What is in a Dialogue

(orig. 3.4.95)

This paper discusses the notion of dialogue in an attempt to propose a formalization. Thus it explores dialogical aspects that may be relevant to characterize a dialogue, and argues in favor of a restrictive definition of dialogue, and a lax notion of dialogical elements.

Material Description: (C1-C4)

C1.- two (or more) participants

C2.- Exchange illocutions

C3.- Hold beliefs

C4.- Beliefs change in at least one participant

Explicit Entailments (D1-D4)

D1.- Two or more participants

D2.- Exchange mutually intelligible illocutions

D3.- Illocutions are expression of their beleifs

D4.- Beliefs change as a consequence of the dialogical interchange

This is not enough. Crucial aspects that need to be made explicit:

Contradiction centered, atomic and complex dialogues. Two phases. A reductive view. What am I committing to when I speak of "belief". A brief discussion.

A reductionsist and simple proposal: Tarski-like dialogical structures and dialogical-frames

a) Relevant dialogical elements:

Participants

Illocutionary particles

Dialogical Protocol

Intentional Stance

Forms of Contradiction

Reconstructive devices

Terminating conditions

b) Canonical or typical interpretations

c) Other semantical issues

d) Elements I purposefully left out. A brief discussion.

Some specific systems. Properties.

Comments on Lorenz, Lorenzen, Hamblin, Rescher (Why I think my approach is better than these).

Abstract

By analyzing everyday dialogues, and families of dialogues, one immediately realizes that many dialogues share distinctive common traits.

This fact had already been appreciated by Aristotle and other theorists such as Searle, Flores and Jacques, who in turn have advanced more o less explicit characterizations and classifications of dialogical structures. But none of these authors was concerned with dialogical systems, as such, as the focus of his research, thus these attempts never intended to be complete or exhaustive.

On the other hand, Karl Otto Apel, and the Erlagen School has, in a sense, focused on dialogical systems and structures, but their characterizations and taxonomical efforts were restricted to dialogues as forms of colective argument, and consequently left out many everyday examples of dialogue.

In this paper we will advance a characterization of dialogues that is:

More inclusive than that of previous writers

Coincides with classical characterizations in crucial aspects, but allows for important new refinements and distinctions

Permits a classification of dialogues that is simple, convincing and technically fruitful.

Facilitates formalization and automation of dialogues

(orig. 30.3.95)

A discussion of general everyday dialogues shows that most of them contain some common features. One may be tempted to use these as the basis for characterization and further formalization of dialogues. But in so doing, one is left with the impression that characterizations are far to much ad-hoc.

One alternative is to break down the process of description of complex dialogues in two different layers. One layer formed by "atomic dialogues" which can be readily combined to form "complex" dialogues, and a second layer consisting of "dialogical elements" that are in turn used to describe the atomic dialogues.

In this paper we explore both alternatives, and show the superiority of the second approach, based on its descriptive, taxonomical, formal and operational advantages.

Abstract

Arguments that show that dialogues are abundant, practical, necessary for action coordination in multiagent environments, useful in learning and irreductible to monologues as argumentation devices.

(orig. 30.3.95)

Dialogues capture, in a very natural form, two crucial aspects of natural reasoning: the fact that beliefs (or knowledge bases) are not static and the fact that not everything that can be known by a reasoning agent is known to be known. In this article we explore this second aspect, that we shall call "opacity", through a simple monological concept, that of restricted deduction.

First, we discuss the intuitive notion of opacity and show how opacity is present in various guises in natural dialogical contexts. Secondly, we give a precise characterization of opacity and show alternative monological treatments of the notion. In a third section of this paper we use the notion of opacity to argue in favor of the irreductibility of dialogues. Finaly, we justify the inclusion of restricted deduction mechanisms in the formal treatment of dialogues.

When people engage in argument they are rarely aware of all their beliefs and committments, if they were --and didn't hold contradictory views, and they were sincere and purposeful-- they wouldn't need to argue. But, as Aristotle showed, it is through the exploration of one's own ideas in contrast with the questioning of others, that we become aware of their consequences, and thus of their ultimate validity.

This simple observation reveals two interesting facts:

First, that people need dialogical reasoning whenever the strenght of their deductive systems, or the quality of their premises, is insufficient to deal with a problem (or more specifically, to settle its truth conditions).

Second, that a formalization of dialogical reasoning should include means to express this limitation.

This article will attempt to explore these two facts in depth.

1.- Real world Opacity

A fundamental quality of most logical systems, at least for those developed during the first half of this century, has been completeness. This is clearly interesting for dealing with mathematics and other foundational topics, but it is not at all clear that natural reasoning systems, such as those used in legal reasoning, or in everyday situations, are anything close to complete.

In fact, if we observe living organisms, we can quickly identify all sorts of conducts that show a basic incompleteness of their deduction mechanisms. For example, mistakes, changeability, dubitation, conflict, ... can usually be shown to appear because an agent's deductive capabilities are in some sense "limited", and although the agent may hold adecuate premises, their pertinent conclusions are not allways present.

Sometimes it is a problem of speed. The agent could in principle reach the correct conclusion, but in reality, there wasn't enough time available (an thus, only an approximate, partial, or altogether wrong conclusion was reached).

Other times it is a lack of "inference" mecanisms, inadecuate analogies, inadecuate refinement of conditions, insufficient information on the problem. And, then again, partial, inaccurate or wrong conclusions are obtained.

And in ohter occassions, it is the fact that no inference or deduction mechanims were available, and the premises were insufficient.

As I will show, not all these conditions are interesting. We will concentrate in those of limitation by iteration of rules of inference, and those of incomplete rules of inference.

2.- Towards a formal characteristics of opacity

We say that an agent's beleif base is "opaque" if it does not include its deductive closure. And we say that an agent deductive system is restricted if it is limited (in the sense of allowing weaker forms of deduction, such as time-constrained deductions or iteration-bounded rules of inference).

3.- Opacity and the irreducibility of dialogues

4.- On formalizing opacity in dialectical systems

(orig. 30.3.95)

Dialectical argument, dialogical learning and illocutionary coordination of actions are all forms of dialogue, and all entail non-static sets of beliefs in at least one of the participants in the dialogical illocutionary exchange.

The fact that a form of dynamic belief system appears, naturally, in a dialogical contexts suggest that other dialogical features may be relevant to the formalization and treatment of some kinds of non-monotonicity.

