Persuasion dialogue in online dispute resolution in pdf/05 Walton Godden PD... · some remarks on how persuasion dialogues might be modelled is such a way as to allow them to be implemented

Persuasion dialogue in online dispute resolution

DOUGLAS WALTON and DAVID M. GODDENDepartment of Philosophy, University of Winnipeg, Winnipeg, Manitoba, R3B 2E9, Canada

E-mail: [email protected]

Abstract. In this paper we show how dialogue-based theories of argumentation can contribute to

the construction of effective systems of dispute resolution. Specifically we consider the role of

persuasion in online dispute resolution by showing how persuasion dialogues can be functionally

embedded in negotiation dialogues, and how negotiation dialogues can shift to persuasion dia-

logues. We conclude with some remarks on how persuasion dialogues might be modelled is such a

way as to allow them to be implemented in a mechanical or computerized system of dialogue or

dialogue management.

Key words: ADR, alternative dispute resolution, argument, argumentation, dialogue, negotiation,

ODR, online dispute resolution, persuasion, persuasion dialogue

Participants in the ICAIL 2003ODRWorkshop of June 28, 2003 in Edinburghexpressed the view that negotiation had dominated as the model for argu-mentation use in early work, and that persuasion dialogue should be stressedmore as an important model as well. This paper presents amodel of persuasiondialogue as an argumentation framework and shows how the model isimportant forODR.Specifically, this paper considers the questionof the role ofpersuasion dialogue in the ODR process, especially in relation to negotiationdialogue. That is, we consider how to model the functional embedding ofpersuasion dialogues within a negotiation dialogue, and under what circum-stances a negotiation dialogue might shift to a persuasion dialogue.

In the first section of this paper, we present a brief overview of ODR. Theaim here is not to provide a general survey of the historical or theoreticaldevelopment of ODR. Katsh and Rifkin (2001, ch. 2) have already provideda comprehensive historical overview of ODR, which will already be familiarto most readers. Rather, the point of this section is to briefly highlight a fewsalient features of ODR, which we feel demonstrate the need for a more in-depth investigation of the normative dialogue models which contribute todispute resolution. This leads to the next section where we note the preva-lence of the negotiation dialogue in ODR, and make some observations

Artificial Intelligence and Law (2006) � Springer 2006DOI 10.1007/s10506-006-9014-0

regarding how negotiation is frequently characterized in ODR (sect. 2). Thisis followed in section 3 by a brief overview of the basic types of dialoguesmodeled in argumentation theory, as well as a section describing the primaryfeatures of persuasion dialogues as they have been studied in argumentationtheory (sect. 4). In the fifth section, we consider a model of the dialogueprocess proposed by Mochol (2004) for use in ODR discourse models andsupport systems. Following our proposal of a modification to Mochol’smodel (sect. 6), we consider the occasions where persuasion dialogues mightbe functionally embedded in a negotiation dialogue, and how this functionalembedding could be represented in the model of the negotiation process (sect.7). In section 8 we discuss those circumstances under which a negotiationdialogue might shift to a persuasion dialogue. We conclude in section 9 withsome remarks on how persuasion dialogues might be modelled is such a wayas to allow them to be implemented in a mechanical or computerized systemof dialogue or dialogue management.

1. ODR overview

Online Dispute Resolution (ODR) is typically presented as a form ofAlternative Dispute Resolution (ADR) and as a means for resolving disputes.Disputes, in turn, are presented as a form of conflict (Rule 2002, p. 21). ODRis presented as a preemptive way of managing and resolving conflict before itreaches litigation (Rule 2002, p. 2). ODR occurs at the intersection of ADRand internet-based communication technology. ADR is based upon resolu-tion-oriented, normatively structured models of communication, allowing foran orderly, rational exchange of information and opinion. Clearly, then, theability of ADR to successfully produce resolutions requires successful com-munication, thus indicating the need for normatively sound models ofcommunication in ADR.

While ADR supplies the theoretical bases of ODR, computer technologyprovides an online environment affecting such factors as information (access,storage, and processing) and communication. Of primary interest for thepurposes of this paper, technology provides ODR with a means of commu-nication which is by-and-large faster, more versatile, and more accessiblethan other, previously available options. At its base, the role of technology –the so called ‘‘fourth party’’ in ODR – is to provide a structured commu-nication and information environment that can contribute to the efficient,effective resolution of disputes. These occur in a variety of basic frameworks.This can be part of arbitrated settlements whereby an arbitrator sets outhouse rules, or mediated exchanges where a mediator suggests some groundrules, or even situations where the communicative exchange is structured bythe technology as, for example, in automated negotiation, solution set

DOUGLAS WALTON AND DAVID M. GODDEN

databases, blind bidding and other forms of automated arbitration andfinally multivariable resolution optimization programs (Rule 2002, pp. 55–59). As such, just as with its parent ADR, it is crucial that the communica-tional models used in ODR are theoretically sound.

Yet, very little attention has been paid to the types of dialogues that canbe at work in ADR, the normative structures that characterize each of thesedialogues, and the procedural rules that govern them. By contributing to thedevelopment of effective models of these various types of dialogue, we hopethat argumentation theory can help to contribute to a sound theoretical basison which communication models used in ODR can be built.

2. ODR and negotiation dialogue

Typically, the primary type of dialogue considered by ADR has been seen asthe negotiation dialogue. Negotiation is seen as standing among conciliation,mediation, arbitration and even private judging as approaches to disputeresolution (e.g. Patterson and Seabolt 2001, pp. 11–17; Lodder and Bol 2004,p. 9), and as alternatives to litigation (Rule 2002, p. 2). Yet, negotiation isseen as the basis of many, if not most of these approaches to dispute reso-lution. Mediation, for example, is viewed as a kind of ‘‘assisted or facilitatednegotiation’’ (citing Leeson and Johnson 1988, p. 133; Patterson and Seabolt2001, p. 53), while arbitration is seen as a contract-based form of disputeresolution (Patterson and Seabolt 2001, p. 115). As a result, while arbitratedsettlements are not arrived at directly through negotiation, the decision toenter into arbitration is presented as the result of negotiation.

