Changing of Participants’ Attitudes in

Argument-based Negotiation

Mare Koit

Institute of Computer Science, University of Tartu, Liivi 2, Tartu, Estonia

Keywords: Negotiation, Argument, Dialogue Model, Reasoning, Attitude, Knowledge Representation.

Abstract: We are modelling argument-based negotiation where the initiator is convincing the partner to do an action.

The initiator is using a partner model which evaluates the hypothetical attitudes of the partner related to the

action under consideration. The partner when reasoning operates with an actual model – the actual attitudes

which still are hidden from the initiator. Both models are changing during negotiation as influenced by the

presented arguments. The choice of an argument by a negotiation participant depends, on one hand, on the

attitudes related to the action, and on the other hand, on the result of reasoning based on these attitudes. The

paper studies how the participants are changing their attitudes during a dialogue. A human-human dialogue

illustrates the results of the analysis of a small dialogue corpus. A limited version of the model is

implemented on the computer.

1 INTRODUCTION

Negotiation is a form of interaction in which a group

of agents, with a desire to cooperate but with

potentially conflicting interests try to come to a

mutually acceptable division of a scarce resource or

resources (Rahwan et al., 2003). Negotiation

dialogues are aimed at reaching an agreement

between participants when there is a perceived

divergence of interest. However, the participants are

also cooperative, at least to the extent that they are

willing to enter into joint interaction to agree on a

division of the resource at issue (DeVault et al.,

2015).

Argumentation-based negotiation is the process

of decision-making through the exchange of

arguments (Lewis et al., 2017).

According to Scherer’s typology, attitudes are

relatively enduring, affectively coloured beliefs,

preferences, and predispositions towards objects or

persons (Scherer, 2000). Attitude is a psychological

tendency that is expressed by evaluating a particular

entity with some degree of favour or disfavour.

Attitudes refer to people’s evaluations of entities in

their world. Attitudes often serve a key mediational

role in behaviour change i.e., attitude change can

mediate the impact of some influence treatment on

behavioural compliance (Petty and Briñol, 2015).

Our aim is to develop a dialogue system (DS)

which interacts with the user in a natural language

following norms and rules of human

communication. For that reason, we study human-

human spoken dialogues. We have worked out a

formal model of negotiation dialogue (Koit and

Õim, 2014; Koit, 2018a) which includes a reasoning

model and communicative strategies applied by the

participants in order to achieve their communicative

goals. Two kinds of attitudes of communication

participants have been introduced in the model –

respectively, related to a communication partner and

related to a negotiation object, in our case, an action

(Koit, 2018b).

Communicative space has been determined in

order to model the attitudes related to

communication partners. In the current paper, we

will concentrate on reasoning of communication

participants and accordingly, on their attitudes

related to a negotiation object. We analyse human-

human argumentation dialogues where the initiator

is convincing her partner to make a decision to do an

action, i.e. the negotiation object is an action. We

study how the participants can influence each other

when negotiating and how they change their

attitudes related to the action under consideration

using arguments. The only possibility is to analyse

the wording of the people not the insight.

770

Koit, M.

Changing of Participants’ Attitudes in Argument-based Negotiation.

DOI: 10.5220/0007472807700777

In Proceedings of the 11th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2019), pages 770-777

ISBN: 978-989-758-350-6

 2019 by SCITEPRESS – Science and Technology Publications, Lda. All r ights reserved

The remainder of the paper is structured as

follows. Section 2 introduces our dialogue model

which includes a reasoning model about doing an

action. Attitudes of a communication participant in

relation to the action will be represented as

coordinates of the vector of motivational sphere of

the reasoning subject. The attitudes are changing in

dialogue as influenced by the arguments of

communication participants. So far, we have applied

the reasoning model on simple artificial dialogues.

In this paper, we are aiming to evaluate the model on

actual human-human dialogues. For that, we analyse

a dialogue corpus which will be introduced in

Section 3. Section 4 presents a case study – a

dialogue example from the corpus which

demonstrates how the reasoning model describes

changing of the attitudes related to the negotiation

object. Section 5 discusses the reasoning model and

introduces the implemented DS. Section 6 draws

conclusions.