In this paper we analize the similarities and differences in the mobility of the belief bases in the three dialogical contexts mentioned above. We show that while all three contexts share the same two dialogical "phases", namely contradiction detection, and contradiction correction, the three differ in the type and extent of belief revision. We then explore the connection between belief revision effects and contradiction correction strategies and devices, and show that a wide variety of possible belief revision effects and features can be obtained from very simple contradiction correction devices. Finally, we first show that classical Belief revision formalims, such as Galdenfords, Kazuno, .... correspond to, but do not exhaust, some of the correction devices presented, and then propose new Belief Revision Formalisms based on other ccd's.

Abstract

This is a brief description of the DETEC program, elaborated in 1986-88 in IBM's Mexico Scientific Center as a propositional and first order theorem-prover. It is based on Lorenz' Dialogical proposal for the semantics of Intuitionistic Logic, corrected and extended to Classical Logic by Lorenzen.

Abstract

This is a concise description of the DIALOREN program. An ESE-add on that performed TMS functions on top of IBM's Expert System Environment. DIALOREN proposed two stages in the process of updating KB, one was to detect contradictions during consultation or addition of rules, the other was the automatic correction of those detected contradictions. Foir the first stage a dialogical theorem prover was used (DETEC), for the other, a conservative "Lakatosian" theory correction mechanism was implemented.

Abstract

When one wants to automatically correct a first-order knowledge base, several alternatives are open. Here we explore the formal characteristics of a simple and convenient correction mechanism, based on I. Lakatos notion of the Evolution of Scientific Theories, in which corrections to an inconsitent Knowledge Base are accomplished by the addition of atomic or negated atomic subformulae to one sentence in a minimal inconsistent set. We prove that the process is effective and correct, but NP (?).

Abstract

This is an excercise to explore the descriptive advantages of a dialogical framework. I take a published account of an agent-based Knowledge System ( or "ABKS" for short), and reify its dialogical components. I show that a compact and lucid description is possible, in which for every feature in the ABKS there is a dialogical concept available. Second, I show that important dialogical aspects such as phases, contradiction detection, opacity and dynamicity of belief systems (that have been classical in the dialogical literature) are present in the ABKS architecture or intended behaviour. Finaly, I argue in favor of a compact generalization of this ABKS.

(orig. 29.3.95)

Dialectical tradition has used the notion of "contradiction" as its theoretical fulcrum. But Aristotle's notion of contradiction is quite different from that of Hegel, or even that of the Erlagen School.

In this paper I will discuss the theoretical aspects involved in the classical notion of contradiction in dialectical arguments, then I will show how similar, but not identical, notions are present in ellucidatory and action dialogues. From these distinctions I will argue that an elementary notion of contradiction can be extracted, and derived or ellaborated notions can then be proposed. Finally, I show how these different notions may be given a formal treatment and how they can be implemented.

Notions such as incompatible, contradictory, inconsistent, incoherent, inadmissible, ....

A full detailed description of what a dialogue-based argument is, has been and can be. From Aristotle to Perelman, from Dialectics to Rethoric. And, cognate forms of argument such as parlamentarian debates, disputes and impugnations. Lexical and formal analysis, from a philosophical and empirical perspective.

Abstract

One typical, and common form of dialogue is that of ellucidation, that is, to exchange illocutionary phrases in order to clarify meaning, or to elicit or discover new knowledge, and to make beleifs (or knowledge) explicit. All these forms share a common goal, that of obtaining new knowledge, and many share the same ellucidatory devices.

In this paper we make those intuitions explicit.

"By the way of the word is the wise man known" or so says the Eclesiastes, refering to the fact that wise people say intelligent things, and dumb people show their limitations in conversation. The crux of this passage is that language is a performative act not only in linguistic terms, but also in cognitive, and specially, epistemological terms. But, what does it have to do with learning? I claim that this performative competence is essentially a learning capability, and as such it has its own rules and regulations that I will attempt to make explicit here.

Abstract

An interesting practical problem in Knowledge based systems has been to construct, and update, knowledge bases. And although soft methodologies of different sorts have been proposed, rarely have they included a sound theoretical framework. In this paper we argue for the use of dialogical concepts in the formulation of a methodology and the proposal of an interactive tool for knowledge ellicitation, based in our research on ellucidatory dialogues.

Abstract

There are situations where speakes seem to be concerned mainly by "speaking their mind out", not necessarily in convincing the listener of anything, not even minding the reactions of the listener, but recognizing the need of a listener and the need to express themselves in a more or less coherent fashion. The purpose of this kind of dialogue is to organize, or "settle" the feelings of thte main speaker, and thus it proceeds in the same general fashion as other ellucidatory dialogues, but this case is much more asymetrical thatn others.

In this article I first explore the structural aspects of these "cathartic" dialogues.and establish some relationships with other forms of ellucidatory dialogues. Then I explore the relationship between emotions and this kind of dialogue, and suggest a further analysis between emotions and performative talk.

In [ Zeldin 95] the author cites the following opinion:

"Who do you have your best conversations with? With my dog, he really iunderstands me"[1]

Absurd as it may seem, this opinion reflects a classical, albeit extreme, form of cathartic dialogue. The 44 year old woman Zeldin refers to, is not a lunatic, she is just asserting the fact that in order to clarify her thoughts, in order to express her ideas, whenever a strong emotional state is involved, the listener's role is radically subordinated to that of the speaker, but it is hardly dispensable.

Everday situations abound where people speak for, apparently, no reason at all. It seems that there is no the point of winning an argument over an opponent, it is not clear if they are negotiating a position, or demanding a specific committment or action. What is clear in these situations is that people want to speak, and that their emotional state is (originaly) exalted, and becomes more peaceful only after they have "spoken their mind out".

What are the characteristic elements of these situations, can they be called dialogues, and tif so do they have anything in common with othr forms of dialogue. I will try to answer these three questions, and then I will focus my analysis on other emotional aspects of dialogue.

A few clear examples, followed by a brief generalization-abstractio

What the speakers usually do.

What are the kinds of things listeners do

A closer look at apparent goals

A closer look at real effects

Is there a Protocol? (explicit, implicit)

A dialogical description (recapitulation)

A diagramatical comparison with other dialogues

Is there a way of linking emotions with certain dialogical actions.