In taking its theoretical bases from ADR, ODR seems to have inherited anegotiation-based approach. But, there are other hereditary and culturalinfluences which might also explain this feature of ODR. The prevalence ofnegotiation-based approaches is perhaps explained by ODR’s origins in theliterature of business management and conflict-management, where negoti-ation is seen as the primary mode of much, if not most, business commin-ucation. It might also be due to the legal influence of seeing ODR as analternative to a litigious approach which is seen as adversarial and hencebetter modelled under a persuasion-based model. Lodder and Bol (2004)propose a negotiation-based model for ODR mainly because parties try tosettle their disputes on their own, without any third-party. Finally, ODRmight often be seen as a negotiation since the primary issue can often beviewed as one of dividing up some property, asset or resource in such a wayas to maximize the interests of the participants. As a result, the tendency hasbeen to model ODR on negotiation-based models while neglecting the rolethat other types of dialogue can have in the ODR process.

PERSUASION DIALOGUE IN ONLINE DISPUTE RESOLUTION

Generally in the ADR literature, negotiation is portrayed as a kind ofbargaining, and usually as a means of dividing up some set of scarce re-sources among the parties involved. Patterson and Seabolt (2001) offer thefollowing summary of types of negotiation and negotiators. There are twoapproaches to negotiation: distributive and interest-based. In distributive, orpositional, negotiation, the only goals considered are the explicit, short-termgoals of the parties to have a desired share of the scarce resource, and themaximum concession they are willing to make to the other side in getting it.Here, the negotiation process is seen largely as a series of trade-offs. Interest-based, or integrative, negotiation, on the other hand, involves an attempt touncover unstated, long-term goals that both (or all) parties might have incommon, or at least that might be more compatible than the opposed short-term goals. These can then be used as a means to propose alternative solu-tions aimed at allowing each party to achieve their long-term interests.Further, there are two styles of negotiating, or attitudes a party can taketowards negotiation: competitive and cooperative. Competitive negotiatorstend to see negotiation as an adversarial process, and seek to maximize theirown interests without concern for the satisfaction of the other participant(s).Cooperative negotiators, on the other hand, tend to see negotiation as acooperative process designed to facilitate the maximization of a mutuallyachievable goal set of the (two) parties involved in negotiation. They seek toachieve their most important goals while reaching a settlement that will allowa negotiating partner to achieve her main goals if at all possible (Pattersonand Seabolt 2001, pp. 26–30).

Following Craver (1994, p. 47ff), Patterson and Seabolt (2001, p. 30)describe the basic structure of the negotiation process has having the fol-lowing six phases.

I Preparation PhaseII Preliminary PhaseIII Information PhaseIV Competitive/Distributive PhaseV Closing PhaseVI Cooperative Phase

While these phases are presented as procedural, they are primarily de-scribed according to the kind of strategic manoeuvring that each party willengage in.

Another type of dialogue that has been central in the development ofODR is deliberation. In a deliberation dialogue, the goal is for the partici-pants to arrive at a decision on what to do (for instance to solvesome problem or adopt some policy), given the need to take action.Hitchcock et al. (2002) set out a formal model of deliberation dialogue in


which participants make proposals and counter-proposals on what to do.In this model (p. 5), the need to take action is expressed in the form of agoverning question like, ‘‘How should we respond to the prospect of globalwarming?’’ In subsequent stages of the dialogue, policies are put forwardas proposals for action, and the arguments for and against each proposalare considered. Finally, an attempt is made to close the dialogue by reach-ing agreement on what is taken to be the best policy.

1

From the point of view of argumentation theory, a principal theoreticalerror apparent in some of the ADR literature is the mischaracterization ofnegotiation as type of formal dialogue. There is a prevalent attitude, espe-cially in the literature of business and management, that negotiation is theprimary mode of business communication. Patterson and Seabolt, for in-stance, write that ‘‘People negotiate all the time. From two people decidingwhat movie to see, to the representative of labor and management bargainingover a labor contract’’ (2001, p. 21). Problematically, on prevailing models inargumentation theory, only the contract bargaining is an example of anegotiation dialogue. Two people deciding on what movie to see is a para-digmatic example of a deliberation dialogue and, while they often use thesame kinds of arguments, the normative structures of these two dialogues arequite different.

Similarly, Patterson and Seabolt (2001, p. 21) follow Goldberg et al.(1992, p. 17) in offering the following definition of ‘‘negotiation’’. Negotia-tion is ‘‘communication for the purpose of persuasion’’. This definition ishighly problematic, from the point of view of argumentation theory, becauseit mistakenly conflates two quite different types of dialogue: negotiationdialogues and persuasion dialogues. Dialogues of different types are struc-tured differently and have different goals and different structural, dialecticaland normative properties. Not only does this mean that the arguments inthese different types of dialogue must be evaluated differently. Moreimportantly for ODR it means that different types of dialogue must bemanaged by different procedural rules.

Importantly, ADR models of negotiation typically allow that argumen-tative dialogue can be embedded in the negotiation process, particularly inthe distributive phase. It is here that, for instance, that Patterson and Seaboltremark that ‘‘negotiators are most likely to use competitive tactics such asargument’’ (2001, p. 36). That is, on the ADR model, negotiation is notsimply a process of making offers and counter-offers. Rather, offers andcounter-offers can be supported with reasons, and the rejection of offers canbe defended with argument. As Patterson and Seabolt describe, ‘‘negotiatorswill take turns making offers and justifying why each offer is reasonable. Ifthe other side buys the argument, they will concede ground and come up (ordown) on their offer’’ (ibid.). This points to a theoretical need to be able tomodel how such argumentation is embedded in the negotiation process.


Before addressing this question, it is worthwhile to briefly inventory the typesof dialogue commonly treated in argumentation theory.

