Recent advances in computational models of natural argument

Recent Advances in Computational Modelsof Natural ArgumentChris Reed,1,* Floriana Grasso2,†

1School of Computing, University of Dundee, Dundee DD1 4HN, UK2Department of Computer Science, University of Liverpool, Peach Street,Liverpool L69 7ZF, UK

This article reviews recent advances in the interdisciplinary area lying between artificial intel-ligence and the theory of argumentation. The article has two distinct foci: first, examining theways in which argumentation has inspired new models of logical and computational intelli-gence, and second, exploring how AI techniques have been used and extended to model andhandle real world argument in a wide variety of domains including law, education, medicine,and e-commerce. © 2007 Wiley Periodicals, Inc.

1. INTRODUCTION

Our aim here is to give a brief overview of current research in the interdisci-plinary area lying between artificial intelligence ~AI! and the theory of argumen-tation. We begin with a summary of the background with pointers to earlier reviews,and then in that context, sketch the current landscape. The objective is to provide abackdrop against which the articles in this special issue might take center stage.

There are two distinct ways in which AI has developed systems involvingargumentation. The first is in using concepts and intuitions about argument to inspireand provide foundations for the development of formal systems ~and often, specif-ically, formal nonclassical logics!. This approach we might term modeling withargument. A second approach is to construct models that reflect aspects of, orabstractions of, real-world practices of argumentation between humans. Thisapproach we might term modeling of argument. In occasional places, the two comeclose, but for the most part, given research projects fit squarely into one approachor the other.

*Author to whom all correspondence should be addressed: e-mail: [email protected].

†e-mail: [email protected]

INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS, VOL. 22, 1–15 ~2007!© 2007 Wiley Periodicals, Inc. Published online in Wiley InterScience~www.interscience.wiley.com!. • DOI 10.1002/int.20187

2. MODELING WITH ARGUMENT

2.1. History

Traditional symbolic AI models of reasoning are typically founded on first-order predicate calculus or some subset thereof. Such first-order reasoning sys-tems, however, are obliged to make a number of assumptions: that a given problemis fully specified ~such that the solution to a problem lies within the closure of thedatabase of clauses!; that the specification is consistent; and that any new factsthat are introduced are consistent with the specification and do not lead to retrac-tion of any propositions from it, that is, the accrual of new information is mono-tonic. Systems built on these assumptions are inadequate for dealing with situationsthat are incomplete, uncertain, or dynamic.

Of course, many real-world situations are indeed dynamic, with reasoningsystems building, perforce, uncertain and incomplete models of the world, dueprimarily to limited sensing abilities. To manipulate such representations of theworld, a variety of nonmonotonic reasoning techniques have been proposed ~foran introductory review, see, e.g., Ref. 1, pp. 195–229!. These techniques haveproved to be very successful and have become more numerous and more refinedas a result. There is, however, a relatively small number of key papers that, asantecedents, represent the main phyla of the current panoply of species of non-monotonic reasoning. Reiter’s2 default logic represents one of the first and intro-duces the concept of reasoning in the face of default or presumptive information;McDermott and Doyle’s3 nonmonotonic logic is the ancestor of modal interpreta-tions of such reasoning; Reiter’s4 reasoning with the Closed World Assumptiontackles one of AI’s oldest and thorniest problems, the frame problem,5 its imple-mentation as negation as failure6 forms a cornerstone of logic programming ingeneral and Prolog in particular, and a set of circumscription-based techniquesintroduced in Refs. 7 and 8 have more recently been used in components of a moregeneral solution to the frame problem in the context of “cognitive robotics.”9 Whatis remarkable is that each of these apparently different techniques ~and thereforetheir many modern descendants! have been brought together as specializations ofa formal framework described by its authors as an argument system.10

The theory is based on a logic programming style predicate logic language~where expressions are either atomic or rules of the form A1, . . . , Anr B! extendedwith the ] connective, from which are formed “nonmonotonic rules” ~the workdoes not assume truth functional semantics!. From a set of “deductive” and “non-monotonic” rules, it is then possible to define arguments as trees with just twolayers, at their single root a conclusion and at the leaves atomic facts or argumentssupporting the conclusion through the rules. Using just these concepts, which theyterm collectively an argument system, Lin and Shoham proceed to capture each ofdefault logic, nonmonotonic logic, negation as failure, and circumscription as exam-ples of argument systems ~Ref. 10, pp. 247–253!.

Lin and Shoham make no claims about—indeed, no mention of—the relation-ship between their abstract notion of an argument system and real argument; thelatter seems to have given brief insight at the work’s inception. In their definitionof an argument system, they include demands that all base facts be included, that

2 REED AND GRASSO

International Journal of Intelligent Systems DOI 10.1002/int

all deductively valid inferences be executed and included, and that inconsistencyis barred. Clearly, these are not generally features of real argumentation.

2.2. Development

Two later landmarks aimed to bring formal models of argumentation closer toreal-world practice. The first is Krause et al.’s11 logic of argumentation, LA, thatuses a labeled deductive system to support the construction of an argumentationtheorem prover, ATP. One feature of argumentation focused on in LA is the varietyof means by which arguments can be aggregated: With various arguments lendingsupport to or detracting from a proposition, there is a problem of how to combinethe information into an evaluation with respect to that proposition. Krause et al.define means of aggregation that work over arbitrary dictionaries of argumentstrength, including binary, multivalued, and probabilistic approaches. They extendtheir account to develop acceptability classes of arguments, from arbitrary well-formed arguments, through consistent arguments, then arguments with no rebut-ters, then arguments with no defeaters of any sort, and finally, logical tautologies.They then associate linguistic terms ~supported, plausible, probable, confirmed,certain! with these classes, in a drive to bring together cognitive science analysesand formal rigour.

Dung’s work12 can be seen as an extension to that of Lin and Shoham. Dung’snotion of an argumentation framework is similar in scope to that of a Lin andShoham argument system, but it does not specify the internal structure of an argu-ment ~rather, an argument is seen as synonymous with the conclusion it tries toestablish!, and it also introduces attacks—ordered pairs of arguments in which thefirst “represents an attack against” the second. Upon this foundation, Dung pro-ceeds to define two key concepts: first, the acceptability of an argument A withrespect to a set of arguments S—that for all counterarguments against A there arecounter-counterarguments in S, and second, the admissibility of a set of argumentsS—just in case each argument in S is acceptable.