From emotion to speech

From speech to emotion

Abstract

Conversation is rarely emotionally neutral. Often, conversation is the result of an emotional state, or need, and the availability of listeners. And always, dialogue and conversation among people, alter emotional states.

In this article I want to explore how certain types of speech performances trigger emotional states within a dialogue. In fact I will first concentrate on how and why people loose their tempers when engaging in particular kinds of dialogue, and then I will extend my discussion to ohter emotional states and their relationship with dialogical performances. It will be argued that some emotional states can be adequately explained in (strictly) dialogical terms

This is a modest attempt to put in perspective, and in a concise form, the main ideas behind Searle's theory of Speech Acts, and its Flores-Winograd recasting, as far as they involve multiple agents and as much as they concern dialogical aspects.

Thus I will not touch upon specific aspects of these theories, unless they bear upon this general theory of dialogues.

In order to make some sense of the whole monster, I will divide my work into four main sections: First, historical background, inheritances influences and relationships. Second, Speech Act Theory and basic Illocutionary Logic. Third, "Conversations for action" and "Conversations to create possibilities". Fourth, what is missing in these theories (if anything).

1.- The Pragmatical Inheritances of Searle

There are three important antecedents to Searle's Speech Acts Theory, and the three have deep roots in the "Pragmatics" tradition, although Searle doesn't acknowledge this fact.

The first one is the notion that context diambiguates meaning in natural language. This we will refer to as the "contextual awareness" fact.

Second, the acknowledgement that language is used not only, not mainly and not predominantly to represent the world, but also to act. This we will call the "language acts" conception.

Finally, the realization that truth is but one of the many formalizable aspect of language. This we will call the "non-classical formal tradition".

We will discuss these three influences one by one.

Contextual Awarenes:

Indexicals, referential opacity, Stalnaker and Montague.

Literal sense and intended meaning. Grice's communicative rules. Implicatures.

Language Acts:

Wittgenstein and language games. Languace as a socially controlled activity, Language practice as following rules.

Austin on becoming competent in the use of language. Perfomative competence. Actors vs agents

The original lists of Wittgenstein and Austin. Austin's teleological distinctions, and the derived ontology.

Non-classical formal tradition:

Bar-Hillel, Montague, Hintikka, ... Possible world semantics, functions and modalities. Gazdar. Logic as Grammar.

2.- Speech Acts Theory and Illocutionary Logic

The basic forms of language acts. Declaratives, Committments, Propositions, Assessments.

Canonical forms, its components and implications.

Illocutants or speaker-hearer pairs.

Basic Illocutionary operations, and is Illocutionary Logic, logical?

The full list of acts and trees for english illocutions and operations

What Searle wanted to do and what he actually did.

3.- Conversations for Action, and Conversations to Create Possibilities

Flores' radical ecclecticism, a background notice.

A Conversation as sequence of spech acts. The original views. What Flores adds to Searle. Society as sets of (coordinatable) committments. Tools to communicate effectively. The Coordinator.

(Mis)-understanding Computers and Cognition. The Winograd factor, and AI, and Hubert Dreyfuss. The resonance of one book, and its sequels.

The idea of invention, speculation and creation of possibilities.

Beyond the Coordinator: Groupware, simplified conversations and the workflow paradigm.

4.- What they haven't seen

Impact of these ideas in Computer Science. Illocutionary based groupware and agent architectures and languages. Who is doing what today.

What is being done that is not a Searle descendent.

What have they missed.

Not, at least in the way we use the word "converse" in everyday language, nor in the technical way adopted by Flores and Winograd. But for a completely different kind of reasons.

Here I will try to structure a characterization of what I call a negotiation-as-a-dialogue concept, and show it encompasses most aspects of what people refer to as "negotiation", and will show how that is defferent from the notion of conversation. Second, I will explore Flores' notion of conversation, and what he calls the negotiation phase, and show it is essentially ellucidatory, and not ruly negotiational (in my terminology).

Finally, I will argue in favor of a systematic dialogical approach to the concept of negotiation, and its empirical consequences.

P Noriega and C. Sierra

IIIA

In this article we introduce the notion of computational environments that allow structured forms of negotiation among heterogeneous agents over a network.

In particular, we advance a theoretical framework for characterization, formalization and implementation of "negotiation environments" and show how this framework can be instantiated to create a typical negotiation environment:

1.- Motivation

Internet based agent negotiations.

a)Polimorphism of agents: human, reactive, deliberative...

Variety of trading conducts, diversity of markets, products, protocols participants.

But there is some underlying uniformity.

b) Usage. Explosion of free roaming agents will clog the net. Excesive advertisement and search activities, lack of guarantees, unstable protocols, difficult to standarize conventions.

Structured itrading and negotiation

Our proposal: Create A common context. Physical-virtual environment, shared conventions (ontology, language,...), Eligibility conditions for participants, simple central management, easy universal access, topic-instantiated general framework.

Will deal with only one example, the fish market model of bidding based negotiation, and comment on extensions.

2.- Basic Intuitions

What is a Fish Market.

site

purpose

participants (roles, expectations, values)

observed conducts (illocutionary exchanges vs. goods)

negotiation principles

negotiation protocols

admission (eligibility) criteria

solvency, bonding and commitment upholding

price, money and other values

strategy vs protocol

A first abstraction (a dialogical analogue)

illocution exchange

illocutionary particles

illocutonary content

real world correspondence to illocuted statements

role description

goal description

dialogical protocol

terinating conditions

3.- Dialogical Framework

Basic definitions (dialogue, dialogical context, dialogical framework,....)

Formal instantiation. The FM framework:

illocution exchange

illocutionary particles

illocutonary content

real world correspondence to illocuted statements

role description

goal description

dialogical protocol

commitment management and update

terminating conditions

Properties

4.- Implementational Issues

FM environment architecture

Trading Hall

Seller agents

Buyer agents

Callers and arbitrers

Agent admission board

Goods Clearinghouse

The trading dialogue language shell

Trading illocutions and tokens

Good-dependent linguistic elements

Valuation and vallidation

Turn-taking, rounds and bidding.

Terminating moves

Exception handling

The bidding sequence manager

Role script managers (definition, enablers, validators,...)

Agent eligibility tests and trading-habilitation procedures

Arbitration and sanctions

Network related shell

Hub management

Protocol conventions

External agent management

The economic-reality checker

Market-specific ontology

Risk and liability checks

Bonding and solvency issues

5.- The Fish-market model

Actual implementation of toy downward bidding FM.