3. The classification of types of dialogue

Having considered the prevailing dialogue models commonly employed inODR models, we now offer a general classification of the types of dialogue. Aclassification system for the main types of dialogue in which argumentationoccurs was presented by Walton in The New Dialectic (1998). There each typeof dialogue is defined as an abstract normative model that sets standards forjudging how a given argument should be used correctly for some conversa-tional purpose in a given case. The six basic types of dialogue described inThe New Dialectic are persuasion, inquiry, negotiation, information-seeking,deliberation, and eristic (strife) dialogue. The properties of these six types ofdialogue are summarized in Table I below.

These models of dialogue are highly simplified in that there are only twoparticipants, a proponent and a respondent, and each takes turns makingmoves that represent speech acts like asking a question or putting forward anargument. There are no third parties like moderators or referees who ensurethat procedures are followed or decide the outcome. Real dialogues tend tobe much more complex. For example in a trial there are many participants:the plaintiff, the defendant, the judge, the lawyers for both sides, possibly ajury, and so forth.

The dialogue structures outlined above are commitment-based in the senseof Hamblin (1970, 1971). Commitment refers to what an arguer has gone onrecord as accepting, according to the evidence of what she said and did,

Table I. Types of dialogue

Type of dialogue Initial situation Participant’s goal Goal of dialogue

Persuasion Conflict of

opinions

Persuade other party Resolve or clarify issue

Inquiry Need to have

proof

Find and verify

evidence

Prove (Disprove) hypothesis

Negotiation Conflict of

interests

Get what you

most want

Reasonable settlement

that both can live with

Information-

Seeking

Need

information

Acquire or give

information

Exchange information

Deliberation Dilemma or

practical choice

Co-ordinate goals

and actions

Decide best available

course of action

Eristic Personal conflict Verbally hit out

at opponent

Reveal deeper

basis of conflict


indicated by the text and context of discourse in a case. This way of seeingargument as commitment-based makes the evaluation of a given argumentdetermined by the evidence of the given text of discourse in which the argu-ment was put forward. This commitment model is contrasted with the beliefmodel. The problem with the belief model is that an arguer’s actual beliefsmay be extremely difficult to determine, and doing so is a psychological task.

The six dialogue types classified above do not represent all the types thatare possible, but only the most common and important types that haveproved most necessary for the study of fallacies and other logical problems ofargumentation. An important feature of these types of dialogues is that, inmany instances, a text of discourse can be mixed, meaning that it combinesseveral types of dialogue. For example, legislative debates, like a debate onwhether to build a new dam, combine deliberation with information-seekingdialogue. Experts on dams, like hydro engineers, are brought in to presentfacts about the dam project proposed. The legislative debate may also involvepersuasion dialogue, as one party tries to persuade another. Also, there canbe dialectical shifts, or changes from one type of dialogue to another during aconnected sequence of argumentation. To cite a common example, a con-tractor and a homeowner may be engaged in deliberation on whetherinstalling a new concrete basement is a good idea. They may, at some pointshift to negotiation on how much the installation would cost. But then theargumentation may shift to an information-seeking dialogue as the con-tractor informs the homeowner on city regulations concerning requirementsfor thickness of concrete for house basements.

In some shifts, the new dialogue contributes to the success of the previousone. In such a case, we say that second dialogue is functionally embedded inthe first. In other cases, the second dialogue interrupts or even blocks theprogress of the first one. Blockages of this sort are often associated withinformal fallacies. Reed (1998) studied functional embeddings of dialogues asa problem for computer modeling of argumentation. An example commonlyused is that of two agents deliberating on how to hang a picture, where thedialogue shifts to negotiation when one proposes that the other should goand get the hammer and nail.

Negotiation may be contrasted with persuasion dialogue and inquiry. Thegoal of an inquiry is to prove something, or to disprove it or show that it canbe proved, by amassing and verifying all the relevant evidence. Persuasiondialogue is also about trying to find the truth of matter, but the matter is acontested issue on which there is uncertainty and lack of knowledge. Theviewpoints on either side of the controversy can only be evaluated looking atthe arguments on both sides and weighing up which side meets the burden ofproof (Prakken 1991). Negotiation is not primarily about finding the truth ofa matter, and if a participant treats it that way, he or she will do poorly in thenegotiation. At any rate, to contrast these types of dialogue more precisely,


we need to get a better account of the central characteristics of persuasiondialogue.

4. Persuasion dialogue

In a persuasion dialogue, various kinds of moves are allowed, including theasking of questions, the answering of these questions, and the putting for-ward of arguments. In persuasion dialogue, the one party, called the pro-ponent, has a particular thesis to be proved, and her goal is to use rationalargumentation persuade the other party, called the respondent, to come toaccept that thesis. The goal of the proponent is to prove this particularproposition that has been designated at the beginning of the dialogue as herultimate thesis (Walton 1999). The assumption is that the respondent, at theoutset, does not accept it. ‘Persuasion’ in this sense refers not to psycho-logical persuasion but to rational persuasion where the proponent presentsan argument containing only premises that the respondent is committed to,and uses this argument to get the respondent to become committed to theconclusion of the argument – a claim to which the respondent was notcommitted at the beginning of the dialogue (Bench-Capon 2002).

There are two kinds of cases. In a dispute, the respondent is committed tothe opposite (negation) of the proponent’s thesis. In a dissent, all therespondent has to do is to cast reasonable doubt on the proponent’s thesis.Persuasion, in this sense, refers to the change in the respondent’s commit-ments due to the proponent’s chain of argumentation in the dialogue. Beforethe respondent was not committed to the proponent’s thesis, but now he is.The proponent builds up a chain of argumentation using only premises therespondent has become committed to, and her ultimate goal of rationalpersuasion is only successful when the end point of the chain of argumen-tation is the her thesis in the dialogue (called the ultimate probandum in law).Thus there are four basic requirements that determine when the proponent’sargumentation in a persuasion dialogue is successful (Walton 1999, p. 121).