The theory that Dung develops is then shown to be a powerful tool in address-ing an important class of problems in game theory. Finally, Dung claims thata range of nonmonotonic reasoning systems are in fact forms of argumentationand goes into some detail in the case of default logic, negation as failure, andPollock’s13–15 theory of defeasible reasoning. Like Pollock, Dung is motivated byeveryday human argument, claiming in his introduction that the “theory capturesnaturally the way humans argue to justify their solutions to many social problems”~p. 324!, and then that the work constitutes “a formal account of the principle ofargumentation.”

2.3. Recent Advances

There is a wide range of formal techniques founded upon these twoapproaches: Vreeswijk,16 Kowlaski and Toni,17 and Bondarenko et al.18 have beenparticularly influential, and two excellent reviews of recent work can be found in

COMPUTATIONAL MODELS OF NATURAL ARGUMENT 3


Chesñevar and Maguitman19 and Prakken and Vreeswijk.20 Three areas justify fur-ther brief mention here.

The first is the structuring of communication protocols for interaction betweensoftware agents. One of the earliest examples21 grew from the work on LA11 intro-ducing a distributed component, demonstrating how arguments ~in the LA sense!could be exchanged between agents that maintained discrete belief databases. Theacceptability criteria could then be operationalized by agents in order to judge ~atleast in part! whether incoming arguments should successfully change the agent’sbeliefs or not. Similarly, work has also extended the Dung approach, with Amgoudand Cayrol’s22 model of argument exchange building directly upon Dung’s notionof acceptability, and Kakas et al.23 using the Dung model as a foundation for pro-tocol specification. Various rhetorical notions ~such as the appeals ad ! have alsoinspired multiagent implementations, from Sycara’s early work24 founded on sim-ple intuitions about rhetotorical appeals, through to detailed accounts of persua-sion such as Refs. 25 and 26.

The general approach has been developed and extended to cover negotiationin Ref. 27. Argument-based negotiation, as distinguished from game-theoreticapproaches characterized by weightier assumptions and more specific, less flexi-ble communication, has become a key area of development.28–32 But as recogni-tion of Walton and Krabbe’s33 dialogue typology, and then Walton’s34 explicationof it, has started to filter through to multiagent systems research, the latter startedto introduce machinery for handling additional dialogue types and then for exploit-ing the very fact that different types of dialogue may be available. Thus Dignumet al.35 explore a number of different types of dialogue, including information-seeking and persuasion ~including Walton and Krabbe’s specialization, “rigorouspersuasion”!, Atkinson et al.36 focus on persuasion, McBurney et al.37 explore delib-eration, and Reed38 explores mechanisms by which dialogues of different typescan be bid and accepted and how one dialogue can be functionally embedded inanother. Indeed, the proliferation of research in the area led to a specification ofdesiderata for agent argumentation,39 and comparative work has been carried outby Norman et al.40 and McBurney and Parsons.41

One criticism of this approach is founded on a problem of verificationdescribed by Wooldridge.42 One of the fundamental tenets of agent theory is theautonomy of agent activity: it is an anathema to many researchers to require directaccess to agents’ internal states—such access should instead, it is argued, be medi-ated through the agent communication language. Unfortunately, if it is exactlythat communication language that is under investigation—particularly if the aimis to assess an agent’s compliance with the specification of such a language—thenusing the language itself to carry out the investigation leads to a regress. ~So, forexample, if a part of the specification of a communication language’s inform prim-itive is that an agent is sincere, then inquiring of an agent if it is, in fact, sinceremight elicit an inform in response—and there is no way of knowing if that informis sincere.! This conundrum has led to a proposal for the definition of agent com-munication languages built not on mental states, but ~at least partly! in terms ofexternally visible and verifiable commitments. Such a commitment-based approachhas examined extant commitment-based dialogue models43– 45 for means of

4 REED AND GRASSO


constructing conversation protocols. The leading proponent of the approach isSingh: in Ref. 46, he introduces a tripartite commitment structure, that in Ref. 47,inter alia, is developed and implemented. One of the aims of the work—to handlethe problem of verification—is then explored in more detail in Ref. 48. The flowof commitment change in agent communication has been explored using CTL* inRef. 49. A review of some of this work can be found in Ref. 50. Propositionalcommitment in this sense is increasingly a cornerstone of agent-based argumenta-tion, with traditional theories of dialectic systems being implemented more or lessdirectly. With McBurney’s desiderata for agent argumentation comes, therefore,the need for evaluative and comparative metrics for dialectic systems, which arejust starting to be explored.51,52

The second and emerging area of recent work is in using argument-basedtechniques for belief revision. So, for example, dynamic environments that can besensed require complex reasoning to maintain an accurate belief database, andCapobianco et al.53 use argument structures in a defeasible reasoner to achievethis. Argument-based protocols can be harnessed for structuring belief revision54

and argumentation internalized for solo-agent belief revision.55

Finally, a specific class of AI models of reasoning—those based on Bayesianprobability—are also being extended through argumentation structures. One ofthe most sophisticated is Vreeswijk’s56 Bayesian belief network into which hasbeen introduced Dungian style notions of argument, attack, and defeat. Das57 takesa similar tack, but founded upon LA. Gratton58 offers a probabilistic account spe-cifically of counterargumentation, and Saha and Sen59 use Bayesian belief modelsto execute decision making during the course of argumentation.

3. MODELING OF ARGUMENT

One key area of application of these nonmonotonic logics is in representinglegal reasoning, where defaults and defeasibility of rules have close analogies inthe law, and where there are established procedures for the relative prioritizationof conflicting rules. An early model of this kind is offered by Loui,60 in which aconcept of defeat between arguments is developed that closely matches legal prac-tice, involving directness, specificity, and preferences between inferences. It ishere too that formal models of burden of proof ~e.g., Ref. 61!, prima facie reason-ing,62 and case-based reasoning63 occur, complementing more pragmatic avenuesfollowed in argumentation theory typified by Ref. 64. A comprehensive review oflegal systems using models of argumentation is provided in Ref. 65.