2 DIALOGUE MODEL

We are modelling negotiations between two

participants A and B in a natural language. One of

them (let it be A) initiates the dialogue by requesting

her partner B to agree to do an action D. If B refuses

then A in negotiation tries to influence him by

presenting various arguments for doing D. The

arguments are based on the partner model – the

image A has about B’s attitudes related to different

aspects of the action D. The partner B, in his turn,

may present counterarguments based on his actual

attitudes. The counterarguments show which beliefs

of A about B’s attitudes were wrong and therefore,

how A has to change her partner model. The

dialogue finishes with B’s decision: to do D or not.

Depending on the decision, A either has achieved or

not her initial communicative goal.

2.1 Reasoning Model

After A has made a proposal or request to the partner

B to do the action D, B can respond with agreement

or rejection, depending on the result of his

reasoning. Rejection can be (but not necessarily)

supported with an argument against doing D. These

arguments can be used by A as giving information

about the reasoning process that brought B to his

decision. Therefore, a reasoning model should be

included into the dialogue model.

There are various formal approaches to

reasoning, e.g. the Elaboration Likelihood Model

(Cacioppo et al., 1986; Petty et al., 2018), Social

Judgment Theory, Social Impact Theory, etc.

Nevertheless, we use a naïve, ‘folk’ theory in our

reasoning model (D’Andrade, 1987; Davies and

Stone, 1995; Õim, 1996). Our model is based on the

studies in the common-sense conception of how the

human mind works in such situations. The general

principles of the model are analogous to the BDI

(Belief-Desire-Intention) model (Grosz and Sidner,

1986; Allen 1995; Boella and van der Torre, 2003)

but it has some specific traits (cf. Koit and Õim,

2014).

First, along with desires we also consider other

kinds of motivational inputs for creating the

intention to do an action in a reasoning subject (e.g.

whether the subject considers the action pleasant or

useful to him/her or s/he is forced to do it

independent on his/her immediate wish).

Secondly, we suppose that people start, as a rule,

from this conception, not from any consciously

chosen scientific one. We want to model a naïve

‘theory’ that people themselves use when they are

interacting with other people and trying to change

their attitudes, to predict and influence their

decisions.

Our reasoning model consists of two parts

including a model of human motivational sphere,

and reasoning procedures.

(1) We represent the model of motivational

sphere of a reasoning subject by the following

vector of attitudes related to the reasoning object –

the action D:

= (w(resources

), w(pleasant

w(unpleasant

), w(useful

), w(harmful

w(obligatory

), w(punishment-not

), w(prohibited

w(punishment-do

)).

We suppose in our model that the attitudes have

numerical values (weights). Here w(pleasant

), etc.

mean the weight of pleasant, etc. aspects of D;

w(punishment-not

) – the weight of the punishment

for not doing D if it is obligatory; w(punishment-

) – the weight of the punishment for doing D if it

is prohibited. Further, w(resources

) = 1 if the

subject has all the resources necessary to do D

(otherwise 0); w(obligatory

)/w(prohibited

) = 1 if

D is obligatory/prohibited for the reasoning subject

(otherwise 0). The values of other weights can be

non-negative natural numbers on the scale from 0 to

10.

Some comments are necessary here. Definitely,

people do not operate with numerical weights in

their reasoning. Instead, they rather use words of a

natural language to characterize the attitudes. For

Changing of Participants’ Attitudes in Argument-based Negotiation

771

example, the pleasantness of an action can be

evaluated by such words and expressions as

excellent, very pleasant, etc. Still, the words can

approximately be represented on a numerical scale.

Instead, fuzzy logic can be used. Further, the aspects

of actions considered here are not fully independent.

For example, harmful consequences of an action as a

rule are unpleasant for a subject (but unpleasant will

not always be harmful). However, we do not assume

the independence of the aspects in the reasoning

process.

(2) The second part of the reasoning model

consists of reasoning procedures that regulate, as we

suppose, human action-oriented reasoning.

According to our model, the reasoning process

can be triggered by three main types of determinants

– wish-, needed- and must-determinant (Õim, 1996).

The process itself consists of a sequence of steps

where such aspects participate as resources of the

reasoning subject for doing D, positive aspects of D

or its consequences (pleasantness, usefulness, and

also punishment for not doing D if it is obligatory),

and negative aspects (unpleasantness, harmfulness,

and punishment for doing D if it is prohibited).

