Modelling Attitudes of Dialogue Participants

Reasoning and Communicative Space

Mare Koit

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

Keywords: Reasoning, Communicative Space, Dialogue Corpus, Knowledge Representation.

Abstract: The paper introduces a dialogue model, concentrating on attitudes of dialogue participants. Two kinds of

attitudes are under consideration: (1) attitudes related to different aspects of a negotiation object (in our

case, doing an action) which direct reasoning in communication, and (2) attitudes related to a

communication partner (dominance-subordination, cooperation-antagonism, communicative distance, etc.)

which are modelled by using the concept of communicative space. Telemarketing calls in the Estonian

dialogue corpus are analysed in order to illustrate communicative space and to find out linguistic cues for

automatic recognition of different coordinates. A limited version of the dialogue model is implemented on

the computer.

1 INTRODUCTION

When communicating, people express their attitudes

which depend on their individual characteristics,

social roles, topic of conversation, etc. Social

attitude or interpersonal stance is an affective style

that can be employed in an interaction with a person

or a group of persons. It consists of conveying a

particular feeling in the interaction, for instance,

being friendly, dominant, hostile or polite (Scherer,

2005). These attitudes can be conveyed by words

and voice features but also by nonverbal means –

facial expression, body movement, and gestures

(Knapp and Stone, 2009). There are some important

cues generated with speech: tone, volume, speed of

voice can change, depending on the social attitude.

Similarly, the relations we have with other people

influence our body gestures. These verbal and

nonverbal cues are used in an interaction in an

unconscious process to communicate and understand

an attitude. Depending on the gestures and facial

signals one is displaying, a social attitude can be

perceived (Ravenet et al., 2012).

The results of studies of human-human

communication can be used when modelling

interaction with the computer. Different features

have to be taken into account in order to make it

possible for a user to interact with the computer like

with another human, by using verbal as well as

nonverbal means. Attitudes expressed by gestures

and facial signals are especially important when

modelling socio-emotional agents.

Our aim is to develop a dialogue system (DS)

which interacts with the user in a natural language

following norms and rules of human communica-

tion. For that reason, we study how people are using

their language when communicating and how they

are expressing their attitudes by language means.

We have worked out a formal model of negotiation

dialogue (Koit and Õim, 2014; Koit, 2015). In the

current paper, we will further develop the model

concentrating on attitudes of communication

participants. We consider two kinds of attitudes: (1)

related to a negotiation object when reasoning in

communication, and (2) related to a communication

partner. We introduce the concept of communicative

space – a mental space which coordinates

correspond to the attitudes of communication

participants (their social distance, degrees of

cooperation and intensity of communication, etc.).

When communicating, the participants can ‘move’

in communicative space from one ‘point’ to another

and depending on their locations, they choose

suitable communicative strategies in order to

achieve their communicative goals.

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

Koit, M.

Modelling Attitudes of Dialogue Participants - Reasoning and Communicative Space.

DOI: 10.5220/0006653205810588

In Proceedings of the 10th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2018) - Volume 2, pages 581-588

ISBN: 978-989-758-275-2

581

coordinates of the vector of motivational sphere of a

reasoning subject. The attitudes are changing in

dialogue as influenced by arguments of a dialogue

partner. Section 3 considers the concept of

communicative space which dimensions express

attitudes of a communication participant in relation

to a communication partner. The dimensions will be

illustrated with dialogue examples from the Estonian

dialogue corpus. Section 4 considers an

implementation of the dialogue model based on

information states. Section 5 discusses how

communicative space can be used when developing

a dialogue system. Section 6 draws conclusions.

2 DIALOGUE MODEL

We are modelling negotiations between two

participants A and B in a natural language. The

initiator A makes a proposal to his communication

partner B about an action D. The communicative

goal of A can be either doing or not doing the action

by B. If B rejects A’s communicative goal (and/or

there are obstacles that do not allow B to accept the

goal) then a negotiation will start where both

participants can present their arguments and

counterarguments. The dialogue finishes with B’s

decision (to do D or not).

2.1 Reasoning Model

When negotiating, a communication participant

must know how to direct the functioning of the

partner’s psychological mechanisms in communica-

tion, in order to change his/her attitudes in relation

to the object of reasoning.

In our reasoning model, we use a naïve, ‘folk’

theory (D’Andrade, 1987; Davies and Stone, 1995).

With respect to a naïve theory of reasoning, there are

three kinds of determinants which can cause humans

to reason about an action D (Õim, 1996). The internal

determinants are the wishes of the subject related to

the action (WISH-determinants) and his/her

considerations that it would be needed, reasonable, or

necessary to do D (NEEDED-determinants). The

external determinants (or MUST-determinants)

originate from outside the subject and operate through

the concept of punishment which is a reaction to

subject’s not fulfilling obligations or prohibitions.