Standards

Examples

6.- Discussion

What is nice about this example

What is too simplistic

What is impossible to achieve

Some economic and mathematical conjectures on sixe and diversity of traders, and their effect on prices

e.g CONJ1. "If there is a large number of buyers and sellers active, and the number of buyers is mkt-large w.r.t sellers, downward bidding benefits sellers. (and dual)"

7.- Extensions

More general price-based negotiation

Value-based trading (diversity of goods, unprecise descriptions, dynamic language trading)

General non crisp negotiation

Intermediated negotiation

Abstract

In practice, many multiagent interactions are "opaque", in the sense that relevant elements of the shared ontology and meaning are not known or established a priori, but only through the agents' interactions. While this condition may be unavoidable in many cases, it is not necessarily easy to contend formally or empirically with. Nevertheless, there are natural contexts where this "opacity" is --at most-- apparent, and apriori objective meaning and ontologies are not only available but effectively enforced. We call these environments "lucid"

In this paper we attempt to characterize the notion of a "lucid" context in which dialogical interactions among agents ocurr, and argue in favor of their formal treatment and the convenience of their computational interpretation.

1. Motivation

Intuition: What it means to be opaque and lucid

Opacity and situatedness are natural, but so are lucid contexts.

Empirical need for lucidity.

Convenience of a formal treatment of lucidity and artificial lcd's

Limits to lucidity

An overview of associated work

2. Towards a formal treatment of lucid dialogical contexts

A definition of lucid contexts

Relevant elements for a formal treatment

Context, illocutory exchanges, intensional comittments, dialogues, protocols and scenes

Some formal properties and nice examples

3. Computational models of lcd's

A proposal for an architecture

Ontological description

Satisfaction of illocutory comittments

Protocols for agent interactions

Scene description

I

4. Discussion

Why are they interesting

Pragmatic and theoretical issues:

Frame problem

Situations

Speech act issues. Conversations for action

1. Motivation

Intuition: What it means to be opaque and lucid

Opacity and situatedness are natural, but so are lucid contexts.

Empirical need for lucidity.

Convenience of a formal treatment of lucidity and artificial lcd's

Limits to lucidity

An overview of associated work

2. Towards a formal treatment of lucid dialogical contexts

A definition of lucid contexts

Relevant elements for a formal treatment

Context, illocutory exchanges, intensional comittments, dialogues, protocols and scenes

Some formal properties and nice examples

3. Computational models of lcd's

A proposal for an architecture

Ontological description

Satisfaction of illocutory comittments

Protocols for agent interactions

Scene description

4. Discussion

Why are they interesting

Pragmatic and theoretical issues:

Frame problem

Situations

Speech act issues. Conversations for action

Dialogues are pervasive and, to a large degree, unavoidable in everyday situations. Legal arguments, political debate, domestic disputes, didactic explanations, interviewing, psychotherapy, coordination of actions, negotiation, all tend to involve some form of dialogical interaction. But dialogues are also unavoidable since what is accomplished through them cannot be accomplished in a strictly "monological" setting, because some fundamental ontological, rethorical or epistemic features would be lost. Thus, dialogues may be worth studying, although they are not simple entities. Certainly not from a formal perspective.

On one hand, dialogues involve multiple participants, who exchange illocutions in rich and complex languages. Thus, classical --i.e. monological, truth-semantical, non-dynamic,-- formal devices, are inadequate to deal with these complexities. But in addition, dialogues are typically "opaque" (or "situated " or "unstructureded"), in the sense that In many dialogical situations meaning is not necessarily established in an objective, a priori, form; nor are interventions subject to an objective, a priori, clearly expressible protocol. In these cases, participants "react" to the illocutions, depending on the conditions or elements present in the given "context" or "situation". In typical dialogues, participants confirm, adjust, refine or establish their own meanings, intentions, beliefs and actions according to their individual interpretation and what the other participants are saying [c.f. Wittgenstein; Searle; McCarthy & Hayes; Winograd & Flores; Barwise & Perry].

The first kind of complexity may be addressed through ad-hoc "dialogical" structures [ c.f. Hamblin, Rescher, Hintikka, Carlson for different approaches], the second one may be sometimes avoided since the degree of "opacity" of dialogues can vary greatly.

Certainly there are situations which are inherently situational, such as psychotherapy or everyday conversation, where interpretational contexts are inevitably opaque, in the sense that meaning and commitments are mostly established through highly unstructured dialogical interactions. But then there are other contexts where a priori univocal shared interpretations ("transparency") is not only desired but enforced.

One example of enforced "transparency" is that of "structured negotiation environments" --such as auctions and other similar forms of mediated trading and negotiation. These naturally transparent contexts enforce univocity of interpretation, protocols and roles as a means for achieving efficiency, and guaranteeing acceptable conditions to participants. [c.f. North]

Take for instance the "downward-bidding" or "dutch" auction --so called because it is the way cut flowers have traditionally been traded in the Netherlands, as well as many other goods in many places [McAfee & McMillan 1987: 702]. In a dutch auction an extraordinarily simple negotiation convention is used: an intermediary, the main auctioneer, calls prices for a given item in a descending sequence, stopping when a bidder first indicates acceptance for the current price. But this simple process betrays a complex "institution" in which a "marketplace" with explicit and shared trading conventions --and meanings-- opens in a given "marketdate" when a specific group of sellers and buyers, submitting to those conventions, exchange a given collection of items under the supervision of market "intermediaries".

Some distinctive structured negotiation environments --auctions, stock markets, trading fairs, labor mediated-collective-bargaining-- have evolved over many years and are remarkably similar in structure and functionality across cultures and negotiated items. Being natural examples of transparent contexts, they are attractive candidates for formalization. They may also turn out to be interesting models for implementation of artificial agent-interaction environments. The kind of contexts, for instance, that may facilitate heterogeneous agents to negotiate. The kind of environments Internet is likely to adopt [Weld 1995: ???. Nosotros SARA].

In this paper we propose a formal approach to the description and modelling of these structured negotiation environments. We will give them a "dialogical" outlook by focusing on the illocutions that are uttered by participants in the negotiations --in the content of these illocutions and the illocutionary exchange process too. We draw our intuitions from the natural structured negotiation environments, such as auctions --and we will use the downward-bidding auction as an illustrative example throughout--; but the framework will be applicable to other forms of structured agent interactions, not only auctions, and not only negotiation.

