}^+ \setminus T_{\mathcal{I}}^- = H_{\mathcal{J}}^+ \setminus T_{\mathcal{J}}^-$ . This implies that $b {\leadsto} a \in \mathcal{C}_{AF}({\mathbf{AF}})$ and ${\mathcal{J}} \models b$ which, in its turn, implies that $a \in H_{\mathcal{J}}^-$ . This is a contradiction and, consequently, ${\mathcal{I}}$ is an equilibrium model.

Theorem 2 captures the relation between the stable extensions of an AF and its translation into a logical theory. As mentioned above, this relation relies on the reasoning principles AT and CW: An $\mathbf{AF}= {\langle \mathbf{A},\mathbf{R} \rangle}$ is translated into a logical theory $\mathcal{C}_{AF}({\mathbf{AF}})$ using the notion of attack introduced in AT. The stable extension ${E_{\mathcal{I}}}$ of this AF is then retrieved from the equilibrium model ${\mathcal{I}}$ of $\mathcal{C}_{AF}({\mathbf{AF}})$ using the CW principle.

4.2 Set attack argumentation frameworks

We may also extend the results of the previous section to SETAFs using the language $\mathcal{C}_{SF} = {\{{\leadsto}, \wedge\}}$ and a similar translation.

Definition 7 Given some finitary set attack framework $\mathbf{SF}= {\langle \mathbf{A},\mathbf{R}_a \rangle}$ , we define

(9) \begin{eqnarray}{ l ,C, l}\Gamma_{\mathbf{R}_a} &\underline{\underline{\mathrm{def}}}& {\Big\{\ { \bigwedge A {\leadsto} b }\ \ \Big|\ \ { (A,b) \in \mathbf{R}_a}\ \Big\}}, \end{eqnarray}

and $\mathcal{C}_{SF}({\mathbf{SF}}) \underline{\underline{\mathrm{def}}} \mathbf{A} \cup \Gamma_{\mathbf{R}_a}$ .

Similar to Definition 6, translation $\mathcal{C}_{SF}({\cdot})$ applies the notion of attack introduced in AT to translate an AF into a logical theory. In this case, the set of attacking arguments becomes a conjuntion in the antecedent of the attack connective.

Theorem 3 Given some finitary $\mathbf{SF}$ there is a one-to-one correspondence between its stable extensions and the equilibrium models of $\mathcal{C}_{SF}({\mathbf{SF}})$ such that

  1. i. if ${\mathcal{I}}$ is an equilibrium model of $\mathcal{C}_{SF}({\mathbf{SF}})$ , then ${E_{\mathcal{I}}}$ is a stable extension of $\mathbf{SF}$ ,

  2. ii. if E is a stable extension of $\mathbf{SF}$ and ${\mathcal{I}}$ is a total interpretation such that $T_{\mathcal{I}}^+ \!=\! \mathbf{A}$ and $T_{\mathcal{I}}^- \!=\! {Def}({E})$ , then ${\mathcal{I}}$ is an equilibrium model of $\mathcal{C}_{SF}({\mathbf{SF}})$ .

Proof sketch. The proof follows as in Theorem 2 by noting that any interpretation ${\mathcal{I}}$ and set of arguments B satisfy: $B \subseteq {E_{\mathcal{I}}}$ iff ${\mathcal{I}} \models b$ for all $b \in B$ iff ${\mathcal{I}} \models \bigwedge B$ .

4.3 Argumentation frameworks with evidence-based support

Let us now extend the language of SETAFs with the LP implication (1), in other words, we consider the language possessing the following set of connectives $\mathcal{C}_{EF} = {\{{\leadsto},\wedge,{\Rightarrow}\}}$ , so that we can translate any EBAF as follows:

Definition 8 Given any finitary evidence-based framework $\mathbf{EF}={\langle \mathbf{A},\mathbf{R}_a,\mathbf{R}_s,\mathbf{P} \rangle}$ , we define its corresponding theory as: $\mathcal{C}_{EF}({\mathbf{EF}}) \,\underline{\underline{\mathrm{def}}}\, \mathbf{P} \cup \Gamma_{\mathbf{R}_a} \cup \Gamma_{\mathbf{R}_s}$ with

(10) \begin{eqnarray}{ l ,C, l}\Gamma_{\mathbf{R}_s} &\underline{\underline{\mathrm{def}}}& {\Big\{\ { \bigwedge A {\Rightarrow} b }\ \ \Big|\ \ { (A,b) \in \mathbf{R}_s}\ \Big\}}, \end{eqnarray}

and $\Gamma_{\mathbf{R}_a}$ as stated in (9).

Note that, in contrast with AFs and SETAFs, the theory corresponding to an EBAFs do not contain all arguments as atoms, but only those that are prima-facie $\mathbf{P}$ . This reflects the fact that in EBAFs not all arguments can be accepted, but only those that are prima-facie or are supported by those prima-facie. Supports are represented using the LP implication ${\Rightarrow}$ and supported arguments are captured by the positive evaluation of each interpretation $H_{\mathcal{I}}^+$ . The following result extends Proposition 9 to EBAFs including the relation between supported arguments and models.