When reasoning about doing D, the general

balance of the ‘weights’ of positive and negative

aspects should be computed. The weights express

beliefs/attitudes of a reasoning subject in relation to

different aspects of the object of reasoning (in our

case, doing an action).

The reasoning itself depends on the determinant

which triggers it (respectively, WISH, NEEDED, or

MUST), thus, we can describe three different

prototypical reasoning procedures.

Our reasoning model is a kind of BDI (belief-

desire-intention) model (Bratman, 1999). It consists

of two parts: (1) a model of human motivational

sphere which includes attitudes of a reasoning

subject in relation to the aspects of the action under

consideration, and (2) reasoning procedures (Koit

and Õim, 2014).

We represent the model of motivational sphere

as a vector with numerical coordinates which

correspond to the different aspects of the action D:

= (w(resources

), w(pleasantness

), w(unplea-

santness

), w(usefulness

), w(harmfulness

), w(obli-

gatory

), w(punishmentfornot_doing

), w(prohibi-

ted

), w(punishmentfor_doing

)).

Here w(resources

) = 1 if the reasoning subject

has all the resources needed for doing D (or 0 if

some of the resources are missing), w(obligatory

) =

1 if the action is obligatory for the subject (otherwise

0), w(prohibited

) = 1 if the action is prohibited

(otherwise 0). The values of the remaining

components can be natural numbers on the scale

from 0 to 10. Still, our numerical scales are

simplifications – when reasoning, a human subject

hardly ever uses numerical weights. (S)he rather

operates with words of a natural language or with

fuzzy evaluations in order to express his/her

attitudes in relation to the object of reasoning.

We use two vectors (w

and w

) in our model

of dialogue. 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

(attitudes) of D’s aspects and it is used by B when

reasoning about doing D. The other vector w

the partner model which includes A’s beliefs

concerning B’s evaluations and it is used by A when

planning the next turn in dialogue. We suppose that

A has some preliminary information about B in order

to compose the initial partner model before making

the proposal to do D. Both the models will change as

influenced by arguments presented by the

participants in negotiation. For example, every

argument presented by A targeting the pleasantness

of D should increase the corresponding values of

(pleasantness

) as well as w

(pleasantness

2.2 Communicative Strategies and

Tactics

A communicative strategy is an algorithm used by a

ICAART 2018 - 10th International Conference on Agents and Artiﬁcial Intelligence

582

participant for achieving his/her goal in

communication (Koit and Õim, 2014). The initiator

A can realize his communicative strategy in different

ways, for example, he can entice, persuade or

threaten the partner B to do or, respectively, avoid

doing D (respectively, he stresses the pleasantness or

usefulness of doing/not doing D or punishment for

not doing/doing D if it is obligatory/prohibited). We

call these ways of realization of a communicative

strategy communicative tactics. The partner B uses a

similar communicative strategy. Some algorithms

are presented in (Koit, 2017).

When reasoning in order to make a decision, B

considers her resources as well as different positive

and negative aspects of doing D. If the positive

aspects weigh more than negative then the decision

will be “do D” otherwise “do not do D”. 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

(for example, to do the action D) and to change her

attitudes (the model w

), A uses a partner model

. A stresses the positive and downgrades the

negative aspects of the action. Various arguments

for doing/not doing D are presented in a systematic

way, for example, when persuading B to do D, A

stresses time and again the usefulness of D. The

partner B can similarly stress or downgrade a certain

aspect of the action by her counterarguments.

3 COMMUNICATIVE SPACE

The models w

and w

capture the attitudes of

communication participants in relation to the action

D under consideration. In order to model the

attitudes of a participant in relation to a

communication partner we use the concept of

communicative space (Koit, 2015).

Concepts of space are fundamental to our

understanding of human action and interaction.

Healey et al. (2008) apply the concepts of (physical)

space, place and communication space to the

analysis of a corpus of interactions from an online

community. Some studies represent social attitude

with two dimensions: a dominance dimension (also

called power, control or agency), that represents the

degree of control one has on another, and a liking

dimension (also called appreciation, affiliation or

communion), that represents the degree of

appreciation, liking of another (Carney et al., 2005;

Hess et al., 2005). In other studies these dimensions

are used among other dimensions like formality or

trust (Burgoon et al, 1984). Concept of social

attitude or interpersonal stance in interaction (being

polite, distant, cold, warm, supportive,

contemptuous, etc.) is considered in (Carofiglio,

2009). Ravenet et al. (2012) have figured out a table

showing the influence of dominance and liking in

the nonverbal behavior depending on the gender of

the speaker. A two-dimensional T. Leary’s

interpersonal circle (IPC) has been introduced as a

framework to classify four types of interpersonal

attitude (Dominant-Hostile, Dominant-Friendly,

Submissive-Hostile and Submissive-Friendly).