(R1) The respondent is committed to all the premises of the arguments(R2) Each single argument in the chain of argumentation is structurally

correct(R3) The chain of argumentation has the proponent’s thesis as its (ulti-

mate) conclusion(R4) Arguments meeting (R1), (R2), (R3) and (R4) are the only means that

count as fulfilling the proponent’s goal in the dialogue

The following diagram (Figure 1) gives the reader an idea of how a chainof argumentation works in a persuasion dialogue, following the four basicrequirements R1–R4.


Let’s say the respondent is committed to premises 1 and 2. Thus when theproponent uses the linked argument based on these premises (shown at thebottom of the diagram), the respondent becomes committed to conclusion 1.And let’s say that the respondent is committed to premise 3. The proponentcan then use premise 3, along with conclusion 1 to generate another argu-ment going to conclusion 2. But just in case the respondent might decide toretract commitment to premise 3 or conclusion 1, the proponent has anotherline of argument available. He could use premise 4 as a basis of evidentialsupport for conclusion 2. In any event, once he has gotten the respondent toaccept conclusion 2, assuming the respondent is also committed to premise 5,he can use both these propositions in another linked argument to support theultimate conclusion. Thus if you look over the whole chain of argumentation,its end point is the ultimate conclusion and its start points are premises thatthe respondent is committed to.

Some possible exceptions to these requirements concerning hypotheticaluses of arguments have been discussed in (Hamblin 1970, chapter 7). Dif-ferent formal models of persuasion dialogue called rigorous persuasiondialogues or RPD’s have been constructed in (Walton and Krabbe 1995).RPD’s are precise and formally rigorous but do not model realistic naturallanguage argumentation fully, because natural language persuasion dialogueis more flexible and open in certain ways. In a permissive persuasion dialogue(PPD), the moves a participant can make and the ways a respondent canreply are more flexible. The key problem in all these formal model of dialogueis that of retraction.

Figure 1. Diagram of a chain of argumentation.


One particular type of persuasion dialogue that has been widely recog-nized is the critical discussion, where the goal is to resolve a conflict ofopinions by rational argumentation (van Eemeren and Grootendorst 1984,1992). A successful critical discussion ends with a resolution of the conflict,for otherwise it is ‘‘not clear whether the discussion has had any point’’ (vanEemeren and Grootendorst 1984, p. 86). However, it has been recognizedthat there are types of persuasion dialogue other than critical discussion,where the dialogue may be regarded as successful if the discussion has thrownlight on the issue by bringing out strong and persuasive arguments, makingthe position on each side clearer and better defended. This clarification effectis called the maieutic function in (Walton and Krabbe 1995), referring theSocratic midwifery of assisting the birth of new ideas by rational dialogue, orthe rational process of discovery.

There are four stages in a critical discussion (van Eemeren and Groot-endorst 1992, p. 35): the confrontation stage; the opening stage; the argu-mentation stage; and the closing stage. At the confrontation stage the conflictof opinions to be resolved is clarified and identified. The viewpoint (point ofview) of the one party needs to be identified, and there has to be someexpression of doubt or disagreement about the viewpoint by the other party.At the opening stage, the two parties come to an agreement to resolve thisconflict of opinions by engaging in rational argumentation. At the argu-mentation stage, one party takes on the role of proponent (protagonist) andthe other party takes on the role of respondent (antagonist, of opponent),and each side puts forward arguments to support its viewpoint. Thesearguments fit argumentation schemes (see Walton 1996) representing theproper form each argument should take. At the closing stage, the two partiescome to a common evaluation of the outcome of the dialogue and decide whohas won.

The argumentation stage of the critical discussion is governed by ten rules(van Eemeren and Grootendorst 1987, pp. 184–293). These ten rules can beexpressed in a simplified form as follows. (1) Parties must not prevent eachother from advancing arguments. (2) An arguer must defend her argument ifasked to do so. (3) An attack on an arguer’s position must relate to thatposition (and not some other position). (4) Giving relevant arguments for aviewpoint is the only way it can be defended. (5) An arguer can be held to hisimplicit premises as commitments. Rules (6) and (7) can be combined into asingle requirement expressed by the following condition. An argument mustbe regarded as conclusively defended if its conclusion has been inferred byargumentation schemes (structurally correct form of inference) from premisesthat have been accepted by both parties at the outset of the discussion. (8)Arguments must be structurally correct, or be capable of being made so bythe addition of implicit premises. (10) Formulations must not be undulyvague or ambiguous.


Violations of these rules are associated with known informal fallacies. Forexample, committing the ad baculum fallacy of using threats or force isassociated with rule (1). Fallacies relating to burden of proof, like argumentfrom ignorance or begging the question, could be violations of rule (2).Violations of rule (3) can commit the straw man fallacy of misrepresenting anopponent’s commitments to make his argument look weaker, and more easilyrefutable. Rule (4) requires relevant argumentation, barring arguments thatmiss the point or go to the wrong conclusion (ignoratio elenchi) includingemotional arguments like ad hominem or ad populum. The critical discussionis not a formal model of dialogue, but the rules have normative bite, and arethus a useful platform for studying informal fallacies and other phenomena.

5. Mochol’s model of negotiation dialogue

Having reviewed the basic characteristics of persuasion dialogues in thecontext of the various dialogue types studied within argumentation theory,we now turn to the question of how persuasion dialogues connect withthe negotiation dialogues prevalent in ODR. To address this question at atheoretical level, it is best to start with a specific model of negotiation dia-logue as it occurs in ODR. We have chosen the model presented by Mocholat the previous ICAIL ODR Workshop (held in Edinburgh 2003).