Proposition 10 Let $\mathbf{EF}$ be some framework and ${\mathcal{I}}$ be some HT-model of $\mathcal{C}_{EF}({\mathbf{EF}})$ . Then, the following hold:

  1. i. if a is supported w.r.t. ${E_{\mathcal{I}}}$ , then ${\mathcal{I}} \models^+ a$ ,

  2. ii. if a is defeated w.r.t. ${E_{\mathcal{I}}}$ , then ${\mathcal{I}} \models^+ \sim\! a$ ,

  3. iii. ${E_{\mathcal{I}}}$ is conflict-free.

If, in addition, ${\mathcal{I}}$ is an $\leq$ -minimal HT-model, then

  1. i. a is supported w.r.t. ${E_{\mathcal{I}}}$ iff ${\mathcal{I}} \models^+ a$ ,

  2. ii. a is defeated w.r.t. ${E_{\mathcal{I}}}$ iff ${\mathcal{I}} \models^+ \sim\! a$ ,

  3. iii. ${E_{\mathcal{I}}}$ is self-supporting.

Example 9 (Ex. 1 continued) Consider now framework $\mathbf{EF}$ representing the Tweety scenario.

(11) \begin{eqnarray}{birdTweety} &{\Rightarrow}& {flyTweety,} \end{eqnarray}
(12) \begin{eqnarray}{penguinTweety} &{\Rightarrow}& {birdTweety,}\end{eqnarray}
(13) \begin{eqnarray}{penguinTweety} &{\leadsto}& {flyTweety} \end{eqnarray}
\begin{eqnarray}&&{penguinTweety.} \notag&\end{eqnarray}

As mentioned in Example 1, framework $\textbf{EF}_{1}$ has a unique stable extension

\begin{equation*}{\{{penguinTweety},\, {birdTweety}\}},\end{equation*}

which does not include the argument flyTweety. In other words, Tweety cannot fly. Interestingly, $\mathcal{C}_{SF}({\textbf{EF}_{1}})$ has also a unique equilibrium model ${\mathcal{I}}_{12} = {\langle {\mathbf{T}}_{12},{\mathbf{T}}_{12} \rangle}$ where ${\mathbf{T}}_{12}$ stands for the set:

\begin{gather*}{\{ {penguinTweety},\, {birdTweety},\, {flyTweety},\, \sim\!{flyTweety}\}}.\end{gather*}

This equilibrium model precisely satisfies the two arguments in that stable extension: ${\mathcal{I}}_{12} \models {penguinTweety}$ and ${\mathcal{I}}_{12} \models {birdTweety}$ . Note that we get ${\mathcal{I}}_{12} \not\models {flyTweety}$ from the fact that ${\mathcal{I}}_{12} \models^+ \sim\!{flyTweety}$ . In fact, this correspondence holds for any EBAF as shown by Theorem 4 below. Though more technically complex, the proof of Theorem 4 is similar that those of Theorems 2 and 3. In particular, it is necessary to prove the following relation between equilibrium models and supportable arguments:

Proposition 11 Let $\mathbf{EF}$ be some framework and ${\mathcal{I}}$ be some equilibrium model of $\mathcal{C}_{EF}({\mathbf{EF}})$ . Then, the following statement holds:

  1. i. a is supportable w.r.t. ${E_{\mathcal{I}}}$ iff ${\mathcal{I}} \models^+ a$ .

In contrast with the results for supported arguments stated in Proposition 10, this property does not hold for arbitrary $\leq$ -minimal models. This fact can be illustrated by considering a simple $\mathbf{EF}_13$ such that $\mathcal{C}_{EF}({\textbf{EF}_{13}}) = {\{ a,\, a {\Rightarrow} b \}}$ . Let ${\mathcal{I}}_{13} = {\langle {\mathbf{H}}_{13},{\mathbf{T}}_{13} \rangle}$ be some interpretation with ${\mathbf{H}}_{13} = {\{ a \}}$ and ${\mathbf{T}}_{13} = {\{ a, \sim\! a \}}$ . It is easy to see that ${\mathcal{I}}_{13}$ is a $\leq$ -minimal model of $\mathcal{C}_{EF}({\textbf{EF}_{13}})$ , though it is not an equilibrium model (because it is not a total interpretation). It can also be checked that a is not defeated and, consequently, that b is supportable w.r.t. $E_{{\mathcal{I}}_{13}} = \varnothing$ . On the other hand, the unique equilibrium model of $\mathcal{C}_{EF}({\textbf{EF}_{13}})$ is ${\mathcal{J}}_{13} = {\langle {\mathbf{H}}'_{13},{\mathbf{T}}'_{13} \rangle}$ with ${\mathbf{H}}'_{13} = {\{ a, b \}}$ and ${\mathbf{T}}'_{13} = {\{ a, b \}}$ . Here, both a and b are supportable (and supported) w.r.t. $E_{{\mathcal{J}}_{13}} = {\{a,b\}}$ .