At the less formal end of the spectrum, there are a variety of models andsystems that employ argumentation to structure and support interaction with a vari-ety of stored data. Such “knowledge engineering” has enjoyed significant success.One good review is offered in Ref. 66; here the focus is maintained on the majorlandmarks. One of the earliest analyses is in Birnbaum’s67 analysis of argumentmolecules, which touched briefly on many of the central issues in argumentationtheory, including the structure of argumentation schemes, of diagramming, and ofargument contexts. The similarly rich model of representation put forward in Ref. 68includes a variety of structures based on a mentalistic account of the writers and



readers of Comment pages in newspapers. Sillince and Minors69 provide yet anotherargument representation language, focusing on handling field-dependent argu-ment strength, providing what Krause et al.11 would regard as a data dictionary forevaluation. In some cases, argument-based knowledge engineering has been drivenby specific applications, with good examples in safety critical computing, whereBayesian models have enjoyed particular success.70,71

One problem with many of these approaches throughout AI is that they focusexclusively on the structure of argument, with a functional analysis of an argument’scomponents, and a means of evaluating or classifying argument parts and wholes.From an argumentation-theoretic point of view, they focus on the sort of argumentthat one offers or puts forward—not the sort of argument in which people engage.That is, in the terminology of O’Keefe,72 these models focus upon argument1, theargument “product,” ignoring the fact that arguments are also identified, in com-mon parlance, with a type of interaction—the process of argument2. Formal mod-els of the process of argument are common in argumentation theory as an upshotof research into the rules governing fallaciousness, with early work by Hamblin,43

Mackenzie,73 and Woods and Walton,74 and formal properties explored by Krabbe.44

In AI, process-oriented models of argument are much rarer. One notable excep-tion is Ref. 75, which employs the situation calculus5 to characterize contributionsfrom interlocutors. Within AI and law, the nature of the domain suggests a greateremphasis on dialogic models. Thus, Gordon,76 in his Pleading’s Game, includesturn taking between Plaintiff and Defendant as a fundamental component in a modelthat also integrates abductive reasoning and a defeasible interpretation of Toulmin-ian77 warrants. Prakken78 models the disputational status of claims as the dialogueproceeds, labeling moves as “in” or “out” according to dialogical rules ~in effect,he is implementing a substantial part of what Walton and Krabbe33 describe asstability adjustments!. AI models of the dynamic processes in argumentativeexchange have been termed computational dialectics, with early workshops on thetopic at American Association for Artificial Intelligence ’94 and Formal and AppliedPractical Reasoning ’96 exploring many of the themes that now characterize thefull range of computational models of, and with, argument.

Similar techniques have also been used for supporting argumentation in avariety of other domains. Gordon’s own subsequent work on the Zeno system79

offers a good example, but systems that offer generic support for various forms ofargument are quite widespread. One of the earliest is Matwin et al.’s80 Negoplan,which provided expert system support for negotiation in particular, and many morerecent systems have focused on argumentation as a mechanism for supporting andinteracting with predominantly human negotiation and decision making.81,82 Morerecent research been focused specifically on the online community, with applica-tions developed for e-democracy83,84 and online dispute resolution85 with popularimplemented and deployed solutions such as SmartSettle ~see www.smartsettle.com!. The gIBIS system86 was developed in an attempt to structure policy discus-sion ~or what Walton34 would probably term “deliberation”!, using Rittel andWebber’s87 IBIS information structures.

One novel emphasis in the gIBIS work is on diagrammatic presentation of theinformation to facilitate navigation, summary, and interaction with arguments in a

6 REED AND GRASSO


complex domain. Diagrammed arguments have been demonstrated to be usefultools for summarizing a range of topics88 and have an important role to play ineducation,89 so it is not surprising, therefore, that one very active area of recentresearch has explored computer-based models of argument diagram generation.The overview presented by Ref. 90 covers many recent developments. Examplesinclude Reason!Able91 aimed at teaching, Araucaria92 aimed at research and cor-pus mark-up, ArguMed93 aimed at legal argument, and ClaiMaker94,95 aimed atorganizing academic documents in a semantically rich network. As with all dia-gramming, one of the most important challenges is to determine the level of detailthat is included in a diagram and in the way that diagramming is carried out. Thereis a fundamental trade-off between, on the one hand, the complexity of the dia-gram ~and consequently of the interface required to produce it!, and on the other,the clarity of that diagram ~and simplicity of the interface!. Of course, greatercomplexity allows greater flexibility, expressiveness, and generality. For the mostpart, the designers of these systems have selected a particular point of trade-offbetween expressiveness and simplicity, determined in large part by the intendeduse and users.

One potential extension to these systems of diagramming that is particularlyliable to reduce clarity is the ability to handle dialogue. At the time of writing,there is no good method for diagramming complex dialogic argument, though workis underway in a number of areas to abstract from the detail of dialogue to provideuseful starting points for computational interpretation—a good example is Mann’s96

work on Dialogue Macrogame Theory, Krabbe’s97 work on Profiles of Dialogue,and the directions indicated by work based in monological models such as Refs. 90and 92. Further work at the boundaries between diagramming, AI, argumentationtheory, and discourse analysis is required to tackle the problems presented bydialogue.

Teaching ~or rather, learning! of both argumentation and domain-specific skillscan also be supported through dialogic models of argument. Thus Ref. 98 offersan early example of the use of a dialogue model in teaching in the medical domainand Ref. 99 in other science domains. Mackenzie’s100 DC system was adopted asthe framework, around which a combination of strategies were implemented in anattempt to ~partially! automate pedagogic interaction. This work was then sub-sequently extended to other domains in Ref. 101. DC’s shortcomings with respectto educational dialogue have been tackled in a series of papers arising from thatwork that develop a model of a derived system called DE.102 Finally, even duplic-itous argumentation can be used in support of educational goals103,104—thoughthe ethical implications of computer systems that argue untruthfully are complex.105

In yet other domains, more ad hoc models of argument are used as motivationand scaffold for computer-assisted learning systems: the CATO system in law,106

Cavilla-Sforza et al.’s107 tools for teaching science ~and, in particular, the dialec-tical nature of scientific development—they use paleontology as a case study!, theDREW system108 again for scientific discussion, and the mathematical proof expla-nation and hinting system P. Rex.109 Finally, argumentation is also employed ped-agogically in areas in which the textual form of presentation is crucial. Healtheducation is a canonical example. Grasso’s110,111 DAPHNE system, for example,



exploits argument schemas offered in Ref. 112 to persuade users to adopt healthiernutritional lifestyles; Reiter et al.’s113 STOP project follows a similar path, but issituated in a realistic domain, replete with length limitations and nontextual data,used to produce letters that are tailored to particular audiences to encourage themto stop smoking.