Mochol (2004) provides functional models of three different discoursesystem design patterns for three types of dialogue: deliberation, negotiationand a third which she calls argumentation. Each design system is composedof a set of three models: one describing the system components, a seconddepicting the use case of the system, and a third mapping the processes oractivities involved in each system. While the component model describes thenature and relation of each functionally defined part of the system, the usecase model describes the ways in which it is envisioned that users of differenttypes would interact with the system as a whole. The activity or processmodel is presented as a kind of a flow-chart that describes the sequence ofprocesses or activities by which the system executes its designed function.Mochol then offers a general model of the discourse system design patternbased on the features that are generally common to all three dialogue sub-types. Since it is the process, or activity, diagram which models the proce-dural aspects of the system, it is primarily this model that describes theargumentative and discursive aspects of the system. As such, it is this modelwhich concerns us for our present purposes.

Figure 2 below is the model of the negotiation process given by Mochol(2004, p. 70).

When considering Mochol’s model of a negotiation system, there areseveral noteworthy features that merit a brief discussion. The first of these


features is that the model does not allow that the negotiation end in dis-agreement. As such the model is descriptively inaccurate since not all nego-tiations end in agreement. Furthermore, even if the model is conceived of as anormative model of an ideal negotiation process, it should still provide fordisagreement as a possible outcome, since sometimes the most rational courseof action in a negotiation could be to end the discussion (e.g. when anegotiator whose initial position is manifestly unreasonable steadfastly

Figure 2. Mochol’s model of the negotiation process.


refuses to make concessions over the course of negotiation). As such, we feelthat models of negotiation discourse systems should provide for disagree-ment as a possible outcome.

Perhaps the most important feature of this model, though, is that it leavesthe actual process or activity of negotiation as a black box. (In the abovediagram, we have highlighted the ‘‘negotiate’’ box in light grey.) Yet, fromthe perspective of argumentation and discourse theory, it is precisely at thispoint where all of the work in the model occurs, and it is here where themodel must soundly reflect the normative, structural and pragmatic featuresof argumentative discussions such as the negotiation dialogue. Moreover,when considering the question of where, within the overall negotiation pro-cess, persuasion dialogues could be functionally embedded, it will largely bewithin the actual negotiation dialogue itself that the embedding occurs.

6. A proposed revision to mochol’s model of negotiation dialogue system

In light of these observations, we propose a revision to the above model ofthe negotiation process. In the diagram below, we offer a model of the actualnormatively structured discursive activities involved in the activity of nego-tiation. We propose that this model replace that section of Mochol’s modelcalled ‘‘negotiation’’ (i.e. the recursive section between pre-negotiation andpost-settlement). In the proposed model, ‘‘P’’ represents an agent or partic-ipant in the negotiation process.

2

‘‘O’’ represents an offer. At its basis, thenegotiation process is represented as a sequence of offers made by one orother of the participants, which are discussed and either accepted or rejectedin a structured way. In the event that an offer is not accepted, it could berevised, or a counter-offer could be proposed. Alternately, it might bedetermined that no satisfactory resolution is possible. The process is recur-sive; as it recycles, offers are revised or counter-offers are made. Each time theprocess returns to the first step where a new offer is made, the number of theoffer under discussion increases by one. In the event that a counter-offer ismade, the roles of the participants in the dialogue switch (this is reflected bythe change in values of i and r.). Finally, in each ‘decision box’ (representedby diamonds) it is assumed that the negotiator will make her decision basedon her commitments (those claims in her commitment store) as well as hergoal-set (including her maximally and minimally acceptable resolution, andother long-term goals).

Here (Figure 3, below), then, is the revision we propose to Mochol’smodel, which we feel will better allow it to capture the normative dialecticalstructure of the negotiation process.

Admittedly, the proposed model of the discursive activity of negotiation ishighly simplified. For example, on this model there are only two parties


involved in the negotiation, and each offer originates from one or other ofthese parties. Developments of this model might allow for more participants(whether interested parties or neutral ones), and provide that offers (i.e.proposals for the resolution of the dispute) could originate from any partyinvolved, including the computer system itself. Further, the model representsnegotiation largely as distributive as opposed to interest-based. The activityinvolved is represented as a kind of bargaining, instead of an integrativeactivity where the inexplicit, long-term interests of the negotiators areexplored in an attempt to find novel solutions. Developments of the proposedmodel might include sub-routines aimed at uncovering and making explicitthe broader, long-term interests of participants in the negotiation as a meansto discovering alternative resolutions to the dispute which might not other-wise have been proposed as an offer or counter-offer.

Figure 3. Basic negotiation dialogue.


So, the proposed model is highly simplified and could be developed inseveral respects. That said, it has several advantages. In the first place, itreflects the interactive, turn-taking activities that characterize and regulatethe process of negotiation. This is an important normative and proceduralfeature of all argumentative discussions including negotiations, and onewhich we feel should be structurally implemented into all ODR systems ifthey are to be effective in managing a discussion aimed at reaching resolu-tions satisfactory to the parties involved. Yet, not all communicative envi-ronments available to ODR can (without some further modification) regulatediscussion in this way. For example, many synchronous interfaces allow forparties to interrupt one another by composing and sending messages beforetheir interlocutors have had a chance to reply to an initial message (e.g. ininstant messaging, or chat-room forums). As Rule has observed (2002, p. 52)this not only gives the faster typist a significant advantage, it also frustrateseffective communication which is a necessary requirement for any successfulODR system. Indeed, we would argue that it violates the basic Griceanprinciples of co-operation (1967/1989) and the idea that meaningful dialogueinvolves turn-taking.

3

In general, we feel that, in order to be effective as ameans to achieving resolution, ODR requires structured dialogue models inwhich the communicational activities of participants are normativelyregulated.

4

A second advantage of the proposed model is that reveals the internalstructure of the negotiation process in such a way as to show how and wherepersuasion dialogues might occur within it.