The following result shows that, indeed, this correspondence holds for any EBAF:

Theorem 4 Given some finitary $\mathbf{EF}$ , there is a one-to-one correspondence between its stable extensions and the equilibrium models of $\mathcal{C}_{EF}({\mathbf{EF}})$ such that

  1. i. if ${\mathcal{I}}$ is an equilibrium model of $\mathcal{C}_{EF}({\mathbf{EF}})$ , then ${E_{\mathcal{I}}}$ is a stable extension of $\mathbf{EF}$ ,

  2. ii. if E is a stable extension of $\mathbf{EF}$ and ${\mathcal{I}}$ is a total interpretation such that $T_{\mathcal{I}}^+ \!=\! {Sup}({E})$ and $T_{\mathcal{I}}^- \!=\! {Def}({E})$ , then ${\mathcal{I}}$ is an equilibrium model of $\mathcal{C}_{EF}({\mathbf{EF}})$ .

5 Translation of $\mathcal{C}_{LP}$ -program to regular programs

In this section, we show how $\mathcal{C}_{LP}$ -programs can be translated into regular ASP programs. An important practical consequence of this fact is that current state-of-the-art ASP solvers (Faber et al. Reference Faber, Pfeifer, Leone, Dell’Armi and Ielpa2008; Gebser et al. Reference Gebser, Kaufmann and Schaub2012) can be applied to $\mathcal{C}_{LP}$ -programs. Let us introduce such a translation as follows:

Definition 9 Given a $\mathcal{C}_{LP}$ -program P, by $\delta{P}$ we denote the result of

  1. 1. replacing every positive literal a in the body of a rule by $a \wedge \neg\!\sim\! a$ ,

  2. 2. replacing every negative literal ${not}\, a$ in the body of a rule by $\neg a \vee (a \!\wedge\! \sim\! a)$ ,

  3. 3. replacing all occurrences of ${\Rightarrow}$ by $\to$ .

Proposition 12 Any $\mathcal{C}_{LP}$ -program P and interpretation ${\mathcal{I}}$ satisfy: ${\mathcal{I}} \models^+ P$ iff ${\mathcal{I}} \models^+ \delta{P}$ .

Proposition 12 shows how we can translate any $\mathcal{C}_{LP}$ -program into an equivalent theory that does not use the new connectives ${not}\, $ and ${\Rightarrow}$ . The result of the translation in Definition 9 is almost a standard logic program, but for two points. First, strong negation has to be understood in a paraconsistent way, so an atom can be true and false at the same time. This can be addressed by using new auxiliary atoms to represent strongly negated atoms. Footnote 6 Second, step 2 introduces a disjunction in the body, which is not allowed in the standard syntax of logic programs. This can be addressed in polynomial-time also by using auxiliary atoms (similar to Tseitin Reference Tseitin1968). The following definition addresses these two issues.

Definition 10 Given a $\mathcal{C}_{LP}$ -program P, by $\tau{P}$ we denote the result of applying the following transformations to $\delta{P}$ :

  1. 1. replacing every explicit literal of the form $\sim\! a$ by a fresh atom $\tilde{a}$ ,

  2. 2. adding rules $a' \leftarrow \neg a$ and $a' \leftarrow a \wedge \tilde{a}$ for each atom $a \in At$ with a’ a new fresh atom, and

  3. 3. replacing each occurrence of $\neg a \vee (a \wedge \tilde{a})$ in the body of any rule by a’.

Given a total interpretation ${\mathcal{I}}$ , we also denote by $\tau{{\mathcal{I}}}$ an interpretation that, for every atom $a \in At$ , satisfies:

  1. 1. $\tau{{\mathcal{I}}} \not\models^- a$

  2. 2. $\tau{{\mathcal{I}}} \models^+ a$ iff ${\mathcal{I}} \models^+ a$

  3. 3. $\tau{{\mathcal{I}}} \models^+ \tilde{a}$ iff ${\mathcal{I}} \models^- a$

  4. 4. $\tau{{\mathcal{I}}} \models^+ a'$ iff either ${\mathcal{I}} \not\models^+ a$ or both ${\mathcal{I}} \models^+ a$ and ${\mathcal{I}} \models^- a$ .

Proposition 13 Any $\mathcal{C}_{LP}$ -program P and total interpretation ${\mathcal{I}}$ satisfy that ${\mathcal{I}}$ is an equilibrium model of P iff $\tau{I}$ an equilibrium model of $\tau{P}$ .

The result of Definition 10 is a standard logic program. Proposition 13 shows that we can use this translation in combination with standard ASP solvers to obtain equilibrium for $\mathcal{C}_{LP}$ -program and stable extensions of all the AFs considered in this paper. The second consequence of this translation is that deciding whether there exists any stable extension of some $\mathcal{C}_{LP}$ -program is in ${\mathrm{\Sigma}^{\mathrm{P}}_{{2}}}$ in general and in NP if the program is normal (Dantsin et al. Reference Dantsin, Eiter, Gottlob and Voronkov2001). This complexity results are tight because hardness follows from Corollary 1 and the hardness results for finding answer sets for these classes of programs (Dantsin et al. Reference Dantsin, Eiter, Gottlob and Voronkov2001). Therefore, deciding whether there exists any stable extension of some $\mathcal{C}_{LP}$ -program is ${\mathrm{\Sigma}^{\mathrm{P}}_{{2}}}$ -complete in general and NP-complete for normal $\mathcal{C}_{LP}$ -programs. Furthermore, this result directly applies to EBAFs so that deciding whether there exists any stable extension is NP-complete.