The emphasis on presentation of information is typical of systems that includea natural language generation component, and it is not surprising that ideas andtheories of argumentation have been brought to bear throughout the area. Naturallanguage processing offers a prime example of early research in AI that exploresargumentation. In a technical report, Kamp114 focuses on the problem of a logicalinterpretation of enthymemes—one that has taxed philosophers of argument forsome time ~see, e.g. Ref. 115!. Since then, computational systems for buildingarguments of one sort or another have been relatively common. Reichman’s116

model focused on a stack model of the shifts in topic during an argument, ratherlike the dynamics prescribed by logics of dialogue.44 A more product-oriented viewis put forward in Ref. 117, in which the structure of large arguments is used as abasis for determining the linguistic coherence of the textual expression of thosearguments. Maybury118 develops a plan-based account with individual plan oper-ators such as “convince” built on definitions of the mental states of the speakerand hearer. A hybrid of these approaches was put forward in Ref. 119 that builtCohen-like structures from Maybury-like plan operators. Elhadad120 takes a some-what different approach, based on Anscombre and Ducrot’s121 theory of argument,according to which the argument generation process is seen as one of managingtopoi-based and lexical-based constraints. Gilbert et al.122 construct a broad archi-tecture for argument-based HCI integrating both natural language understandingand generation part employing both operator composition and constraint manage-ment. Finally, Zukerman et al.123 demonstrate a model for producing “nice” argu-ments based on a Bayesian underpinning, and Carenini,124 following a similar path,based his GEA system on decision trees. Green’s model125 also uses a Bayesiancharacterization as a part of argument generation in the domain of genetic coun-seling. Finally, Carofiglio126 and Carofiglio and de Rosis103 combine a Bayesianunderpinning with Toulmin77 structures to tackle representation of uncertain infor-mation. All these various systems are founded on the assumption that argumenta-tion offers a relatively simple and intuitive means of presenting complexinformation at the human–computer interface.

One remarkable feature in this work has been a focus on structural compo-nents of argumentation in a highly logical style. Relatively little work has focusedupon the rich body of research in rhetoric and the heuristic structures and audience-centered approach developed there, including, as a prime example, Ref. 112. This,despite the fact that a very wide potential role has been described for “technolo-gies of persuasion” in AI and computer applications in general.127 Notable excep-tions to the trend include Grasso,110,128 who has worked at bringing rhetoricalconcepts and analyses to bear on operational problems in language generation,Bench-Capon,129,130 who introduces values into formal models of argument toaccount for persuasive structure, Ref. 119, in which a wide range of specificallyrhetorical moves are characterised as planning operators, and Ref. 105, in which

8 REED AND GRASSO


formal systems of context are harnessed to represent and reason about structuresof rhetoric. Tapping in to the emotional component or “mode” of argument131 issimilarly sparsely explored, with some tentative explorations in Refs. 132–135. Asimilarly surprising omission is the growing tradition of pragma-dialectics,136

which, as it is founded in speech act theory, might be expected to offer a good fitfor speech-act-based computational models of natural language processing.

There are, finally, three particularly striking outstanding issues that cut rightacross models of argument oriented toward knowledge engineering, toward natu-ral language processing, and toward agent communication. The first issue is therather narrow view taken of argument structure: Snoeck Henkemans137 points out,in a tradition following Freeman138 and Yanal139 and many others, that the identi-fication of “an argument,” and the means by which basic components can be com-posed into linked and convergent structures, are far from satisfactory. Many areaswithin AI might be well placed to contribute to this discussion. The second issueis the mechanism by which argumentation is driven. The goal machinery thatleads to arguments being automatically generated has been only briefly touchedupon,40,140,141 and yet is clearly fundamental to the endeavor. The challenge entailsspecific problems in architecture design, natural language production, knowledgerepresentation, practical reasoning, and so on. The third issue is the potential foruse of argumentation schemes.136,142 In almost all of the areas of AI in which argu-mentation has acted as a catalysis for development of new techniques andapproaches, there have been nascent concepts of stereotypical patterns of argu-ment, developed on an ad hoc and intuition-driven basis.143–145 As research effortwithin argumentation theory turns to such scheme-based reasoning, the potentialfor collaboration and cross-disciplinary utilization becomes much greater, withthe potential to develop in tandem both theoretical and implemented models ofargumentation schemes.

4. CONCLUDING REMARKS

This review offers a brief insight into the breadth of computational models ofargument, both in AI’s modeling with argument in formal systems, and its model-ing of argument in a more or less naturalistic style. The breadth of work coveredhere represents reasonably accurately the breadth of the workshop series in Com-putational Models of Natural Argument, which was hosted with the InternationalConference on Computational Science in its first year, and alternately at the Inter-national Joint Conference on Artificial Intelligence and the European Conferenceon Artificial Intelligence thereafter. The articles in this special issue represent resub-mitted and revised versions of a subset of the papers at the first three of thoseworkshops from 2001, 2002, and 2003, and naturally provide depth in a few ofthose areas.

References

1. Rich E, Knight K. Artificial intelligence, 2nd edition. New York: McGraw Hill; 1991.2. Reiter R. A logic for default reasoning. Artif Intell 1980;13:81–132.



3. McDermott DV, Doyle J. Nonmonotonic logic I. Artif Intell 1980;13:41–72.4. Reiter R. On closed world data bases. In: Gallaire H, Minker J, editors. Logic and data-

bases. New York: Plenum Press; 1978. pp 55–76.5. McCarthy J, Hayes P. Some philosophical problems from the standpoint of artificial intel-

ligence. In: Meltzer B, Michie D, editors. Machine intelligence, vol 4, Edinburgh: Edin-burgh University Press; 1969. pp 463–502.

6. Clark K. Negation as failure. In: Gallaire H, Minker J, editors. Logic and databases. NewYork: Plenum Press; 1978. pp 293–322.