7. Functional embedding of persuasion dialogues in a negotiation dialogue

Considering the proposed model, there are several places in a negotiationdialogue where a persuasion dialogue could be functionally embedded. Thatis, there are several places where there could be a shift within the process ofnegotiation to a process of persuasion, only to resume the negotiationprocess later. For example, in the proposed model of negotiation, we haveassumed that participants need not defend or justify their offers with rea-sons. Offers can be made solely in the hope that they will be accepted, andthat this will maximize the desired goal-set of the party making the offer.Similarly, a negotiator can reject an offer without having to justify herrejection with a reason. Because negotiations are primarily agreement-based, the making or rejection of an offer does not bring about any burdenof proof on the relevant party to justify their move in the negotiation.Negotiators are not required to justify their offers or defend their rejectionswith reasons. Yet, while not required to do so, negotiators can do so, andthey may even see a strategic advantage in doing so. Negotiators might well


choose to support their offers or their rejections of offers with reasons, inan attempt to negotiate a better settlement. (Better settlements might beseen as more fair, or simply as more lucrative, for the individual negotia-tor.) Yet if this happens, the dialogue has shifted from a negotiation to apersuasion dialogue.

So, in the negotiation process itself, there are two primary occasions wherepersuasion dialogues can be embedded: (i) following an offer (where reasonsare provided in support of the offer), and (ii) following a rejection of an offer(where reasons are provided to demonstrate the unacceptability or unrea-sonableness of the offer). These two functional embeddings can be repre-sented by modifying the model of the basic negotiation dialogue in the waysindicated in the following diagram (Figure 4, below).

In the above model, the persuasion dialogue (PD) itself is represented as a‘black box’ (the grey activity state bubbles), the structural and proceduralfeatures of which are not given. (This topic is briefly taken up below (insection 9).) What is important to recognize is that the entire argumentativeprocess of the persuasion dialogue is contained entirely within the negotia-tion dialogue. Further, the persuasion dialogue can result in one of twooutcomes, where the thesis (claim or standpoint) under dispute is eithersuccessfully established, or is not successfully established.

We have tried to simplify the above model by making some assumptionsabout the thesis that will be argued in each of the two situations wherepersuasion dialogues can occur. In the first situation, where a negotiatorsupports an offer with reasons and argument, we assume that the thesis beingasserted is that the respondent should accept the offer. Thus, should theproponent be successful in her argumentation, this claim would be added tothe commitment store of the respondent who is then rationally obliged toabide by this commitment and accept the offer. Should the argumentationfail, the model simply behaves as if no reasons had been offered. That said,certain claims will likely have been added to, or removed from, the com-mitment stores of the negotiators over the course of the argumentation. Sincethe model prescribes that negotiators will make decisions based on their goalsas well as their commitments, the argumentation may well affect the sub-sequent course of negotiations. For example, the persuasion dialogue mightwell affect how a negotiator revises her offer, or how a counter-offer isformulated.

Similarly, where a negotiator offers reasons for her rejection of an offer,we have assumed that the thesis being argued is that the offer is unacceptable(or unreasonable). For this reason, if the argumentation is successful themodel requires that the negotiator having made the offer revise her offer so asto accommodate whatever aspects of it were unacceptable or unreasonable(since these will now be reflected by claims in her commitment store). On theother hand, should the argumentation here be unsuccessful, the model again


simply behaves as if no reasons had been offered. As was said before though,negotiators must act on the basis of their commitments as well as their goals,and so argumentation can well influence the process of negotiation even whenthe main claim at issue is not successfully established.

Having set forth the model, and given a brief description of it, severalobservations can be made. The first is that the model shows how the func-tional embedding of different types of dialogues as sub-routines in a largerdiscourse process can quickly complicate the model. Indeed, the simplicity ofthe negotiation model is due largely to the limited number and type of movesthat can be made within it. If we were to provide a model that allowed for allof the various kinds of communicative acts which might occur at any junc-ture in the negotiation process (e.g. requests for information or clarification,asking questions about goals, interests or commitments, etc.) the model

Figure 4. Negotiation dialogue with functional embedding of persuasion dialogue.


would quickly become so complicated that no diagrammatic representationof it would be beneficial, if indeed it was even possible.

A second important feature of the proposed model is that is shows theimportance of persuasion dialogues on negotiation by prescribing the effectsof persuasive argument on the negotiation process. Specifically, as was dis-cussed above, effective persuasion can compel a negotiator to accept or revisean offer. It can also effect changes in the commitment store of a negotiatorwhich will subsequently guide that negotiator in the decisions she makesthroughout the rest of the negotiation process. The effects of persuasion inthe process of negotiation are, therefore, substantial.

Finally, there are still other places where persuasion might have a role innegotiation. For example, in the post-settlement phase, when the fairness oroptimization of the result of the negotiation dialogue is evaluated. (InMochol’s model this is discussed in terms of the ‘‘optimal offer’’.) Thejudgment that a resolution is fair or optimal is one that requires reasons, andcould become subject of dispute if, for instance, the judgment is objected toby some party either internal or external to the negotiation itself.

8. Shifting from negotiation to persuasion dialogues

Having considered how persuasion dialogues can be functionally embeddedin the process of negotiation, it remains to be seen whether there are otherways in which persuasion is connected with negotiation. A second way inwhich negotiation can switch over to persuasion is if the entire dialogue shiftsfrom one type to another. When dialogues are functionally embedded, onedialogue occurs wholly within another as a sub-routine. When the sub-rou-tine ends, the primary dialogue resumes, albeit perhaps in some changed waythat reflects the outcome of the nested dialogue. When a dialogue shift oc-curs, a dialogue which begins as having the characteristics of one dialoguetype shifts into a dialogue of a different type. Here, the structural and pro-cedural features of the initial dialogue itself change and, as a result, theprocedural regulations and normative standards which govern the dialoguechange accordingly.