6 Discussion

LP and AFs are two well-established KRR formalisms for dealing with nonmonotonic reasoning (NMR). In particular, Answer Set Programming (ASP) is an LP paradigm, based on the stable model semantics, which has raised as a preeminent tool for practical NMR with applications in diverse areas of AI including planning, reasoning about actions, diagnosis, abduction and beyond (Baral Reference Baral2003; Brewka et al. Reference Brewka, Eiter and Truszczynski2011). On the other hand, one of the major reasons for the success of AFs is their ability to handle conflicts due to inconsistent information.

Here, we have shown that both formalisms can be successfully accommodated in Nelson’s constructive logic. In fact, it is easy to see that by rewriting attacks using definition (7), the translation of any AF becomes a normal $\mathcal{C}_{LP}$ -program. For instance, by rewriting the attack (13), we obtain the equivalent formula:

(14) \begin{eqnarray} {penguinTweety} &{\Rightarrow}& \sim\!{flyTweety,} \end{eqnarray}

which is a $\mathcal{C}_{LP}$ -rule. In this sense, we can consider $\mathcal{C}_{SF}({\textbf{EF}_{1}})$ in Example 9 as an alternative representation of the Tweety scenario in LP. Note that both the unique equilibrium model ${\mathcal{I}}_{9}$ of program $P_{9}$ (Example 5) and the unique equilibrium model ${\mathcal{I}}_{12}$ of this program satisfy:

\begin{eqnarray*}{l C l}{\mathcal{I}}_{9} \not\models {flyTweety}\\{\mathcal{I}}_{9} \models {not}\, {flyTweety}{\mathcal{I}}_{12} \not\models {flyTweety}\\{\mathcal{I}}_{12} \models {not}\, {flyTweety}\end{eqnarray*}

In other words, in both programs we conclude that Tweety cannot fly. However, there are a couple of differences between these two representations. First, in contrast with ${\mathcal{I}}_{9}$ , we have that ${\mathcal{I}}_{12}$ is not consistent: ${\mathcal{I}}_{12} \models^+ {flyTweety}$ and ${\mathcal{I}}_{12} \models^+ \sim\!{flyTweety}$ . Second and perhaps more interestingly, in $\mathcal{C}_{SF}({\textbf{EF}_{1}})$ , the “normality” of the statement “birds can fly” does not need to be explicitly represented. Instead, this normality is implicitly handled by the strong closed word assumption CW, which resolves the contradictory evidence for flyTweety by regarding it as false. In this sense, $\mathcal{C}_{LP}$ -programs and AFs can be seen as two different syntaxes of the same formalism based on the principles NC and CW highlighted in the introduction. In addition, another principle of this formalism is the fact that evidence must be founded or justified: this clearly shows up in normal LP and EBAFs where true literals can be computed by some recursive procedure, but also in Dung’s AFs where, as we have seen, defeat can be understood as a proof of falsity.

Regarding practical aspects, we can use $\mathcal{C}_{LP}$ -programs as a unifying formalism to deal with both logic programs and AFs. This directly allows to introduce variables in AFs through the use of grounding. Going further, full first-order characterizations of AFs can be provided by applying the same principles to first-order constructive logic (full first-order characterization of consistent logic programs has been already provided by Pearce and Valverde Reference Pearce and Valverde2004). Besides, constructive logic immediately provides an interpretation for other richer syntaxes like the use of disjunctive targets in Collective Argumentation (Bochman Reference Bochman2003) or the use of arbitrary propositional formulas to represent attacks in Abstract Dialectical Frameworks (Brewka and Woltran Reference Brewka and Woltran2010; Brewka et al. Reference Brewka, Strass, Ellmauthaler, Wallner and Woltran2013).

7 Conclusion and future work

We have formalized the principles NC and CW in Nelson’s constructive logic and shown that this is a conservative extension of logic programs which allow us to reason with contradictory evidence. Furthermore, this allows us to translate argumentation frameworks in a modular way and using the object language such that attacks and supports become connectives in logic using the object level. As a consequence, we can combine both formalisms in an unifying one and use proof methods from the logic or answer set solver to reason about it.