7. McCarthy J. Applications of circumscription to formalizing commonsense knowledge. ArtifIntell 1986;28:89–118.

8. Lifschitz V. Pointwise circumscription. In: Ginsberg M, editor. Readings in nonmonotonicreasoning. San Francisco, CA: Morgan Kaufmann; 1987. pp 179–193.

9. Shanahan M. Solving the frame problem. Cambridge, MA: MIT Press; 1997.10. Lin F, Shoham Y. Argument systems: A uniform basis for nonmonotonic reasoning. In:

Proc First Int Conf on Knowledge Representation and Reasoning ~KRR’89!. San Fran-cisco, CA: Morgan Kaufmann; 1989. pp 245–255.

11. Krause P, Ambler S, Elvang-Gøransson M, Fox J. A logic of argumentation for reasoningunder uncertainty. Comput Intell 1995;11:113–131.

12. Dung PM. On the acceptability of arguments and its fundamental role in nonmonotonicreasoning, logic programming and n-person games. Artif Intell 1995;77:321–357.

13. Pollock JL. Defeasible reasoning. Cogn Sci 1987;11:481–518.14. Pollock JL. A theory of defeasible reasoning. Int J Intell Syst 1991;6:33–54.15. Pollock JL. Justification and defeat. Artif Intell 1994;67:377– 407.16. Vreeswijk GAW. Abstract argumentation systems. Artif Intell 1997;90:225–279.17. Kowlaski RA, Toni F. Abstract argumentation. Artif Intell Law 1996;4:275–296.18. Bondarenko A, Dung PM, Kowalski RA, Toni F. An abstract argumentation theoretic frame-

work for default reasoning. Artif Intell 1997;93:63–101.19. Chesñevar CI, Maguitman AG. Logical models of argument. ACM Comput Surv 2000;

32:337–383.20. Prakken H, Vreeswijk G. Logics for defeasible argumentation. In: Gabbay D, Guenthner

F, editors. Handbook of philosophical logic, vol 4, 2nd ed. Dordrecht: Kluwer; 2002.pp 218–319.

21. Parsons SD, Jennings NR. Negotiation through argumentation: A preliminary report. In:Proc 2nd Int Conf on Multi-Agent Systems ~ICMAS’96! Kyoto; 1996. pp 267–274.

22. Amgoud L, Cayrol C. A model of reasoning based on the production of acceptable argu-ments. Ann Math Artif Intell 2002;34:197–216.

23. Kakas A, Maudet N, Moraitis P. Layered strategies and protocols for argumentation-basedagent interaction. In: Rahwan I, Moraitis P, Reed C, editors. Argumentation in multi-agentsystems ~ArgMAS 2004!. Berlin: Springer; 2005. pp 64–77.

24. Sycara K. Persuasive argumentation in negotiation. Theor Decis 1990;28:203–242.25. Ramchurn SD, Jennings NR, Sierra C. Persuasive negotiation for autonomous agents: A

rhetorical approach. In: Working Notes of the 3rd Int Workshop on Computational Modelsof Natural Argument ~CMNA2003!, Acapulco, Mexico; 2003. pp 9–18.

26. Bentahar J, Moulin B, Chaib-draa B. Specifying and implementing a persuasion dia-logue game using commitments and arguments. In: Rahwan I, Moraitis P, Reed C, edi-tors. Argumentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005.pp 130–148.

27. Parsons SD, Sierra C, Jennings N. Agents that reason and negotiate by arguing. J LogicComput 1998;8:261–292.

28. Jennings NR, Parsons S, Noriega P, Sierra C. On argument-based negotiation. In: Proc IntWorkshop on Multi-Agent Systems, Boston; 1998. Available at: http://www.ecs.soton.ac.uk /;nrj/download-files/iwmas.ps.

29. Maudet N. Negotiating dialogue games. J Auton Agents Multi Agent Syst 2003;7:229–233.30. McBurney P, van Eijk RM, Parsons S, Amgoud L. A dialogue-game protocol for agent

purchase negotiations. J Auton Agents Multi Agent Syst 2003;7:235–273.

10 REED AND GRASSO


31. Rahwan I, Sonenberg L, McBurney P. Bargaining and argument-based negotiation: Somepreliminary comparisons. In: Rahwan I, Moraitis P, Reed C, editors. Argumentation inmulti-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005. pp 176–191.

32. Karuntillake NC, Jennings NR. Is it worth arguing? In: Rahwan I, Moraitis P, Reed C,editors. Argumentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005.pp 234–250.

33. Walton DN, Krabbe ECW. Commitment in dialogue. Albany, NY: SUNY Press; 1995.34. Walton DN. The new dialectic. Toronto: University of Toronto Press; 1998.35. Dignum FPM, Dunin-Keplicz B, Verbrugge R. Dialogue in team formation. In: Dignum F,

Greaves M, editors. Issues in agent communication. Berlin: Springer-Verlag; 2001.pp 264–280.

36. Atkinson K, Bench-Capon T, McBurney P. A dialogue game protocol for multi-agent argu-ment over proposals for action. In: Rahwan I, Moraitis P, Reed C, editors. Argumentationin multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005. pp 149–161.

37. McBurney P, Hitchcock D, Parsons S. The eightfold way of deliberation dialogue. Int JIntell Syst 2007;22:95–132.

38. Reed CA. Dialogue frames in agent communication. In: Proc 3rd Int Conf on Multi-AgentSystems ~ICMAS98!. Piscataway, NJ: IEEE Press; 1998. pp 246–253.

39. McBurney P, Parsons S, Wooldridge M. Desiderata for agent argumentation protocols. In:Castelfranchi C, Johnson WL, editors. In: Proc First Int Joint Conf on Autonomous Agentsand Multi-Agent Systems ~AAMAS’02!. New York: ACM Press; 2002. pp 402– 409.

40. Norman T, Carbogim D, Krabbe ECW, Walton D. Argument and multi-agent systems. In:Reed C, Norman TJ, editors. Argumentation machines. Norwell, MA: Kluwer; 2003.pp 15–54.

41. McBurney P, Parsons S. Dialogue games in multi-agent systems. Informal Logic 2002;22:257–274.

42. Wooldridge M. Verifiable semantics for agent communication languages. In: Demazeau Y,editor. In: Proc Third Int Conf on Multi-Agent Systems ~ICMAS 98!. Piscataway, NJ:IEEE Press; 1998. pp 349–356.