There are three reasoning procedures in our

model (WISH, NEEDED, and MUST) which

depend on the determinant that triggers the

reasoning. Each procedure represents the steps that a

subject goes through in the reasoning process when

comparing and summarizing weights of different

aspects of D, and the result is the decision: to do D

or not.

The reasoning procedures include some

principles which represent the interactions between

the determinants, e.g.

•people want pleasant states and do not want the

unpleasant ones

•if the sum of the values of the internal (wish- and

needed-) determinants and the value of the external

(must-) determinant appear equal in a situation then

the decision suggested by the internal determinants

is preferred.

As an example, let us present the reasoning

procedure WISH as a step-form algorithm in Fig. 1

triggered by the wish of the reasoning subject to do

D, that is, D is not less pleasant than unpleasant for

the subject, cf. (Koit, 2016). Here we do not indicate

the action D which remains the same during the

reasoning.

If D is not less useful than harmful then the

reasoning procedure NEEDED can be triggered by

the reasoning subject. Finally, if D is obligatory then

the subject can trigger the reasoning procedure

MUST, cf. (Koit and Õim, 2014). When reasoning, a

subject applies the procedures in a certain order as

motivated by the internal or external determinants.

First of all, s/he tries to apply the procedure WISH.

If it is impossible (the presumption is not fulfilled)

or it gives the decision “do not do D” then the

subject applies the procedure NEEDED and finally,

the procedure MUST, until the decision (do D or

not) is achieved (Koit and Õim, 2014).

Presumption: w(pleasant)



w(unpleasant).

1) Is w(resources) = 1? If not then go to 11.

2) Is w(pleasant) > w(unpleasant) +

w(harmful)? If not then go to 6.

3) Is w(prohibited) = 1? If not then go to 10.

4) Is w(pleasant) > w(unpleasant) + w(harmful)

+ w(punishment-do)? If yes then go to 10.

5) Is w(pleasant) + w(useful) > w(unpleasant) +

w(harmful) + w(punishment-do)? If yes then go to

10 else go to 11.

6) Is w(pleasant) + w(useful)



w(unpleasant) +

w(harmful)? If not then go to 9.

7) Is w(obligatory) = 1? If not then go to 11.

8) Is w(pleasant) + w(useful) + w(punishment-

not) > w(unpleasant) + w(harmful)? If yes then go

to 10 else go to 11.

9) Is w(prohibited) = 1? If yes then go to 5.

10) Decide: do D. End.

11) Decide: do not do D.

Figure 1: The reasoning procedure WISH.

We use two vectors of attitudes – w

and w

–

in our dialogue model. Here w

is the model of

motivational sphere of B who has to make a decision

about doing D; the vector includes B’s (actual)

evaluations of D’s aspects and it is used by B when

he is reasoning about doing D. The other vector w

is the partner model which includes A’s beliefs

concerning B’s attitudes (the hypothetical

evaluations) and it is used by A when she is planning

her next turn in dialogue. Both models w

and w

are changing as influenced by the arguments

presented by the participants in negotiation.

2.2 Communicative Strategies and

Tactics

A communicative strategy is an algorithm used by a

participant for achieving his/her communicative goal

(Koit and Õim, 2014; Koit, 2018a). The initiator A

when having a communicative goal to convince B to

make a decision to do D can realize her

communicative strategy in different ways, e.g. she

can entice, persuade or threaten the partner B to do

ICAART 2019 - 11th International Conference on Agents and Artiﬁcial Intelligence

772

D. Respectively, she stresses the pleasantness or

usefulness of doing D or punishment for not doing D

if it is obligatory. We call these ways of realization

of a communicative strategy communicative tactics.

B similarly applies his communicative strategy

through related communicative tactics. Some

algorithms are presented in (Koit, 2018a).

The initiator A chooses a suitable communicative

strategy and the communicative tactics in order to

direct B’s reasoning to the desirable decision. When

trying to influence B to make the pursued decision

(do the action D) and to change his initial attitudes

(the model w

), A uses a partner model w

. A

stresses the positive and downgrades the negative

aspects of the action. Various arguments for

doing/not doing D will be presented by the

participants in a systematic way. While enticing

(respectively, persuading or threatening) the partner

B for doing D, A attempts to trigger the reasoning

procedure WISH (respectively, NEEDED or MUST)

in B’s mind (Koit and Õim, 2014).