Regarding future work, an obvious open topic is to explore how other argumentation semantics can be translated into the logic. For instance, the relation between the complete semantics for AFs, three-valued stable models semantics for LP (Przymusinski Reference Przymusinski1991; Y. et al. 2009) and partial equilibrium logic (Cabalar et al. Reference Cabalar, Odintsov, Pearce and Valverde2007) suggest that our framework can be extended to cover other semantics such as the complete and preferred. Similarly, the relation between the paracoherent semantics for AFs (Amendola and Ricca Reference Amendola and Ricca2019) and semiequilibrium models (Amendola et al. Reference Amendola, Eiter, Fink, Leone and Moura2016) suggest a possible direction to capture this semantics using the object level. It will also be interesting to see the relation with the semistable semantics for AFs (Caminada et al. Reference Caminada, Carnielli and Dunne2012). The relation with other AFs extensions such as Collective Argumentation (Bochman Reference Bochman2003), Abstract Dialectical Frameworks (Brewka and Woltran Reference Brewka and Woltran2010; Brewka et al. Reference Brewka, Strass, Ellmauthaler, Wallner and Woltran2013) or Recursive Argumentation Frameworks (Barringer et al. Reference Barringer, Gabbay and Woods2005; Modgil Reference Modgil2009; Gabbay Reference Gabbay2009; Baroni et al. Reference Baroni, Cerutti, Giacomin and Guida2011; Cayrol et al. Reference Cayrol, Cohen and Lagasquie-Schiex2016; Cayrol et al. Reference Cayrol, Fandinno, FariÑas del Cerro and Lagasquie-Schiex2021) is also a direction worth exploring. Another important open questions are studying how the principles NC and CW stand in the context of paraconsistent logics (da Costa Reference da Costa1974) and paraconsistent logic programming (Blair and Subrahmanian Reference Blair and Subrahmanian1989); and studying the notion of strong equivalence (Lifschitz et al. Reference Lifschitz, Pearce and Valverde2001; Oikarinen and Woltran Reference Oikarinen and Woltran2011) in this logic and evidence-based frameworks.

Competing interests

The authors declare none.

Footnotes

*

This work is an extended version of a paper presented at the Sixteenth International Conference on Principles of Knowledge Representation and Reasoning and entitled Constructive Logic Covers Argumentation and Logic Programming (Fandinno and Farinas del Cerro 2018). We are thankful to Seiki Akama, Pedro Cabalar, Marcelo Coniglio, David Pearce, Newton Peron, and AgustÍn Valverde for their suggestions and comments on earlier versions of this work. We also thank the anonymous reviewers of the Sixteenth International Conference on Principles of Knowledge Representation and Reasoning for their comments on a preliminary version of this work.

1 We denote by $\sim\! S \underline{\underline{\mathrm{def}}} {\{\ { \sim\! \varphi }\ \big|\ { \varphi \in S}\ \}}$ the of set strongly negated formulas of a given set S. Similarly, we also define $\neg S \underline{\underline{\mathrm{def}}} {\{\ { \neg \varphi }\ \big|\ { \varphi \in S}\ \}}$ .

2 For the sake of clarity, proofs of formal results are moving to an appendix.

3 It is still an open question whether it is definable in terms of Nelson’s connectives or not.

4 Suspicious answer sets are based on a 6-value lattice which add the values suspiciously true and suspiciously false to the four values of $\textbf{N4}$ . In the unique suspicious answer set of $\Gamma_10$ , atom d gets assigned the suspiciously true value instead the true value. A formal comparison with suspicious answer sets is left for future work.

5 This theorem is a particualr case of Theorem 3 below. Recall that full proofs are provided in the appendix.

6 In fact, modern solvers already allow the use of explicit negation and their implementation is done by using new auxiliary atoms to represent strongly negated atoms. However, solvers also include a constraint of the form $a \wedge \sim\! a \to \bot$ for every atom a. This would remove the non-consistent answer sets, something we have to avoid to obtain paraconsistent answer sets.