Abstract

We present an outline of a dialogical notation that can be used to describe the dialogical processes that characterize lucid multiagent interactions such as are present in auctions and other structured negotiation environments. This notation has the advantage that if and when the description is adequate, it spawns a computational interpreter that constitutes a "model" for the description i.e. an actual implementation of a multiagent environment that corresponds to the description.

1. Background

1.1 Motivation

Lucid Dialogical contexts. Practical situations that correspond to..

Intuitive definition of lucid-dial-con

Why it may be interesting to automate them: pragmatical convenience robot-clog, abundance of dialogical contexts, test beds for agent interactions. Inherent complexity, apparent diversity, commonality of structure and components.

An ideal working solution, what should it amount to?

1.2 Our previous work:

Feasibility of adequate description

Convenience of a model-theoretic formalism

Intended computational interpretation

Actual implementations of FM

1.3 What we now want to accomplish

How we intend to generalize those previous efforts

Related work: Wokflow, Prolog, APL.

1.4 What we are going to do in this paper

We will show how it is possible to develop an executable notation for sialogical contexts. One that is rich enough to characterize, precisely, a variety of dialogical contexts (as long as these are "lucid"), and concrete enough to actually have implemented counterparts.

We will describe a "methodology" for description of lucid dial-cont, that will make explicit scenes, scene outlines and basic ontologies of a ldc.

We will then present the basic constructs of the notation that allow for the description to be coded into these constructs.

We show how the basic constructs interact as "building-blocks" whose compatibility conditions correspond to "felicitous termination" conditions, and to efffective composition of effective procedures.

We make precise the notions of an "adequate" description for an lcd, and that of correspondence between a description and an interpreter.

We prove that these constructs and their interpretations actually are sufficient to implement a well-defined class of lcd's.

We finally show how to describe some specific structured multiagent negotiation environments in this notation, and what the corresponding computational implementations are like.

Finally we discuss the significance of this work and suggest directions for further research.

Our work

Plato, or at least so claims Gadamer, used Dialogues not only as a literary device, but as a necessary form to "convince". Is there any difference.

I argue that the borders of Rethoric are fuzzy, at best, and in the worst case, they are the same as those of literary intention.

Dialogue is sometimes used to prove things, other times to convince or to ask, or to promote..i.e. in an intentional stance. But many times, dialogues are also used to clarify a definition. Thus as representational devices. As such, the Toronto and Erlagen schools have used them. That is why they missed out on agents.

Part One: Natural dialogue under artificial cicumstances

One can design situations where dialogical exchanges happen under controlled conditions. Examining these exchanges allow us to reify formal components that enrich our theoretical cogitations.

Examples:

Machine interviews: Eliza like dialogues

Negotiations

Polls and directed questioning

Facing contradictions, natural escapatories...

Part Two: Artificial Dialogues in their natural form.

Likewise, one can take specific real dialogues (transcriptions, parphrasings, evocative accounts,...) and subject them to the same kind of analysis.

Arcipreste de Hita

Classical arguments (Plato, Cicero)

Abstract

The Chiapas conflict has apparently been a long peace-negotiation dialogue, but in close analysis one can find that participants, roles, strategy and protocol have been shifting to accomodate quite different objectives over time.

In this note I attempt to analyze the dialogical elements that are present in the readily available printed documents --which have systematically invoked dialogical terms-- of this notorious conflict . I claim that a good analysis is possible, and that it may have some explanatory value. I don't want to claim that it may have any directive or predictive one. I think it may be interesting and worthwhile.

A discussion of practical situations in which dialogical talk is frequent. A careful assessment of what the intuitive notions detect, and what is missed or exagerated in naive views of dialogue.

(orig.3.4.95)

Abstract

An interesting practical problem in Knowledge based systems has been to construct, and update, knowledge bases. And although soft methodologies of different sorts have been proposed, rarely have they included a sound theoretical framework. In this paper we argue for the use of dialogical concepts in the formulation of a methodology and the proposal of an interactive tool for knowledge ellicitation, based in our research on ellucidatory dialogues.

Abstract

There are situations where speakes seem to be concerned mainly by "speaking their mind out", not necessarily in convincing the listener of anything, not even minding the reactions of the listener, but recognizing the need of a listener and the need to express themselves in a more or less coherent fashion. The purpose of this kind of dialogue is to organize, or "settle" the feelings of thte main speaker, and thus it proceeds in the same general fashion as other ellucidatory dialogues, but this case is much more asymetrical thatn others.

In this article I first explore the structural aspects of these "cathartic" dialogues.and establish some relationships with other forms of ellucidatory dialogues. Then I explore the relationship between emotions and this kind of dialogue, and suggest a further analysis between emotions and performative talk.

In [ Zeldin 95] the author cites the following opinion:

"Who do you have your best conversations with? With my dog, he really iunderstands me"[1]

Absurd as it may seem, this opinion reflects a classical, albeit extreme, form of cathartic dialogue. The 44 year old woman Zeldin refers to, is not a lunatic, she is just asserting the fact that in order to clarify her thoughts, in order to express her ideas, whenever a strong emotional state is involved, the listener's role is radically subordinated to that of the speaker, but it is hardly dispensable.

Everday situations abound where people speak for, apparently, no reason at all. It seems that there is no the point of winning an argument over an opponent, it is not clear if they are negotiating a position, or demanding a specific committment or action. What is clear in these situations is that people want to speak, and that their emotional state is (originaly) exalted, and becomes more peaceful only after they have "spoken their mind out".

What are the characteristic elements of these situations, can they be called dialogues, and tif so do they have anything in common with othr forms of dialogue. I will try to answer these three questions, and then I will focus my analysis on other emotional aspects of dialogue.

A few clear examples, followed by a brief generalization-abstractio

What the speakers usually do.

What are the kinds of things listeners do

A closer look at apparent goals

A closer look at real effects

Is there a Protocol? (explicit, implicit)

A dialogical description (recapitulation)

A diagramatical comparison with other dialogues

Is there a way of linking emotions with certain dialogical actions.

From emotion to speech

From speech to emotion

Abstract

Conversation is rarely emotionally neutral. Often, conversation is the result of an emotional state, or need, and the availability of listeners. And always, dialogue and conversation among people, alter emotional states.