43. Hamblin CL. Fallacies. London: Methuen; 1970.44. Krabbe ECW. Formal systems of dialogue rules. Synthese 1985;63:295–328.45. Mackenzie JD. Four dialogue systems. Studia Logica 1990;49:567–583.46. Singh MP. A social semantics for agent communication languages. In: Proc IJCAI’99 Work-

shop on Agent Communication Languages. Berlin: Springer Verlag; 2000. pp 31– 45.47. Yolum P, Singh MP. Commitment machines. In: Proc 8th Int Workshop on Agent Theo-

ries, Architectures and Languages ~ATAL’01!. Berlin: Springer Verlag; 2002. pp 245–257.48. Venkatraman M, Singh MP. Verifying compliance with commitment protocols: Enabling

open web-based multi-agent systems. J Auton Agents Multi Agent Syst 1999;2:217–236.49. Bentahar J, Moulin B, Meyer JJCh, Chaib-draa B. A modal semantics for an argumentation-

based pragmatics for agent communication. In: Rahwan I, Moraitis P, Reed C, editors.Argumentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005. pp 44– 63.

50. Maudet N, Chaib-draa B. Commitment-based and dialogue-game based protocols—Newtrends in agent communication language. Knowl Eng 2002;17:157–179.

51. Parsons S, Wooldridge M, Amgoud L. Properties and complexity of some formal inter-agent dialogues. J Logic Comput 2003;13:347–376.

52. Wells S, Reed C. Formal dialectic specification. In: Rahwan I, Moraitis P, Reed C,editors. Argumentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005.pp 31– 43.

53. Capobianco M, Chesñevar CI, Simari GR. An argument-based framework to model anagent’s beliefs in a dynamic environment. In: Rahwan I, Moraitis P, Reed C, editors. Argu-mentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005. pp 95–110.

54. Malheiro W, Oliveira E. A reasoning model based on the production of acceptable argu-ments. In: Progress in Artificial Intelligence Knowledge Extraction, Multi-agent Sys-tems, Logic Programming, and Constraint Solving ~EPIA 2001!. Berlin: Springer; 2001.pp 205–218.



55. Paglieri F, Castelfranchi C. Revising beliefs through arguments: Bridging the gap be-tween argumentation and belief revision in MAS. In: Rahwan I, Moraitis P, Reed C, edi-tors. Argumentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005.pp 78–94.

56. Vreeswijk GAW. Argumentation in Bayesian belief networks. In: Rahwan I, Moraitis P,Reed C, editors. Argumentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer;2005. pp 111–129.

57. Das S. Logic of probabilistic arguments. In: Working Notes of the Second Int Workshopon Computational Models of Natural Argument ~CMNA 2002!, Lyon, France; 2002.pp 9–18.

58. Gratton C. Counterexamples and degrees of support. In: Working Notes of the Second IntWorkshop on Computational Models of Natural Argument ~CMNA 2002!, Lyon, France;2002. pp 49–54.

59. Saha S, Sen S. A Bayes net approach to argumentation based negotiation. In: Rahwan I,Moraitis P, Reed C, editors. Argumentation in multi-agent systems ~ArgMAS 2004!. Ber-lin: Springer; 2005. pp 208–222.

60. Loui R. Defeat among arguments. Comput Intell 1987;3:100–106.61. Farley AM, Freeman K. Burden of proof in a computational model of argumentation. In:

Eemeren FH, Van Grootendorst R, Blair JA, Willard CA, editors. In: Proc Third ISSAConf on Argumentation, vol III, Sicsat, Amsterdam; 1995. pp 410– 422.

62. Verheij B. DefLog: On the logical interpretation of prima facie justified assumptions. JLogic Comput 2003;13:319–346.

63. Skalak DB, Rissland EL. Argument moves in a rule-guided domain. In: Proc Int Conf onAI and Law ~ICAIL’91!, Oxford, UK; 1991. pp 1–11.

64. Walton DN. Legal argumentation and evidence. University Park, PA: Pennsylvania StateUniversity Press; 2002.

65. Bench-Capon T, Freeman JB, Hohmann H, Prakken H. Computational models, argumen-tation theories and legal practice. In: Reed C, Norman TJ, editors. Argumentation machines.Norwell, MA: Kluwer; 2003. pp 85–120.

66. Carbogim DV, Robertson D, Lee J. Argument-based applications to knowledge engineer-ing. Knowl Eng Rev 2000;15:119–149.

67. Birnbaum L. Argument molecules: A functional representation of argument structure. In:Proc Second Nat Conf on AI ~AAAI’82!, Pittsburgh, PA; 1982. pp 63– 65.

68. Alvarado SJ, Dyer MG, Flowers M. Argument representation for editorial text. KnowlBased Syst 1990;3:87–107.

69. Sillince JAA, Minors RH. Argumentation, self-inconsistency and multi-dimensional argu-ment strength. Commun Cogn 1992;25:325–338.

70. Fox J, Das S. Safe and sound: Artificial intelligence in hazardous applications. Cam-bridge, MA: MIT Press; 2000.

71. Gurr C. Argument representation for dependable computer-based systems. Informal Logic2002;22:293–321.

72. O’Keefe DJ. Two concepts of argument. J Am Forensic Assoc 1977;13:121–128.73. Mackenzie JD. The dialectics of logic. Logique Analyse 1981;24:159–177.74. Woods J, Walton DN. Arresting circles in formal dialogues. J Philos Logic 1978;7:73–90.75. Brewka G. Dynamic argument systems: A formal model of argumentation processes based

on situation calculus. J Logic Comput 2001;11:257–282.76. Gordon TF. Pleadings game: An artificial intelligence model of procedural justice. Bos-

ton, MA: Kluwer Academic Publishers; 1995.77. Toulmin SE. The uses of argument. Cambridge: Cambridge University Press; 1958.78. Prakken H. Relating protocols for dynamic dispute with logics for defeasible argumenta-

tion. Synthese 2001;127:187–219.79. Gordon TF, Karacapilidis NI. The Zeno argumentation framework. In: Proc Int Conf on AI

and LAW ~ICAIL’97!, Melbourne, Australia; 1997. pp 10–18.80. Matwin S, Szpakowicz S, Koperczak Z, Kersten GE, Michalowski W. Negoplan; An expert

system shell for negotiation support. IEEE Expert 1989;4:50– 62.