While it is more difficult to model precisely at what point in a negotiationdialogue such a shift could occur, it is quite clear that negotiation dialoguescan change to persuasion dialogues or perhaps to deliberation dialogues.Such a shift in dialogue type can be said to have occurred when basic featuresof the dialogue including the participants’ attitude towards the goal of thedialogue, as well as their general approach to this goal and strategicmanoeuvring within the dialogue, have changed from that of one dialoguetype to that of another. One example of such a shift could be when partici-pants no longer ‘barter’ their way to agreement, but instead they undertake


to base any agreement they might reach on the reasoned defense of a claimwith positive supporting evidence and response to critical argument. Here,one of the participants would take up a position on what the best solutionwould be, and then support her standpoint with reasons offered to try toconvince the other participant in the dialogue. The other participant wouldattempt to provide reasoned objections to this standpoint, and perhaps toprovide positive reasons in support of an alternative and opposingstandpoint.

While this approach might at first seem more adversarial than astraightforward negotiated settlement, it can often be an effective approachto agreement when negotiations have broken down as might occur when thetopic of the dialogue is highly personal or when a negotiation dialogue hasbecome highly emotional. For example, Jacobs and Jackson et al. (1987)have observed the benefit of this type of dialogue shift in child custody cases.Suppose the parents are negotiating on who should get custody of the chil-dren. This struggle can become very personal and emotional. But suppose themediator shifts the discussion to a persuasion dialogue on the issue of who isbetter equipped to look after the children. This is a matter more of objectivefacts that can be discussed dispassionately. Jacobs and Jackson describe howthis type of dialogue shift can be effected by mediators of a negotiationthrough the careful and strategic employment of questions designed to framethe issue under discussion in a more de-personalized way, and to elicit rea-soned, as opposed to eristic or emotional, responses to moves made by otherdialogue participants. While well beyond the scope of the present paper, animportant question for ODR is how the need for such a shift could bedetected, and the shift itself effected, by an automated negotiation system.

9. Modeling persuasion dialogues

Having shown a number of ways in which persuasion dialogues are con-nected with negotiation, and the kinds of effects persuasion can have on thenegotiation process, it remains to consider how persuasion dialogues shouldbe modeled. While we feel that this is an important question for ODR, theprovision of such a model is well beyond the scope of the present paper. But afew preliminary comments can be made here which might help to inform thedevelopment of such a model.

In the first place there are at least two contextual features which providecertain constraints on anymodel to be developed. From the point of view of thetechnical aspects ofODR, anymodel of persuasiondialoguemust be applicablewithin the computerized framework ofODR. So, if our aim is to fully automatethe task of mediating the negotiation process, we require a technical, or for-malizable, model of persuasive discussions that can be implemented at the


mechanical or functional level of computer systems.On the other hand, we alsorequire amodel that accurately reflects the normative procedural and structuralfeatures of persuasion as a type of dialogue. That is, persuasion is a type ofhuman dialogue exchange which is governed by a number of rules, both pro-cedural and epistemic, which come from a variety of sources, including thenature of the speech acts which are involved in persuasion.

This brings us to another important point. Our model of persuasiondialogue must reflect the variety of speech acts that can be made as part ofthe complex speech act of persuasion itself. And it must do so in such a wayas to reflect all of the combinatorial possibilities presented by these speechacts and the possible responses to them. Modeling all of these possibilitieswill generate a good deal of complexity which could quickly becomeunmanageable if models are developed along certain lines.

A rule-based model of the sort offered by Hitchcock et al. (2002) con-cerning deliberation dialogues presents a fuller analysis of the structure of thedialogues involved than the diagrammatic type of model considered in thispaper. A rule-based model is a dialectical type of model that begins with aninventory of the possible locutions that can occur within a dialogue of sometype. This includes speech acts (construed as moves within the dialogue) suchas assertion, questioning, hypothesizing, etc. Governing each of these dia-logue moves are a set of procedural rules giving the pre-conditions and post-conditions for each speech act. Such an approach allows that a variety ofspeech acts can be made at any move in the dialogue, just so long as the pre-conditions for that dialogue move are met. This is a more effective way ofrepresenting the dialogue while managing its inherent complexity thanattempting to provide a diagram of the entire dialogue-type.

In addition to these rule-governed speech acts, an effective model ofpersuasion dialogues would have to have several other components. It wouldhave to have a commitment store for each agent in the dialogue. It would benecessary that this commitment store could be changed in light of retractions.Further, the model would require a set of inference rules which would licensethe drawing of certain conclusions inferred from the commitments in thecommitment store. These inference rules could include not only formallydeductive rules such as modus ponens or modus tollens, and rules for inductiveinference (taken from the probability calculus) but also licenses for defeasiblepatterns of inference such as those provided by argumentation schemes (seeWalton 1996).

Finally, the dialogue system would have to include not only logicallybased rules of inference, but also pragmatically based rules. For example, themodel could incorporate Gricean rules of conventional and conversationalimplicature (1967/1989). These would help to specify the practical rulesgoverning certain speech acts and their consequences, and as such could helpto inform the statement of the pre-and post-conditions of individual dialogue


moves as well as more general procedural regulations of the dialogue itself.While this sketch alone does not encompass those considerations which willbe required in the development of a model for persuasion dialogues, it doesprovide a number of considerations and components which will have to betaken account of in any model.

10. Summary and conclusions

In this paper, we have attempted to show the how persuasion dialogues canbe connected with negotiation dialogues in the larger context of OnlineDispute Resolution. Having provided a model of negotiation dialogue, wehave shown how persuasion can be incorporated into this model by locatingthe places at which persuasion dialogues can be functionally embedded intonegotiation dialogues. In doing this, we have demonstrated the effects thatpersuasion can have on the process of negotiation, and the need for ODRsystems to incorporate persuasion into their dialogue models.

More generally, this has shown the need for ODR models to reflect thedifferent types of dialogue and other speech acts which can have a role in theresolution of a dispute handled in an ODR system. Also, we have suggestedthat ODR systems must be able to manage and regulate communication insuch a way that basic cooperative principles are not violated, if they are to beeffective as systems which support or facilitate the resolution of disputes.Finally, we have recommended an approach for developing models of per-suasion dialogues as well as other dialogue systems as they have a role inODR.