3 EMPIRICAL MATERIAL

The current study is based on the Estonian dialogue

corpus (Hennoste et al., 2008). The main part of the

corpus is formed by transcripts of human-human

dialogues recorded in authentic situations. Among

them are phone calls (travel negotiations,

telemarketing calls, directory inquiries, etc.) as well

as face-to-face conversations, in total 1056

transliterated texts (206,485 tokens).

For this study, a small sub-corpus consisting of

five everyday phone calls between acquaintances

has been chosen from the corpus. In the dialogues,

participants are negotiating about doing an action by

one of them. We will consider how the participants

are reasoning in order to make their decisions about

the action and how they are influencing the partner

to change his/her attitudes related to the action.

However, direct access to their minds is impossible.

Instead, we can make conclusions only by analysing

their utterances – a dialogue (text).

In order to describe the reasoning processes of

the participants, we use the models w

and w

and the reasoning procedures introduced in the

previous section. We are wondering how well the

models describe authentic human-human dialogues

and whether they can be used when developing a DS

which interacts with the user ‘like a human’.

The initiator A, starting a dialogue, generates a

partner model w

(using her preliminary

knowledge) and determines the communicative

tactics T

which she will use (e.g. enticement), i.e.

she accordingly fixes a reasoning procedure R

which she aims to trigger in B’s mind (e.g. WISH).

A applies the reasoning procedure in her partner

model, in order to ‘put herself’ into B’s role and to

choose suitable arguments when convincing B to

make a decision to do D.

B has his own model – the vector w

(the exact

values of which coordinates A does not know

similarly like w

is not directly accessible for B).

He in his turn determines a reasoning procedure R

which he will use in order to make a decision about

doing D (the procedure can be different from R

fixed by A) and his communicative tactics T

4 CHANGING THE ATTITUDES:

A CASE STUDY

Let us consider an example from our analysed sub-

corpus in order to demonstrate how both models of

motivational sphere are used in a reasoning process

and how the attitudes of participants captured in the

models are changing.

The following dialogue is a transcript of a phone

call of mother (participant A) to her son (participant

B). A makes a proposal to B to bake gingersnaps (the

action D) and presents a lot of arguments during the

dialogue in order to produce/increase B’s wish to do

the action, until B finally agrees. Transcription of

Conversation Analysis is used in the example

(Sidnell and Stivers, 2012). In the following, we do

not indicate the action D in the vectors w

and w

because it remains the same.

Let us suppose that mother A (knowing her son

B) has created the following partner model: w

(1,6,2,1,1,1,0,0,0), i.e. A believes that B will have all

the resources to bake gingersnaps (the value of the

first coordinate equals to 1), further, the action is

much more pleasant (6) than unpleasant (2) for B, it

is useful for B because gingersnaps will be prepared,

and similarly harm because it needs time (both

values 1), obligatory (1) for B because son is obliged

to fulfil mother’s request, but no punishment (0) will

follow if B will not agree; the action is not

prohibited (0) and therefore no punishment (0) will

follow when doing it. The coordinates of the vector

should be empirically confirmed, based on A’s

preliminary knowledge about B. (Still, an external

observer can hardly ever determine these values

exactly, only analysing the dialogue. Similarly, as

already said above, people do not operate with exact

numerical values when reasoning.)

Changing of Participants’ Attitudes in Argument-based Negotiation

773

We further suppose that mother A will entice her

son B, assuming that B wants to do D. This

assumption is confirmed by the following dialogue

analysis – all the arguments presented by A increase

the pleasantness of the action. (Still, we carry out an

informal analysis here; the automatic analysis of

utterances with the aim to determine the certain

aspects of the action that they influence, remains for

the further work.) The reasoning procedure WISH

applied by A in the initial partner model gives the

decision „do D“ (cf. Fig.1, steps 1, 2, 3, 10).

Therefore, A makes a proposal, optimistically

looking for B’s agreement:

/---/

A: ´küsimus.

A question.