12 REED AND GRASSO


81. Girle R, Hitchcock D, McBurney P, Verheij B. Decision support for practical reasoning.In: Reed C, Norman TJ, editors. Argumentation machines. Norwell, MA: Kluwer; 2003.pp 55–84.

82. Prakken H, Vreeswijk G. Encoding schemes for a discourse support system for legalargument. In: Working Notes of the Second Int Workshop on Computational Models ofNatural Argument ~CMNA 2002!, Lyon, France; 2002. pp 31– 40.

83. Gordon TF, Richter G. Discourse support systems for deliberative democracy. In: Traun-müller R, Lenk K, editors. eGovernment: State of the art and perspectives ~EGOV!. Ber-lin: Springer Verlag; 2002. pp 248–255.

84. Atkinson K, Bench-Capon T, McBurney P. PARMENIDES: Facilitating democratic debate.In: Traunmüller R, editor. Electronic Government ~EGOV 2004!, Zaragoza, Spain; 2004.pp 313–316.

85. Lodder AR. A simple model to structure the information of parties in online ADR. In:Proc Seventh Int Conf on AI and Law, St. Louis, MO; 2001. pp 233–234.

86. Conklin J, Bergman ML. gIBIS: A hypertext tool for exploratory policy discussion. ACMTrans Office Inform Syst 1988;6:303–331.

87. Rittel HWJ, Webber MM. Dilemmas in a general theory of planning. Pol Sci 1973;4:155–169.

88. Horn RE. Using argumentation analysis to examine history and status of a major debatein artificial intelligence and philosophy. In: Proc Fourth Conf of the International Soci-ety for the Study of Argument ~ISSA’98! SicSat, Amsterdam; 1998. pp 375–381.

89. Brna P, Cox R, Good J. Learning to think and communicate with diagrams: 14 questionsto consider. Artif Intell Rev 2001;15:115–134.

90. Kirschner PA, Buckingham Shum SJ, Carr CS. Visualizing argumentation. Berlin: SpringerVerlag; 2003.

91. Van Gelder TJ. How to improve critical thinking using educational technology. In: Ken-nedy G, Keppell M, McNaught C, Petrovic T, editors. Proc 18th Conf of the Austral-asian Society for Computers in Learning in Tertiary Education, Melbourne; 2001.pp 539–548.

92. Reed C, Rowe GWA. Araucaria: Software for argument analysis, diagramming and rep-resentation. Int J AI Tools 2004;14:961–980.

93. Verheij B. Artificial argument assistants for defeasible argumentation. Artif Intell 2003;150:291–324.

94. Li G, Uren V, Motta E, Buckingham Shum S, Domingue J. ClaiMaker: Weaving a seman-tic web of research papers. In: Proc First Int Semantic Web Conf, Sardinia; 2002.pp 436– 441.

95. Buckingham Shum SJ, Uren V, Li G, Sereno B, Mancini C. Modeling naturalistic argu-mentation in research literatures: Representation and interaction design issues. Int J IntellSyst 2007;22:17– 47.

96. Mann WC. Dialogue macrogame theory. In: Proc SIGdial Workshop, Philadelphia; 2002.pp 129–141. See also www-rcf.esc.edu/;billmann/dialogue.

97. Krabbe ECW. Profiles of dialogue. In: Gerbrandy J, Marx M, de Rijke M, Venema Y,editors. JFAK: Essays dedicated to Johan van Benthem on the occasion of his 50th birth-day, vol III. Amsterdam: Amsterdam University Press; 1999. pp 25–36.

98. Pilkington RM, Hartley JR, Hintze D, Moore DJ. Learning to argue and arguing to learn:An interface for computer-based dialogue games. J Artif Intell Educ 1992;3:275–295.

99. Suthers D, Jones D. An architecture for intelligent collaborative educational systems. In:AI-Ed 97, the Eighth World Conf on Artificial Intelligence in Education, Kobe, Japan;1997. pp 55– 62.

100. Mackenzie JD. Question-begging in non-cumulative systems. J Philos Logic 1979;8:117–133.

101. Hobbs DJ, Moore DJ. Computational uses of philosophical dialogue theories. InformalLogic 1996;18:131–163.

102. Yuan T, Moore D, Grierson A. A human-computer debating system prototype and itsdialogue strategies. Int J Intell Syst 2007;22:133–156.



103. Carofiglio V, de Rosis F. Exploiting uncertainty and incomplete knowledge in deceptiveargumentation. In: Proc Workshop on Computational Models of Natural Language Argu-ment. In: Proc Int Conf on Computational Science ~ICCS’2001! Part 1, ~LNCS vol 2073!.Berlin: Springer Verlag; 2001. pp 1019–1028.

104. Sklar E, Parsons S, Davies M. When is it okay to lie? A simple model of contradiction inagent-based dialogues. In: Rahwan I, Moraitis P, Reed C, editors. Argumentation in multi-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005. pp 251–262.

105. Crosswhite J, Fox J, Reed C, Scaltsas T, Stumpf S. Computational models of rhetoricalargument. In: Reed C, Norman TJ, editors. Argumentation machines. Norwell, MA: Klu-wer; 2003. pp 175–210.

106. Aleven V, Ashley KD. An instructional environment for practising argumentation skills.In: Proc 12th National Conf on AI ~AAAI’94!, Seattle, WA; 1994. pp 485– 492.

107. Cavilla-Sforza V, Moore JD, Suthers DD. Helping students articulate, support and criti-cize scientific explanations. In: Proc World Conf on Artificial Intelligence in Education,Virginia; 1993. pp 113–120.

108. Baker M, Quignard M, Lund K. Designing a computer-supported collaborative learningsituation for broadening and deepening understanding of the space of debate. In: ProcFifth Int Conf of the International Society for the Study of Argument ~ISSA’02! SicSat,Amsterdam; 2002. pp 55– 62.

109. Fiedler A, Horacek H. Argumentation within deductive reasoning. Int J Intell Syst2007;22:49–70.

110. Grasso F. Exciting avocados and dull pears: Combining behavioural and argumentativetheory for producing effective advice. In: Proc 20th Annual Meeting of the CognitiveScience Society ~COG-SCI’98!, Wisconsin; 1998. pp 436– 441.