Acknowledgements

Research for this paper was made possible by a Research Grant from theSocial Sciences and Humanities Research Council of Canada.

Notes

1 While we do not argue this thesis here, we feel that there is good reason to contextualizemany, if not most, situations in ODR primarily as deliberations rather than as negotiations.

Essentially, the participants are trying to resolve some problem to the satisfaction of allconcerned, and this resolution typically requires that certain parties take some specific ac-tion. Negotiation might be one means of finding this solution, but there could be otheravailable means.2 In the model, i and r are variables ranging over the participants in the dialogue, whocould be named with letters such as ‘a’ and ‘b’. Since the roles of the participants canswitch when a counter-offer is made, there is a rule for the model which states that ‘‘i=aM

r=b’’ and ‘‘i=bM r=a’’.


3 It is not immediately apparent which of the four principal Gricean Maxims (of quantity,quality, relation and manner) (1967/1989, 28ff) are violated by interruption.4 Almost all pragmatic theories of argumentation have incorporated Grice’s conversationalmaxims into their dialogue models, and we feel that the creation of effective dispute resolu-

tion systems in ODR involves implementing these maxims in to the regulative structural andprocedural features of the dialogue models used in ODR. This question is well beyond thescope of the present paper, and one which we will have to leave for future consideration.

References

Bench-Capon, T. J. M. (2002). Agreeing to Differ: Persuasive Dialogue Between Parties withDifferent Values. Informal Logic 22: 231–245.

Craver C. (1994). Effective Legal Negotiation and Settlement. Michie Co.: Charlottesville,VA.

Eemeren, F. H. van and Grootendorst, R. (1984). Speech Acts in Communicative Discussions.

Foris: Dordrecht.Eemeren, F. H. van and Grootendorst, R. (1987). Fallacies in Pragma-Dialectical Perspective.

Argumentation 1: 283–301.

Eemeren, F. H. van and Grootendorst, R. (1992). Argumentation, Communication andFallacies. Erlbaum: Hillsdale, NJ.

Goldberg, S., Sander, F. and Roberts, N. (1992). Dispute Resolution, Negotiation, Mediation

and Other Processes. Little Brown and Company: Boston.Grice, Paul. ([1967] 1989). Logic and Conversation. In Studies in the Way of Words, 22–40.

Harvard University Press: Cambridge, MA.Hamblin, C. L. (1970). Fallacies Methuen: London.

Hamblin, C. L. (1971). Mathematical Models of Dialogue. Theoria 37: 130–155.Hitchcock, D., P. McBurney and S. Parsons. (2002). A Framework for Deliberation Dialogues.

In Hansen H. V., Tindale C. W., Blair J. A. and Johnson R. H. (eds.), The Proceedings of

Argument and Its Applications: Proceedings of the Fourth Biennial Conference of theOntario Society for the Study of Argumentation (OSSA 2001). OSSA: Windsor, ONcompact disk. Available on Peter McBurney’s web page: http://www.csc.liv.ac.uk/~peter/.

Jacobs, S. Jackson, S., et al. (1987). Ideal Argument in the Real World: Making do inMediation. In Wenzel, J. (ed.) The Proceedings of Argument and Critical Practices: Pro-ceedings of the Fifth SCA/AFA Conference on Argumentation. Speech Communication

Association: Annandale, VA: 291–298.Katsh, Ethan and Janet, Rifkin (2001). Online Dispute Resolution: Resolving Conflicts in

Cyberspace. Jossey-Bass: San Francisco.Leeson, S. and Johnson, B. (1988). Ending it: Dispute Resolution in America. Anderson

Publishing: Cincinnati.Lodder, Arno R., Bol, Setphanie H. (2004). Towards an Online Negotiation Environment:

Legal Principles, Technical Requirements, and the Need for Close Cooperation. In Lod-

der, A. (eds.), et al. Essays on Legal and Technical Aspects of Online Dispute Resolution,7–13. CEDIRE: Centre for Electronic Dispute Resolution: Amsterdam.

Mochol, M. (2004). Discourse Support Design Patterns. In Lodder, A., et al. (eds.), Essays

on Legal and Technical Aspects of Online Dispute Resolution, 61–74. CEDIRE: Centrefor Electronic Dispute Resolution: Amsterdam.

Patterson, S. and Seabolt, G. (2001). Essentials of Alternative Dispute Resolution. (2nd ed.).Pearson Publications: Dallas.


Prakken, H. (1991). On Formalizing Burden of Proof in Legal Argument. In Proceedings ofLegal Knowledge-Based Systems: JURIX 99, the Twelfth Conference, 85–97. GerardNoodt Instituut: Nijmegen.

Reed, C. (1998). Dialogue Frames in Agent Communication. In Demazeau, T. (ed.), Pro-

ceedings of the Third International Conference on Multi-Agent Systems, 246–253. IEEEPress.

Rule, C. (2002). Online Dispute Resolution for Business: For E-commerce, B2B, Consumer,

Employment, Insurance, and Other Commercial Conflicts. Jossey-Bass: San Francisco.Walton, Douglas. (1996). Argumentation Schemes for Presumptive Reasoning. Mahwah,

NJ: Lawrence Erlbaum.

Walton, Douglas. (1998). The New Dialectic: Conversational Contexts of Argument.University of Toronto Press: Toronto.

Walton, Douglas. (1999). Dialectical Relevance in Persuasion Dialogue. Informal Logic 19:

119–143 Available on Douglas Walton’s web page: http://www.io.uwinnipeg.ca/~walton/p_and_p.htm.

Walton, Douglas N. and Krabbe, E. C. W. (1995). Commitment in Dialogue: Basic Conceptsof Interpersonal Reasoning. State University of New York Press: Albany.


Persuasion dialogue in online dispute resolution in pdf/05 Walton Godden PD... · some remarks on how persuasion dialogues might be modelled is such a way as to allow them to be implemented

Documents