(0.6) .hhhhh kas sulle pakuks ´pinget ´piparkookide

´küpsetamine.

Do you like to bake gingersnaps?

(1.7)

B: ´praegu.

Just now?

(0.6)

A: jah.

Yes.

(0.6)

Let us further suppose that for B, the initial model

= (1,1,5,2,1,1,1,0,0), i.e. the resources exist

(value 1), the action is much more unpleasant (5)

than pleasant (1) but nonetheless, more useful (2)

than harm (1) for him. Therefore, B’s initial attitudes

are quite different as compared with A’s guesses.

(Again, here we evaluate the coordinates/attitudes

only approximately, by an informal analysis of the

dialogue.) Thus, B does not want to do D because its

pleansantness is smaller than the unpleasantness (on

the contrary to A’s supposition). Based on w

, B

cannot trigger the reasoning procedure WISH in his

mind because the assumption of the procedure is not

fulfilled (cf. Fig. 1). However, he can trigger the

procedure NEEDED (because D is more useful than

harmful for him) which still gives the decision „do

not do D“ as demonstrated by B’s next utterance:

B: .hhhhhhh ma=i=´tea vist ´mitte.

I don’t know, perhaps not.

As follows from B’s refusal, A has to update the

partner model. The updated model will be w

= (1,6

2,2,1,1,1,0,0,0) because it should give the decision

„do not do D“ by applying the reasoning procedure

WISH (valid for A) like B got (cf. Fig. 1, steps 1, 2,

6, 7, 8, 11). Here A supposes that B applies this same

reasoning procedure (which actually is not the case)

and she does not change her communicative tactics

(enticement).

Now A presents an argument for increasing the

pleasantness:

A: ja=sis gla´suurimine=ja=´nii.

And then glazing and so on.

(0.6)

At the same time, she increases the value of the

pleasantness in her partner model. We suppose (in

our implementation) that every argument increases

(respectively, decreases) the targeted value by one

unit (by 1). Thus, we consider all the arguments to

be equal, having the value/weight 1 (which still is a

simplification although in the reality, the arguments

could have different weights). New partner model

will be w

= (1,2 3,2,1,1,1,0,0,0). The reasoning

procedure WISH gives the decision „do D“ in this

model therefore A is again looking for B’s

agreement.

As influenced by A’s argument, B in his turn

increases the value of the pleasantness (by 1) in his

model: w

= (1,1+1=2,5,2,1,1,1,0,0). B continuosly

applies the reasoning procedure NEEDED which

again gives the decision „do not do D“:

B: ´ei, ´ei, ´ei ei=´ei.

No, no, no, no, no.

(0.9)

Based on B’s rejection, A has to update the partner

model: w

= (1,3 2,2,1,1,1,0,0,0). Now the

reasoning procedure WISH applied by A gives the

decision „do not do D“ like B got. When enticing, A

once more increases the pleasantness presenting the

following argument:

A: me saaksime nad ´vanaema=jurde ´kaasa võtta.

We can take them with us when going

to visit grandmother.

(0.4)

New partner model is w

= (1, 2 3,2,1,1,1,0,0,0), the

reasoning procedure WISH gives the decision „do

D“.

Influenced by the presented argument, B changes

his attitude about the pleasantness, after that w

(1,2+1=3,5,2,1,1,1,0,0). The reasoning procedure

NEEDED, continuosly applied by B, gives the

decision „do not do D“:

B: ´präägu ei=´taha.

I don’t want.

(1.3)

A once more has to decrease the pleasantness in her

partner model, getting w

= (1, 3 2,2,1,1,1,0,0,0)

where the procedure WISH gives the result „do not

do D.

ICAART 2019 - 11th International Conference on Agents and Artiﬁcial Intelligence

774

B: aga (.) noh, kas sa mõtled nagu .hhh kui sa tuled

´koju=vä.

But what do you think – after you

come home?

A: .hhh ei

No.

ma mõtlen: kui mind kodus ei=´ole.

I think, when I’m not home.

The argument presented by A (I’m not home)

implies w

= (1, 2 3,2,1,1,1,0,0,0) and the decision

will be „do D“. This argument increases the

pleasantness of D also for B (he obviously likes to

act alone, while his mother is not home). After the

update, w

= (1,3+1=4,5,2,1,1,1,0,0), and the

procedure NEEDED finally gives the result „do D“

– B agrees:

B: aa.