References

Akama, S. 1987. Constructive predicate logic with strong negation and model theory. Notre Dame Journal of Formal Logic 29, 1 (12), 1827.Google Scholar
Almukdad, A. and Nelson, D. 1984. Constructible falsity and inexact predicates. The Journal of Symbolic Logic 49, 1, 231233.CrossRefGoogle Scholar
Amendola, G., Eiter, T., Fink, M., Leone, N. and Moura, J. 2016. Semi-equilibrium models for paracoherent answer set programs. Artificial Intelligence 234, 219271.CrossRefGoogle Scholar
Amendola, G. and Ricca, F. 2019. Paracoherent answer set semantics meets argumentation frameworks. Theory and Practice of Logic Programming 19, 5–6, 688704.CrossRefGoogle Scholar
Amgoud, L., Cayrol, C. and Lagasquie-Schiex, M.-C. 2004. On the bipolarity in argumentation frameworks. In NMR 2004, Proceedings, J. P. Delgrande and T. Schaub, Eds. 19.Google Scholar
Arieli, O. and Caminada, M. W. 2013. A qbf-based formalization of abstract argumentation semantics. Journal of Applied Logic 11, 2, 229252.CrossRefGoogle Scholar
Baral, C. 2003. Knowledge Representation, Reasoning and Declarative Problem Solving.CrossRefGoogle Scholar
Baroni, P., Cerutti, F., Giacomin, M. and Guida, G. 2011. AFRA: argumentation framework with recursive attacks. International Journal of Approximate Reasoning 52, 1, 1937.CrossRefGoogle Scholar
Barringer, H., Gabbay, D. and Woods, J. 2005. Temporal dynamics of support and attack networks: From argumentation to zoology. In Mechanizing Mathematical Reasoning. LNAI, vol. 2605. Springer Verlag, 5998.Google Scholar
Besnard, P. and Doutre, S. 2004. Checking the acceptability of a set of arguments. In 10th International Workshop on Non-Monotonic Reasoning (NMR 2004), Whistler, Canada, June 6–8, 2004, Proceedings, J. P. Delgrande and T. Schaub, Eds. 5964.Google Scholar
Besnard, P., Doutre, S. and Herzig, A. 2014. Encoding argument graphs in logic. In Information Processing and Management of Uncertainty in Knowledge-Based Systems - 15th International Conference, IPMU 2014, Montpellier, France, July 15-19, 2014, Proceedings, Part II, A. Laurent, O. Strauss, B. Bouchon-Meunier and R. R. Yager, Eds. Communications in Computer and Information Science, vol. 443. Springer, 345354.Google Scholar
Blair, H. and Subrahmanian, V. 1989. Paraconsistent logic programming. Theoretical Computer Science 68, 2, 135154.CrossRefGoogle Scholar
Bochman, A. 2003. Collective argumentation and disjunctive logic programming. Journal of Logic and Computation 13, 3, 405428.CrossRefGoogle Scholar
Brewka, G., Eiter, T. and Truszczynski, M. 2011. Answer set programming at a glance. Communications of the ACM 54, 12, 92103.CrossRefGoogle Scholar
Brewka, G., Strass, H., Ellmauthaler, S., Wallner, J. P. and Woltran, S. 2013. Abstract dialectical frameworks revisited. In IJCAI 2013, Proceedings, F. Rossi, Ed. IJCAI/AAAI, 803809.Google Scholar
Brewka, G. and Woltran, S. 2010. Abstract dialectical frameworks. In Principles of Knowledge Representation and Reasoning: Proceedings of the Twelfth International Conference, KR 2010, Toronto, Ontario, Canada, May 9-13, 2010, F. Lin, U. Sattler and M. Truszczynski, Eds. AAAI Press.Google Scholar
Cabalar, P., Fandinno, J., FariÑas del Cerro, L., Pearce, D. and Valverde, A. 2017. On the properties of atom definability and well-supportedness in logic programming. In EPIA 2017, Proceedings, E. C. Oliveira, J. Gama, Z. A. Vale and H. L. Cardoso, Eds. Springer, 624636.Google Scholar
Cabalar, P., Odintsov, S., Pearce, D. and Valverde, A. 2007. Partial equilibrium logic. Annals of Mathematics and Artificial Intelligence 50, 3–4, 305331.CrossRefGoogle Scholar
Caminada, M., Carnielli, W. and Dunne, P. 2012. Semi-stable semantics. Journal of Logic and Computation 22, 5, 12071254.CrossRefGoogle Scholar
Caminada, M., , S., AlcÂntara, J. and DvorÁk, W. 2015. On the equivalence between logic programming semantics and argumentation semantics. International Journal of Approximate Reasoning 58, 87111.CrossRefGoogle Scholar
Caminada, M. W. A. and Gabbay, D. M. 2009. A logical account of formal argumentation. Studia Logica 93, 2 (11), 109.CrossRefGoogle Scholar
Cayrol, C., Cohen, A. and Lagasquie-Schiex, M.-C. 2016. Towards a new framework for recursive interactions in abstract bipolar argumentation. In Proceedings of COMMA, 191198.Google Scholar
Cayrol, C., Fandinno, J., FariÑas del Cerro, L. and Lagasquie-Schiex, M. 2021. Valid attacks in argumentation frameworks with recursive attacks. Annals of Mathematics and Artificial Intelligence 89, 15737470.Google Scholar
Cayrol, C., Fandinno, J., FariÑas del Cerro, L. and Lagasquie-Schiex, M.-C. 2018. Argumentation Frameworks with Recursive Attacks and Evidence-Based Supports. Rapport de recherche IRIT/RR–2018–01–FR, IRIT, Universite Paul Sabatier, Toulouse. 1.Google Scholar