In this article I want to explore how certain types of speech performances trigger emotional states within a dialogue. In fact I will first concentrate on how and why people loose their tempers when engaging in particular kinds of dialogue, and then I will extend my discussion to ohter emotional states and their relationship with dialogical performances. It will be argued that some emotional states can be adequately explained in (strictly) dialogical terms

This is a modest attempt to put in perspective, and in a concise form, the main ideas behind Searle's theory of Speech Acts, and its Flores-Winograd recasting, as far as they involve multiple agents and as much as they concern dialogical aspects.

Thus I will not touch upon specific aspects of these theories, unless they bear upon this general theory of dialogues.

In order to make some sense of the whole monster, I will divide my work into four main sections: First, historical background, inheritances influences and relationships. Second, Speech Act Theory and basic Illocutionary Logic. Third, "Conversations for action" and "Conversations to create possibilities". Fourth, what is missing in these theories (if anything).

1.- The Pragmatical Inheritances of Searle

There are three important antecedents to Searle's Speech Acts Theory, and the three have deep roots in the "Pragmatics" tradition, although Searle doesn't acknowledge this fact.

The first one is the notion that context diambiguates meaning in natural language. This we will refer to as the "contextual awareness" fact.

Second, the acknowledgement that language is used not only, not mainly and not predominantly to represent the world, but also to act. This we will call the "language acts" conception.

Finally, the realization that truth is but one of the many formalizable aspect of language. This we will call the "non-classical formal tradition".

We will discuss these three influences one by one.

Contextual Awarenes:

Indexicals, referential opacity, Stalnaker and Montague.

Literal sense and intended meaning. Grice's communicative rules. Implicatures.

Language Acts:

Wittgenstein and language games. Languace as a socially controlled activity, Language practice as following rules.

Austin on becoming competent in the use of language. Perfomative competence. Actors vs agents

The original lists of Wittgenstein and Austin. Austin's teleological distinctions, and the derived ontology.

Non-classical formal tradition:

Bar-Hillel, Montague, Hintikka, ... Possible world semantics, functions and modalities. Gazdar. Logic as Grammar.

2.- Speech Acts Theory and Illocutionary Logic

The basic forms of language acts. Declaratives, Committments, Propositions, Assessments.

Canonical forms, its components and implications.

Illocutants or speaker-hearer pairs.

Basic Illocutionary operations, and is Illocutionary Logic, logical?

The full list of acts and trees for english illocutions and operations

What Searle wanted to do and what he actually did.

3.- Conversations for Action, and Conversations to Create Possibilities

Flores' radical ecclecticism, a background notice.

A Conversation as sequence of spech acts. The original views. What Flores adds to Searle. Society as sets of (coordinatable) committments. Tools to communicate effectively. The Coordinator.

(Mis)-understanding Computers and Cognition. The Winograd factor, and AI, and Hubert Dreyfuss. The resonance of one book, and its sequels.

The idea of invention, speculation and creation of possibilities.

Beyond the Coordinator: Groupware, simplified conversations and the workflow paradigm.

4.- What they haven't seen

Impact of these ideas in Computer Science. Illocutionary based groupware and agent architectures and languages. Who is doing what today.

What is being done that is not a Searle descendent.

What have they missed.

Not, at least in the way we use the word "converse" in everyday language, nor in the technical way adopted by Flores and Winograd. But for a completely different kind of reasons.

Here I will try to structure a characterization of what I call a negotiation-as-a-dialogue concept, and show it encompasses most aspects of what people refer to as "negotiation", and will show how that is defferent from the notion of conversation. Second, I will explore Flores' notion of conversation, and what he calls the negotiation phase, and show it is essentially ellucidatory, and not ruly negotiational (in my terminology).

Finally, I will argue in favor of a systematic dialogical approach to the concept of negotiation, and its empirical consequences.

P Noriega and C. Sierra

IIIA

In this article we introduce the notion of computational environments that allow structured forms of negotiation among heterogeneous agents over a network.

In particular, we advance a theoretical framework for characterization, formalization and implementation of "negotiation environments" and show how this framework can be instantiated to create a typical negotiation environment:

1.- Motivation

Internet based agent negotiations.

a)Polimorphism of agents: human, reactive, deliberative...

Variety of trading conducts, diversity of markets, products, protocols participants.

But there is some underlying uniformity.

b) Usage. Explosion of free roaming agents will clog the net. Excesive advertisement and search activities, lack of guarantees, unstable protocols, difficult to standarize conventions.

Structured itrading and negotiation

Our proposal: Create A common context. Physical-virtual environment, shared conventions (ontology, language,...), Eligibility conditions for participants, simple central management, easy universal access, topic-instantiated general framework.

Will deal with only one example, the fish market model of bidding based negotiation, and comment on extensions.

2.- Basic Intuitions

What is a Fish Market.

site

purpose

participants (roles, expectations, values)

observed conducts (illocutionary exchanges vs. goods)

negotiation principles

negotiation protocols

admission (eligibility) criteria

solvency, bonding and commitment upholding

price, money and other values

strategy vs protocol

A first abstraction (a dialogical analogue)

illocution exchange

illocutionary particles

illocutonary content

real world correspondence to illocuted statements

role description

goal description

dialogical protocol

terinating conditions

3.- Dialogical Framework

Basic definitions (dialogue, dialogical context, dialogical framework,....)

Formal instantiation. The FM framework:

illocution exchange

illocutionary particles

illocutonary content

real world correspondence to illocuted statements

role description

goal description

dialogical protocol

commitment management and update

terminating conditions

Properties

4.- Implementational Issues

FM environment architecture

Trading Hall

Seller agents

Buyer agents

Callers and arbitrers

Agent admission board

Goods Clearinghouse

The trading dialogue language shell

Trading illocutions and tokens

Good-dependent linguistic elements

Valuation and vallidation

Turn-taking, rounds and bidding.

Terminating moves

Exception handling

The bidding sequence manager

Role script managers (definition, enablers, validators,...)

Agent eligibility tests and trading-habilitation procedures

Arbitration and sanctions

Network related shell

Hub management

Protocol conventions

External agent management

The economic-reality checker

Market-specific ontology

Risk and liability checks

Bonding and solvency issues

5.- The Fish-market model

Actual implementation of toy downward bidding FM.