Aha.

(0.5) .hhh et ´lähen ostan ´tainast=vä.

Then I’ll go to buy paste, yes?

A: ja=niimodi=jah,

Yes, right.

(1.4)

Nevertheless, the pleasantness of D is less than the

unpleasantness in w

, therefore B does not even now

want to do D, but he only takes it as needed (more

useful than harmful). A presents her next argument

for the pleasantness:

.hhh sinna:: ´Pereleiva ´kohvikusse võiksid minna @

´võiksid seal endale ühe ´kohvi lubada=ja @ (2.7) teha

ostmise ´mõnusaks=ja (0.8) ja=siis tulla ´koju=ja? (1.7)

´piparkooke teha=ja

And you could go to Pereleiva cafe

and take a coffee in order to make

buying pleasant for you, and then go

home to bake gingersnaps.

(1.2)

The reasoning procedure WISH gives „do D“ as

before in the updated partner model w

= (1, 3

4,2,1,1,1,0,0,0). B similarly updates his model:

=(1,4+1=5,5,2,1,1,1,0,0). Now the pleasantness

equals to the unpleasantness therefore B started to

want to do D. He can yet apply the reasoning

procedure WISH. The result will be „do D“ (cf.

Fig.1, steps 1, 2, 6, 9, 10):

B: okei?

OK.

/---/

(Actually, both procedures NEEDED and WISH

give the same positive decision in w

. B prefers to

apply the procedure WISH.) However, A does not

finish the call but she presents an additional (the

last) argument in order to increase B’s wish once

A: .hhhhhhhhhh (0.2) ja ´siis ma tahtsin sulle öelda=et

´külmkapis on: ´sulatatud või tähendab=ned ´külmutatud

ja ´ülessulanud ´maasikad ja ´vaarika´mömm.=hh

And I wanted to tell you that there

are frozen strawberries and raspberries

in the icebox.

B: jah

Yes.

(0.3)

A: palun ´paku endale sealt.

Please help yourself.

/---/

After this argument, both models will change:

= (1, 4 5,2,1,1,1,0,0,0)

=(1,5+1=6,5,2,1,1,1,0,0).

In both models, the reasoning procedure WISH

gives the final decision „do D“.

Only one attitude (the pleasantness) is changing

in the models w

and w

during negotiation. That is

because A over and over again presents the

arguments for the pleasantness of the action for B,

i.e. she continuosly applies the communicative

tactics of enticing by trying to trigger the reasoning

procedure WISH in B’s mind. The partner B in the

beginning of the negotiation does not want to do D

but he takes it only useful. When reasoning, B uses

the procedure NEEDED. Nevertheless, A’s

arguments increase the pleasantness of D for B to

such an extent that the wish to do D arises in his

mind: the pleasantness finally becomes equal to the

unpleasantness which makes it possible to trigger the

reasoning procedure WISH. The balance of the other

weigths in w

contribute to achieve the result of the

reasoning „do D“. This final result is the same in

both models w

and w

regardless of their

difference in the beginning as well as in the end of

the negotiation.

5 DISCUSSION

The corpus analysis demonstrates that our dialogue

model can be used when describing actual human-

human dialogues. A big challenge when applying it

for the dialogue analysis has been creating of initial

models w

and w

. It is hard to determine, only

based on a dialogue text, such models that

adequately describe the attitudes and attitude

changes of the participants during a dialogue.

Another problem is recognition of the

communicative strategies and tactics applied by the

Changing of Participants’ Attitudes in Argument-based Negotiation

775

participants. This needs the linguistic analysis of

utterances in order to understand which aspect of the

action (i.e. the negotiation object) is affected by a

certain utterance (e.g. the pleasantness,

unpleasantness, etc.).

Still, here our primary aim is not the automatic

analysis of dialogues in a natural language but rather

we want to design and develop a DS which follows

norms and regulations of human communication.

When reasoning about doing an action, a subject

is weighing and comparing different aspects of the

action (the availability of resources, its pleasantness,

usefulness, etc.) which are captured in his/her model

of motivational sphere as attitudes.