Cayrol, C., Fandinno, J., FariÑas del Cerro, L. and Lagasquie-Schiex, M.-C. 2018. Argumentation frameworks with recursive attacks and evidence-based supports. In FoIKS 2018, Proceedings.CrossRefGoogle Scholar
da Costa, N. 1974. On the theory of inconsistent formal systems. Notre Dame Journal of Formal Logic 15, 497510.Google Scholar
Dantsin, E., Eiter, T., Gottlob, G. and Voronkov, A. 2001. Complexity and expressive power of logic programming. ACM Computing Surveys 33, 3, 374425.CrossRefGoogle Scholar
Doutre, S., Herzig, A. and Perrussel, L. 2014. A dynamic logic framework for abstract argumentation. In Principles of Knowledge Representation and Reasoning: Proceedings of the Fourteenth International Conference, KR 2014, Vienna, Austria, July 20–24, 2014, C. Baral, G. D. Giacomo and T. Eiter, Eds. AAAI Press.Google Scholar
Dung, P. M. 1995. On the acceptability of arguments and its fundamental role in nonmonotonic reasoning, logic programming and n-person games. Artificial Intelligence 77, 2, 321358.CrossRefGoogle Scholar
DvorÁk, W., Gaggl, S. A., Linsbichler, T. and Wallner, J. P. 2015. Reduction-based approaches to implement modgil’s extended argumentation frameworks. In Advances in Knowledge Representation, Logic Programming, and Abstract Argumentation – Essays Dedicated to Gerhard Brewka on the Occasion of His 60th Birthday, T. Eiter, H. Strass, M. Truszczynski and S. Woltran, Eds. Lecture Notes in Computer Science, vol. 9060. Springer, 249264.Google Scholar
DvorÁk, W., Gaggl, S. A., Wallner, J. P. and Woltran, S. 2011. Making use of advances in answer-set programming for abstract argumentation systems. In Applications of Declarative Programming and Knowledge Management – 19th International Conference, INAP 2011, and 25th Workshop on Logic Programming, WLP 2011, Vienna, Austria, September 28–30, 2011, Revised Selected Papers, H. Tompits, S. Abreu, J. Oetsch, J. PÜhrer, D. Seipel, M. Umeda and A. Wolf, Eds. Lecture Notes in Computer Science, vol. 7773. Springer, 114133.Google Scholar
DvorÁk, W., Szeider, S. and Woltran, S. 2012. Abstract argumentation via monadic second order logic. In Scalable Uncertainty Management - 6th International Conference, SUM 2012, Marburg, Germany, September 17–19, 2012. Proceedings, E. HÜllermeier, S. Link, T. Fober and B. Seeger, Eds. Lecture Notes in Computer Science, vol. 7520. Springer, 8598.Google Scholar
Faber, W., Pfeifer, G., Leone, N., Dell’Armi, T. and Ielpa, G. 2008. Design and implementation of aggregate functions in the DLV system. Theory and Practice of Logic Programming 8, 5–6, 545580.CrossRefGoogle Scholar
Fandinno, J. and FariÑas del Cerro, L. 2018. Constructive logic covers argumentation and logic programming. In Principles of Knowledge Representation and Reasoning: Proceedings of the Sixteenth International Conference, KR 2018, Tempe, Arizona, 30 October - 2 November 2018, M. Thielscher, F. Toni and F. Wolter, Eds. AAAI Press, 128137.Google Scholar
Gabbay, D. and Gabbay, M. 2016. The attack as intuitionistic negation. Logic Journal of the IGPL 24, 5, 807837.CrossRefGoogle Scholar
Gabbay, D. M. 2009. Semantics for higher level attacks in extended argumentation frames part 1: Overview. Studia Logica 93, 2, 357.CrossRefGoogle Scholar
Gabbay, D. M. and Gabbay, M. 2015. The attack as strong negation, part i. Logic Journal of the IGPL 23, 881941.CrossRefGoogle Scholar
Gebser, M., Kaufmann, B. and Schaub, T. 2012. Conflict-driven answer set solving: From theory to practice. Artificial Intelligence 187–188, 5289.Google Scholar
Gelfond, M. and Lifschitz, V. 1988. The stable model semantics for logic programming. In Logic Programming: Proceedings of the Fifth International Conference and Symposium (Volume 2).Google Scholar
Gelfond, M. and Lifschitz, V. 1991. Classical negation in logic programs and disjunctive databases. New Generation Computing 9, 3/4, 365386.Google Scholar
Grossi, D. 2011. Argumentation in the view of modal logic. In Argumentation in Multi-Agent Systems, P. McBurney, I. Rahwan and S. Parsons, Eds. Springer Berlin Heidelberg, 190208.CrossRefGoogle Scholar
Gurevich, Y. 1977. Intuitionistic logic with strong negation. Studia Logica 36, 1–2, 4959.CrossRefGoogle Scholar
Kamide, N. and Wansing, H. 2015. Proof Theory of N4-Related Paraconsistent Logics. College Publications London.Google Scholar
Karacapilidis, N. and Papadias, D. 2001. Computer supported argumentation and collaborative decision making: The hermes system. Information Systems 26, 4, 259277.CrossRefGoogle Scholar
Lifschitz, V., Pearce, D. and Valverde, A. 2001. Strongly equivalent logic programs. ACM Transactions on Computational Logic 2, 4, 526541.CrossRefGoogle Scholar