111. Grasso F, Cawsey A, Jones R. Dialectical argumentation to solve conflicts in advicegiving: A case study in the promotion of healthy nutrition. Int J Hum Comput Stud2000;53:1077–1115.

112. Perelman Ch, Olbrechts-Tyteca L. The new rhetoric. Notre Dame, IN: Notre Dame Press;1969.

113. Reiter E, Robertson R, Osman L. Lessons from a failure: Generating tailored smokingcessation letters. Artif Intell 2003;144:41–58.

114. Kamp H. Enthymemes. TM Series Report TM-4168 SDC. Santa Monica, CA: SystemDevelopment Corporation; 1969.

115. Hitchcock D. Enthymematic arguments. Informal Logic 1985;7:83–97.116. Reichman R. Getting computers to talk like you and me. Cambridge, MA: MIT Press;

1987.117. Cohen R. Analyzing the structure of argumentative discourse. Comput Ling 1987;

13:11–24.118. Maybury MT. Communicative acts for generating natural language arguments. In: Proc

Nat Conf on Artificial Intelligence ~AAAI’93!, Washington, DC; 1993. pp 357–364.119. Reed CA. The role of saliency in generating natural language arguments. In: Proc

16th Int Joint Conf on AI ~IJCAI’99!. San Francisco, CA: Morgan Kaufmann; 1999.pp 876–881.

120. Elhadad M. Using argumentation in text generation. J Pragmatics 1995;24:189–220.121. Anscombre JC, Ducrot O. L’argumentation dans la langue. Liège: Pierre Mardaga; 1983.122. Gilbert MA, Grasso F, Groarke L, Gurr C, Gerlofs JM. The persuasion machine. In:

Reed C, Norman TJ, editors. Argumentation machines. Norwell, MA: Kluwer; 2003.pp 121–174.

123. Zukerman I, McConachy R, Korb KB, Pickett D. Exploratory interaction with a Bayes-ian argumentation system. In: Dean T, editor. In: Proc 19th Int Joint Conf on AI ~IJCAI’99!.San Francisco, CA: Morgan Kaufmann; 1999. pp 1274–1279.

124. Carenini G. GEA: A complete, modular system for generating evaluative arguments. In:Proc Workshop on Computational Models of Natural Language Argument. In: Proc IntConf on Computational Science ~ICCS’2001! Part 1, ~LNCS vol 2073!. Berlin: SpringerVerlag; 2001. pp 959–968.

14 REED AND GRASSO


125. Green N. A study of argumentation in a causal probabilistic humanistic domain: Geneticcounseling. Int J Intell Syst 2007;22:71–93.

126. Carofiglio V. Modelling argumentation with belief networks. In: Working Notes of the4th Int Workshop on Computational Models of Natural Argument ~CMNA 2004!, Valen-cia, Spain; 2004. pp 22–24.

127. King P, Tester J. The landscape of persuasive technologies. Commun ACM 1999;42:31–38.128. Grasso F. Towards computational rhetoric. Informal Logic 2002;22:195–230.129. Bench-Capon TJM. Agreeing to differ: Modelling persuasive dialogue between parties

with different values. Informal Logic 2002;22:231–246.130. Bench-Capon T. Persuasion in practical argument using value based argumentation frame-

works. J Logic Comput 2003;13:429– 448.131. Gilbert MA. Coalescent argumentation. Mahwah, NJ: Lawrence Erlbaum Associates;

1997.132. de Rosis F, Grasso F. Affective natural language generation. In: Paiva A, editor. Affective

interactions. Berlin: Springer-Verlag; 2000. pp 204–218.133. Gilbert MA, Reed C. E-motion: Moving towards the utilization of artificial emotion.

Informal Logic 2002;22:275–292.134. Guerini M, Stock O, Zancanaro M. Persuasion models for intelligent interfaces. In: Work-

ing Notes of the Third Int Workshop on Computational Models of Natural Argument~CMNA2003!, Acapulco, Mexico; 2003. pp 25–30.

135. Carofiglio V, de Rosis F. In favour of cognitive models of emotions. Proc Workshop onMind-Minding Agents at AISB’05, Hatfield, UK, April 2005. pp 171–176.

136. Eemeren FH, van Grootendorst R. Argumentation communication and fallacies. Mah-wah, NJ: Erlbaum; 1992.

137. Snoeck Henkemanns AF. State of the art: The structure of argumentation. Argumentation2000;14:447– 473.

138. Freeman JB. Dialectics and the macrostructure of argument. Berlin: Foris; 1991.139. Yanal R. Dependent and independent reasons. Informal Logic 1991;13:137–144.140. Gilbert MA. Getting good value: Facts, values and goals in computational linguistics. In:

Proc Workshop on Computational Models of Natural Language Argument. In: Proc IntConf on Computational Science ~ICCS’2001! Part 1, ~LNCS vol 2073!. Berlin: SpringerVerlag; 2001. pp 989–998

141. Amgoud L, Kaci S. On the generation of bipolar goals in argumentation-based negotia-tion. In: Rahwan I, Moraitis P, Reed C, editors. Argumentation in multi-agent systems~ArgMAS 2004!. Berlin: Springer; 2005. pp 192–207.

142. Walton DN. Argumentation schemes for presumptive reasoning. Mahwah, NJ: LawrenceErlbaum Associates; 1996.

143. Bex F, Prakken H, Reed C, Walton D. Towards a formal account of reasoning aboutevidence: Argument schemes and generalisations. Artif Intell Law 2003;11:125–165.

144. Reed C, Walton D. Towards a formal and implemented model of argumentation schemesin agent communication. In: Rahwan I, Moraitis P, Reed C, editors. Argumentation inmulti-agent systems ~ArgMAS 2004!. Berlin: Springer; 2005. pp 19–30.

145. Verheij B. Dialectical argumentation with argumentation schemes: Towards a methodol-ogy for the investigation of argumentation schemes. In: van Eemeren FH, Blair JA, Wil-lard CA, Snoeck Henkemans F, editors. In: Proc Fifth Conf of the International Societyfor the Study of Argumentation ~ISSA 2002! Sic Sat, Amsterdam; 2003. pp 1033–1037.



Documents

Recent advances in computational models of natural argument