When attempting to direct B’s reasoning to the

desirable decision (“do D” in our case), A presents

several arguments stressing the positive and

downgrading the negative aspects of D. The choice

of A’s argument is based on one hand, on the partner

model and on the other hand, on the (counter)

argument presented by the partner. Still, B is not

obliged to present a counterargument but he can

simply refuse to do the proposed action if his

reasoning gives a negative decision (like in the

considered example). When choosing the next

argument for D, A triggers a reasoning procedure in

her partner model depending on the chosen

communicative tactics, in order to be sure that the

reasoning will give a positive decision after

presenting this argument. B himself can use the same

or a different reasoning procedure triggering it in his

own model. After the updates made both by A and B

in the two models during a dialogue (A’s model of B

will be updated by A, and B’s model of himself will

be updated by B), the models will approach each to

another but, in general, do not equalize. Although

the results of reasoning in both models can be equal,

as demonstrated the example considered in the

previous section. Therefore, A can convince B to do

D even if not having a complete picture of him.

Our dialogue model considers only a limited

kind of dialogues but although, it illustrates the

situation where the dialogue participants are able to

change their attitudes related to the negotiation

object (doing an action) and bring them closer one to

another by using arguments. The initiator A does not

need to know whether the counterarguments

presented by the partner B have been caused by B’s

opposite initial goal or are there simply obstacles

before their common goal and can be eliminated by

A’s arguments. A’s goal, on the contrary is not

hidden from B. Secondly, as said in Section 2.2, the

different communicative tactics used by A are aimed

to trigger different reasoning procedures in B’s

mind. A can fail to trigger the pursued reasoning

procedure in B but however, she can achieve her

communicative goal when having a sufficient

number of statements for supporting her initial goal.

In the considered example (Section 4), A finally

succeeded to trigger the desirable reasoning

procedure and achieved her communicative goal.

We have implemented the model of negotiation

as a simple DS where the computer plays A’s and

the user B’s role. The participants are interacting in

written Estonian. The computer uses ready-made

sentences for presenting arguments but the user can

optionally use another set of ready-made sentences

or also put in free texts which include specific

keywords or key phrases. Based on the

implementation, we can study how attitudes of the

participants are changing in argumentation dialogue.

6 CONCLUSIONS

We are considering the dialogues where the

participants A and B negotiate doing an action D by

B. Their initial communicative goals can conform or

be opposite. They are presenting arguments for and

against doing D, in order to achieve their goals. The

arguments take into account the counterarguments

presented by the partner. In addition, A’s arguments

are based on her partner model whilst B’s arguments

are based on his model of himself. Both models

include the attitudes related to the availability of the

resources, positive and negative aspects of doing D

which have numerical values in our implementation.

Both models are changing during negotiation. We

study how the models are updated in a dialogue, and

track the changes.

We have worked out a model of argument-based

negotiation which includes a reasoning model. When

reasoning about doing an action, the subject is

weighing, summarizing and comparing different

aspects of the action under consideration. If the

positive aspects weigh more than negative then the

decision will be “do the action” otherwise “do not do

it”.

We have implemented the model of negotiation

as a simple DS. Our future work includes

development of the implementation by adding text

processing tools to DS in order to achieve more

human-like interaction of a user with the system.

ICAART 2019 - 11th International Conference on Agents and Artiﬁcial Intelligence

776

ACKNOWLEDGEMENTS

This work was supported by institutional research

funding IUT (20-56) of the Estonian Ministry of

Education and Research, and by the European Union

through the European Regional Development Fund

(Centre of Excellence in Estonian Studies). The

author is also very thankful to anonymous reviewers

for their valuable comments and suggestions.

REFERENCES

Allen, J., 1995. Natural Language Understanding. 2nd ed.

The Benjamin/Cummings Publ. Comp., Inc.

Boella, G., van der Torre, L., 2003. BDI and BOID

Argumentation. In Proc. of CMNA-03. The 3rd

Workshop on Computational Models of Natural

Argument at IJCAI-2003, Acapulco. Available online

at www.cmna.info

Cacioppo, J.D., Petty, R.E., Kao, C.F., Rodriguez, R.,

1986. Central and peripheral routes to persuasion: An

individual difference perspective. In Journal of

Personality and Social Psychology, 51:1032–1043.