Standards

Examples

6.- Discussion

What is nice about this example

What is too simplistic

What is impossible to achieve

Some economic and mathematical conjectures on sixe and diversity of traders, and their effect on prices

e.g CONJ1. "If there is a large number of buyers and sellers active, and the number of buyers is mkt-large w.r.t sellers, downward bidding benefits sellers. (and dual)"

7.- Extensions

More general price-based negotiation

Value-based trading (diversity of goods, unprecise descriptions, dynamic language trading)

General non crisp negotiation

Intermediated negotiation

Abstract

In practice, many multiagent interactions are "opaque", in the sense that relevant elements of the shared ontology and meaning are not known or established a priori, but only through the agents' interactions. While this condition may be unavoidable in many cases, it is not necessarily easy to contend formally or empirically with. Nevertheless, there are natural contexts where this "opacity" is --at most-- apparent, and apriori objective meaning and ontologies are not only available but effectively enforced. We call these environments "lucid"

In this paper we attempt to characterize the notion of a "lucid" context in which dialogical interactions among agents ocurr, and argue in favor of their formal treatment and the convenience of their computational interpretation.

1. Motivation

Intuition: What it means to be opaque and lucid

Opacity and situatedness are natural, but so are lucid contexts.

Empirical need for lucidity.

Convenience of a formal treatment of lucidity and artificial lcd's

Limits to lucidity

An overview of associated work

2. Towards a formal treatment of lucid dialogical contexts

A definition of lucid contexts

Relevant elements for a formal treatment

Context, illocutory exchanges, intensional comittments, dialogues, protocols and scenes

Some formal properties and nice examples

3. Computational models of lcd's

A proposal for an architecture

Ontological description

Satisfaction of illocutory comittments

Protocols for agent interactions

Scene description

I

4. Discussion

Why are they interesting

Pragmatic and theoretical issues:

Frame problem

Situations

Speech act issues. Conversations for action

1. Motivation

Intuition: What it means to be opaque and lucid

Opacity and situatedness are natural, but so are lucid contexts.

Empirical need for lucidity.

Convenience of a formal treatment of lucidity and artificial lcd's

Limits to lucidity

An overview of associated work

2. Towards a formal treatment of lucid dialogical contexts

A definition of lucid contexts

Relevant elements for a formal treatment

Context, illocutory exchanges, intensional comittments, dialogues, protocols and scenes

Some formal properties and nice examples

3. Computational models of lcd's

A proposal for an architecture

Ontological description

Satisfaction of illocutory comittments

Protocols for agent interactions

Scene description

4. Discussion

Why are they interesting

Pragmatic and theoretical issues:

Frame problem

Situations

Speech act issues. Conversations for action

Dialogues are pervasive and, to a large degree, unavoidable in everyday situations. Legal arguments, political debate, domestic disputes, didactic explanations, interviewing, psychotherapy, coordination of actions, negotiation, all tend to involve some form of dialogical interaction. But dialogues are also unavoidable since what is accomplished through them cannot be accomplished in a strictly "monological" setting, because some fundamental ontological, rethorical or epistemic features would be lost. Thus, dialogues may be worth studying, although they are not simple entities. Certainly not from a formal perspective.

On one hand, dialogues involve multiple participants, who exchange illocutions in rich and complex languages. Thus, classical --i.e. monological, truth-semantical, non-dynamic,-- formal devices, are inadequate to deal with these complexities. But in addition, dialogues are typically "opaque" (or "situated " or "unstructureded"), in the sense that In many dialogical situations meaning is not necessarily established in an objective, a priori, form; nor are interventions subject to an objective, a priori, clearly expressible protocol. In these cases, participants "react" to the illocutions, depending on the conditions or elements present in the given "context" or "situation". In typical dialogues, participants confirm, adjust, refine or establish their own meanings, intentions, beliefs and actions according to their individual interpretation and what the other participants are saying [c.f. Wittgenstein; Searle; McCarthy & Hayes; Winograd & Flores; Barwise & Perry].

The first kind of complexity may be addressed through ad-hoc "dialogical" structures [ c.f. Hamblin, Rescher, Hintikka, Carlson for different approaches], the second one may be sometimes avoided since the degree of "opacity" of dialogues can vary greatly.

Certainly there are situations which are inherently situational, such as psychotherapy or everyday conversation, where interpretational contexts are inevitably opaque, in the sense that meaning and commitments are mostly established through highly unstructured dialogical interactions. But then there are other contexts where a priori univocal shared interpretations ("transparency") is not only desired but enforced.

One example of enforced "transparency" is that of "structured negotiation environments" --such as auctions and other similar forms of mediated trading and negotiation. These naturally transparent contexts enforce univocity of interpretation, protocols and roles as a means for achieving efficiency, and guaranteeing acceptable conditions to participants. [c.f. North]

Take for instance the "downward-bidding" or "dutch" auction --so called because it is the way cut flowers have traditionally been traded in the Netherlands, as well as many other goods in many places [McAfee & McMillan 1987: 702]. In a dutch auction an extraordinarily simple negotiation convention is used: an intermediary, the main auctioneer, calls prices for a given item in a descending sequence, stopping when a bidder first indicates acceptance for the current price. But this simple process betrays a complex "institution" in which a "marketplace" with explicit and shared trading conventions --and meanings-- opens in a given "marketdate" when a specific group of sellers and buyers, submitting to those conventions, exchange a given collection of items under the supervision of market "intermediaries".

Some distinctive structured negotiation environments --auctions, stock markets, trading fairs, labor mediated-collective-bargaining-- have evolved over many years and are remarkably similar in structure and functionality across cultures and negotiated items. Being natural examples of transparent contexts, they are attractive candidates for formalization. They may also turn out to be interesting models for implementation of artificial agent-interaction environments. The kind of contexts, for instance, that may facilitate heterogeneous agents to negotiate. The kind of environments Internet is likely to adopt [Weld 1995: ???. Nosotros SARA].

In this paper we propose a formal approach to the description and modelling of these structured negotiation environments. We will give them a "dialogical" outlook by focusing on the illocutions that are uttered by participants in the negotiations --in the content of these illocutions and the illocutionary exchange process too. We draw our intuitions from the natural structured negotiation environments, such as auctions --and we will use the downward-bidding auction as an illustrative example throughout--; but the framework will be applicable to other forms of structured agent interactions, not only auctions, and not only negotiation.

Abstract

We present an outline of a dialogical notation that can be used to describe the dialogical processes that characterize lucid multiagent interactions such as are present in auctions and other structured negotiation environments. This notation has the advantage that if and when the description is adequate, it spawns a computational interpreter that constitutes a "model" for the description i.e. an actual implementation of a multiagent environment that corresponds to the description.