LÓpez-Escobar, E. 1972. Refutability and elementary number theory. Indagationes Mathematicae (Proceedings) 75, 4, 362374.CrossRefGoogle Scholar
Markov, A. 1953. A constructive logic.Google Scholar
Modgil, S. 2009. Reasoning about preferences in argumentation frameworks. Artificial Intelligence 173, 9–10, 901934.CrossRefGoogle Scholar
Nelson, D. 1949. Constructible falsity. Journal of Symbolic Logic 14, 1 (03), 1626.CrossRefGoogle Scholar
Nelson, D. 1959. Negation and separation of concepts in constructive systems. Constructivity in Mathematics, 208225.Google Scholar
Nielsen, S. H. and Parsons, S. 2007. A generalization of Dung’s abstract framework for argumentation: Arguing with sets of attacking arguments. In Argumentation in Multi-Agent Systems, N. Maudet, Parsons, S. and I. Rahwan, Eds. Berlin, Heidelberg, 5473.Google Scholar
Nieves, J. C., CortÉs, U. and Osorio, M. 2008. Preferred extensions as stable models. TPLP 8, 4, 527543.Google Scholar
Odintsov, S. and Rybakov, V. 2015. Inference rules in Nelson’s logics, admissibility and weak admissibility. Logica Universalis 9, 1, 93120.CrossRefGoogle Scholar
Odintsov, S. P. 2005. The class of extensions of Nelson’s paraconsistent logic. Studia Logica 80, 2(8), 291320.CrossRefGoogle Scholar
Odintsov, S. P. and Pearce, D. 2005. Routley semantics for answer sets. In LPNMR 2005, Proceedings, C. Baral, G. Greco, N. Leone and G. Terracina, Eds. Springer, 343355.Google Scholar
Oikarinen, E. and Woltran, S. 2011. Characterizing strong equivalence for argumentation frameworks. Artificial Intelligence 175, 14–15, 19852009.CrossRefGoogle Scholar
Oren, N. and Norman, T. 2008. Semantics for evidence-based argumentation. In COMMA 2008, Proceedings., P. Besnard, S. Doutre and A. Hunter, Eds. 276284.Google Scholar
Osorio, M., PÉrez, J. A. N. and Arrazola, J. 2005. Safe beliefs for propositional theories. Annals of Pure and Applied Logic 134, 1, 6382.CrossRefGoogle Scholar
Pearce, D. 1996. A new logical characterisation of stable models and answer sets. In NMELP 1996, Selected Papers, J. Dix, L. M. Pereira and T. C. Przymusinski, Eds. Springer, 5770.Google Scholar
Pearce, D. 2006. Equilibrium logic. Annals of Mathematics and Artificial Intelligence 47, 1–2, 341.CrossRefGoogle Scholar
Pearce, D. and Valverde, A. 2004. Towards a first order equilibrium logic for nonmonotonic reasoning. In JELIA 2004, Proceedings, J. J. Alferes and J. A. Leite, Eds. Lecture Notes in Computer Science, vol. 3229. Springer, 147160.Google Scholar
Polberg, S. and Oren, N. 2014. Revisiting support in abstract argumentation systems. Tech. rep., TU Wien, Institut for Informatics.Google Scholar
Przymusinski, T. 1991. Three-valued nonmonotonic formalisms and semantics of logic programs. Artificial Intelligence 49, 1–3, 309343.CrossRefGoogle Scholar
Rasiowa, H. 1969. N-lattices and constructive logic with strong negation.Google Scholar
Reiter, R. 1980. A logic for default reasoning. Artificial Intelligence 13, 1–2, 81132.CrossRefGoogle Scholar
Routley, R. 1974. Semantical analyses of propositional systems of Fitch and Nelson. Studia Logica 33, 3, 283298.CrossRefGoogle Scholar
Sakama, C. and Inoue, K. 1995. Paraconsistent stable semantics for extended disjunctive programs. Journal of Logic and Computation 5, 3, 265285.CrossRefGoogle Scholar
Thomason, R. H. 1969. A semantical study of constructible falsity. Mathematical Logic Quarterly 15, 16–18, 247257.CrossRefGoogle Scholar
Toni, F. and Sergot, M. 2011. Argumentation and Answer Set Programming. Springer Berlin Heidelberg, Berlin, Heidelberg, 164180.Google Scholar
Tseitin, G. 1968. On the complexity of derivation in the propositional calculus. Zapiski nauchnykh seminarov LOMI 8, 234259.Google Scholar
Van Gelder, A., Ross, K. A. and Schlipf, J. S. 1991. The well-founded semantics for general logic programs. Journal of the ACM (JACM) 38, 3, 619649.CrossRefGoogle Scholar
Verheij, B. 2003a. Deflog: On the logical interpretation of prima facie justified assumptions. Journal of Logic and Computation 13, 3, 319346.CrossRefGoogle Scholar
Verheij, B. 2003b. Deflog: On the logical interpretation of prima facie justified assumptions. Journal of Logic and Computation 13, 3, 319346.CrossRefGoogle Scholar
Vorob’ev, N. 1952. A constructive propositional calculus with strong negation. In Doklady Akademii Nauk SSR. Vol. 85. 465468.Google Scholar
Wu, Y. and Caminada, M. 2010. A labelling-based justification status of arguments.Google Scholar
Caminada, M. and Gabbay, D. 2009. Complete extensions in argumentation coincide with 3-valued stable models in logic programming. Studia Logica 93, 2–3, 383403.Google Scholar