D’Andrade, R., 1987. A folk model of the mind. In

Cultural models of Language and thought, D. Holland

and A. Quinn (Eds.), 112–148. London: Cambridge

University Press.

Davies, M., Stone, T., eds., 1995. Folk psychology: the

theory of mind debate. Oxford, Cambridge,

Massachusetts: Blackwell. ISBN 978-0-631-19515-3.

DeVault, D., Mell, J., Gratch, J., 2015. Toward natural

turn-taking in a virtual human negotiation agent. In

AAAI Spring Symposium on Turn-taking and

Coordination in Human-Machine Interaction, 9 p.

AAAI Press, Stanford, CA.

Grosz, B., Sidner, C.L., 1986. Attention, intentions, and

the structure of discourse. In Computational

Linguistics, 12(3), 175–204.

Hennoste, T., Gerassimenko, O., Kasterpalu, R., Koit, M.,

Rääbis, A., Strandson, K., 2008. From human

communication to intelligent user interfaces: corpora

of spoken Estonian. In Proc. of LREC, European

Language Resources Association (ELRA). Marrakech,

Morocco, 2025–2032. Available online at www.lrec-

conf.org/proceedings/¬lrec2008

Koit, M., 2018a. Reasoning and communicative strategies

in a model of argument-based negotiation. In Journal

of Information and Telecommunication TJIT, vol. 2,

14 p. Taylor & Francis Online.

https://doi.org/10.1080/24751839.2018.1448504

Koit, M., 2018b. Modelling attitudes of dialogue

participants – reasoning and communicative space. In

A.P. Rocha and J. van den Herik, eds. Proc. of the

International Conference on Agents and Artificial

Intelligence ICAART’18, vol. 2, 581−588. Portugal,

SciTePress.

Koit, M., 2016. Influencing the beliefs of a dialogue

partner. In Christo Dichev, Gennady Agre, eds.

Artificial Intelligence: Methodology, Systems, and

Applications. Lecture Notes in Computer Science:

17th International Conference, AIMSA. Varna,

Bulgaria. Springer International Publishing, 216−225.

(LNAI 9883). https://doi.org/10.1007/978-3-319-

44748-3.

Koit, M., Õim, H., 2014. A computational model of

argumentation in agreement negotiation processes. In

Argument & Computation, 5, 209−236,

https://doi.org/10.1080/19462166.2014.915233

Lewis, M., Yarats, D., Dauphin, Y.N., Parikh, D., Batra,

D., 2017. Deal or no deal? End-to-end learning for

negotiation dialogues. In Proc. of the Conference on

Empirical Methods in Natural Language Processing,

2443–2453. Copenhagen, Denmark, Assoc. for

Computational Linguistics.

Õim, H., 1996. Naïve Theories and Communicative

Competence: Reasoning in Communication. In

Estonian in the Changing World, 211–231. University

of Tartu Press.

Petty, R.E., Briñol, P., 2015. Processes of social influence

through attitude change. In E. Borgida and J. Bargh,

eds. APA Handbook of Personality and Social

Psychology (Vol.1): Attitudes and social cognition,

509–545. Washington, D. C.: APA Books.

Petty, R.E., Briñol, P., Teeny, J., Horcajo, J., 2018. The

elaboration likelihood model: Changing attitudes

toward exercising and beyond. In B. Jackson, J.

Dimmock, and J. Compton, eds. Persuasion and

communication in sport, exercise, and physical

activity, 22–37. Abington, UK: Routledge.

Rahwan, I., McBurney, P., Sonenberg, L., 2003. Towards

a theory of negotiation strategy (a preliminary report).

In S. Parsons and P. Gmytrasiewicz, eds. Game-

Theoretic and Decision-Theoretic Agents (GTDT).

Proc. of an AAMAS-2003 Workshop, 73–80.

Melbourne, Australia.

Scherer, K.R., 2000. Psychological models of emotion. In

J. C. Borod, ed. The neuropsychology of emotion,

137–162. Oxford.

Sidnell, J., Stivers, T., eds. 2012. Handbook of

Conversation Analysis. Boston: Wiley-Blackwell.

ISBN 978-1-4443-3208-7

Changing of Participants’ Attitudes in Argument-based Negotiation

777