1. Background

1.1 Motivation

Lucid Dialogical contexts. Practical situations that correspond to..

Intuitive definition of lucid-dial-con

Why it may be interesting to automate them: pragmatical convenience robot-clog, abundance of dialogical contexts, test beds for agent interactions. Inherent complexity, apparent diversity, commonality of structure and components.

An ideal working solution, what should it amount to?

1.2 Our previous work:

Feasibility of adequate description

Convenience of a model-theoretic formalism

Intended computational interpretation

Actual implementations of FM

1.3 What we now want to accomplish

How we intend to generalize those previous efforts

Related work: Wokflow, Prolog, APL.

1.4 What we are going to do in this paper

We will show how it is possible to develop an executable notation for sialogical contexts. One that is rich enough to characterize, precisely, a variety of dialogical contexts (as long as these are "lucid"), and concrete enough to actually have implemented counterparts.

We will describe a "methodology" for description of lucid dial-cont, that will make explicit scenes, scene outlines and basic ontologies of a ldc.

We will then present the basic constructs of the notation that allow for the description to be coded into these constructs.

We show how the basic constructs interact as "building-blocks" whose compatibility conditions correspond to "felicitous termination" conditions, and to efffective composition of effective procedures.

We make precise the notions of an "adequate" description for an lcd, and that of correspondence between a description and an interpreter.

We prove that these constructs and their interpretations actually are sufficient to implement a well-defined class of lcd's.

We finally show how to describe some specific structured multiagent negotiation environments in this notation, and what the corresponding computational implementations are like.

Finally we discuss the significance of this work and suggest directions for further research.

Our work

Plato, or at least so claims Gadamer, used Dialogues not only as a literary device, but as a necessary form to "convince". Is there any difference.

I argue that the borders of Rethoric are fuzzy, at best, and in the worst case, they are the same as those of literary intention.

Dialogue is sometimes used to prove things, other times to convince or to ask, or to promote..i.e. in an intentional stance. But many times, dialogues are also used to clarify a definition. Thus as representational devices. As such, the Toronto and Erlagen schools have used them. That is why they missed out on agents.

Part One: Natural dialogue under artificial cicumstances

One can design situations where dialogical exchanges happen under controlled conditions. Examining these exchanges allow us to reify formal components that enrich our theoretical cogitations.

Examples:

Machine interviews: Eliza like dialogues

Negotiations

Polls and directed questioning

Facing contradictions, natural escapatories...

Part Two: Artificial Dialogues in their natural form.

Likewise, one can take specific real dialogues (transcriptions, parphrasings, evocative accounts,...) and subject them to the same kind of analysis.

Arcipreste de Hita

Classical arguments (Plato, Cicero)

Abstract

The Chiapas conflict has apparently been a long peace-negotiation dialogue, but in close analysis one can find that participants, roles, strategy and protocol have been shifting to accomodate quite different objectives over time.

In this note I attempt to analyze the dialogical elements that are present in the readily available printed documents --which have systematically invoked dialogical terms-- of this notorious conflict . I claim that a good analysis is possible, and that it may have some explanatory value. I don't want to claim that it may have any directive or predictive one. I think it may be interesting and worthwhile.

A discussion of practical situations in which dialogical talk is frequent. A careful assessment of what the intuitive notions detect, and what is missed or exagerated in naive views of dialogue.

Abstract (orig. 6.9.96)

What is a ÒprotocolÓ? How does one define a role? What are contexts? A dialogicaly biased approach to classical notions. I want to make clear the notion of "protocolÓ. I believe it needs many subsidiary notions, and that a dialogical outlook may prove illuminating. Thus I will try to focus on issues that are pertinent to individual agents, and those that need to be shared. Those that are obligatory and those that may or may not happen, and so on.

First let's fix a "Context"

A context is a set of agent types, rules (or restrictions) that have to be satisfied by those agents. A timing convention, a well defined set of actions and illocutions (dialogical framework), and possibly other objects. (Location, permanence,...).

Is an "institution" a better concept? I don't think so. Situation? Frame? (Frame problem)

An agent is a (complex, layerd) theory. Agents speak and hear, and may have an internal life. Ocasionally they may perceive through other sensors.."dialogical entities"

A role is a set of restrictions (Rules?) that have to be satisfied by agents. It involves a protocol?. I would prefer to separate them: a "buyer" agent fits a role of a buyer (in general), in a given "context" it has to adapt to that context's specific constraints, such as the protocols --e.g. bidding-- that are used in that context. It is conceivable that a general-purpose agent is tailored to play the role of a buyer, and further taylored to play a succesful buyer in the Fishmarket context. Is that wise? Would one end up with too general an agent? From an empirical perspective the question is irrelevant, I may need a buyer agent for the fishmarket, and I may very well construct an ad-hoc agent, or even just an "agentoid" (a simple interface that allows a human user to participate in a multiagent system,... In such a situation, role-agency and context are indossociable.

A protocol is a socially accepted (possibly non-deterministic) sequencing of illocutions whose standard commitments (obligations?) are to be satisfied by all participating agents in a given context.

I think protocols can be formalized, or at least characterized with some rigour, and one would like to be able to prove that certain properties hold or cannot hold in a given protocol (and its implementation). Eg "fairness", or the "premature bid problem"...

Exemplify with FM, DBA and other similar protocols.

....from Julian Padget conversation: ---------------------Begin JAP DBP ----------

----------------------------End

Questions: Are there many conditions (analogous to pbp) that should be avoided or satisfied by a given protocol? How can one express carefully, or formally, the underlying entailments of "fairness", "transparency",etc. that a market should uphold? How can you prove that a given protocol (or its implementation) satisfies such conditions or entailments? What should a protocol be?

In real-world conditions, how reliable are reliable networks? Are they usually accepted as reliable by users or is reliability user-dependent, link-dependent, or protocol-dependent? Is there a probability-of-failure kind of measure always associated (to a network, protocol)? If so, should the auction protocol implementation be made more robust IN PRACTICE by adding components that reduce such probability (e.g. have a "witnessing" auctioneer that may also hear a bid, and who could be consulted when failures ocurr, or are suspected)?

-----------------end

And Julian's comments:

-------------------Begin

2.- Julian on "reliability".