3.1 Dimensions of Communicative

Space

We specify communicative space by a number of

dimensions that characterize the relationships of

participants in a communicative encounter. Commu-

nication can be collaborative or confrontational,

personal or impersonal; it can also be characterized

by the social distance of participants (near, far), etc.

(Koit, 2015). These dimensions represent a sub-

system of human communicative competence with

deep evolutionary roots, the basic function of which

is to regulate the communication process. Together,

they bring the social aspect of communication into

the model. People have an intuitive, naïve theory of

these dimensions; the values of the coordinates can

be expressed by specific words. Instead, at present

we use numerical values as approximations in our

model (like in the model of human motivational

sphere, cf. Section 2.1).

We determine communicative space as n-

dimensional (n>0) space. At least, the following

dimensions can be specified:

• dominance (on the scale from dominant to

submissive)

• communicative distance to the partner (on the

scale from familiar to remote)

• cooperation (on the scale from collaborative to

confrontational)

• politeness (from polite to impolite)

• personality (from personal to impersonal)

• modality (from friendly to hostile)

• intensity (from peaceful to vehement).

We use the numbers +1, 0 and -1 for the values

of the coordinates of communicative space. For

example, the value +1 on the scale of modality

means friendly interaction and the value -1 means

hostile interaction. Communicative distance is -1 if

the person feels closeness in relation to his/her

communication partner and +1 if (s)he is far from

the partner. On any scale, 0 is the neutral value. Still,

Modelling Attitudes of Dialogue Participants - Reasoning and Communicative Space

583

one could consider a bigger number of values than

three on every scale. It would be also possible to use

continuous scales instead of discrete values.

Communication participants can be located in

different points of communicative space. For

example, a boss can angrily communicate with his

subordinate who, on the contrary, remains neutral or

even friendly. One communication participant can

feel closeness to his/her partner whereas the partner

has different feelings, etc. Moreover, the participants

can also ‘move’ from one point to another during the

encounter. For instance, the participants who were

on confrontational positions at the outset can reach

the collaborative one at the end.

3.2 Dialogue Analysis

With the aim to model human-computer interaction

we are considering human-human communication.

Where do people place themselves in communica-

tive space when communicating and how do they

‘move’ there? We are especially interested in

linguistic means, which would help us to recognize

‘the points’ of communicative space on the basis of

texts of a natural language (in our case, Estonian).

We try to find out some cues which are applicable

without semantic analysis of text because necessary

software is currently missing for the Estonian.

3.2.1 Dialogue Corpus

Our current study is based on the Estonian dialogue

corpus (Hennoste et al., 2008). The corpus includes

different kinds of dialogues: (1) transcripts of

human-human spoken dialogues, (2) written

dialogues collected in simulations by Wizard-of-Oz

method, and (3) log files of interactions with two

DSs which give information in Estonian (movie

programs and dentist advice, respectively). The

biggest part of the corpus – about 1000 spoken

dialogues – is recorded in authentic situations and

transcribed by using the transcription system of

Conversation Analysis (Sidnell and Sivers, 2012).

Each transcription is provided with a header that

lists situational factors (‘meta-knowledge’ about the

dialogue session), which affect language use –

participants’ names, social characteristics, relations

between participants in the situation, specification of

situation (private/public place, private/institutional

conversation), etc.

Dialogue acts (DA) are annotated in the corpus

by using a customized typology which is based on

Conversation Analysis. Custom-made web-based

software is used for annotation. Another custom-

made software tool enables to visualize speech

features (overlapping speech, comments of

transcribers, etc.) by different colors and also to

calculate some statistics (the counts of utterances,

words, different DAs, frequency of words and

certain sequences of DAs, etc.).

3.2.2 Communicative Space in

Telemarketing Calls

In a previous paper (Koit, 2015) we have analyzed

directory inquiries and negotiations between

acquaintances in the Estonian dialogue corpus. For

the current paper, we have chosen a sub-corpus

consisting of 51 telemarketing calls (in total, 35,678

running words, 5744 utterances). Sales clerks of an

educational company call to potential customers –

managers or personnel officers of other companies –

and offer training courses for employees. The

communicative goal of a clerk is to achieve a

positive decision of a customer (to take a course). A

clerk is giving information about his educational

company, collecting information about the customer

by asking questions, arguing for usability of the

courses for the customer in order to awake her

interest to take a course. We are looking for

linguistic cues in the transcripts of the calls which

can give a signal of certain values of the coordinates

of communicative space and therefore will

contribute to their automatic recognition. We prefer

the rule-based approach due to the limited size of

our current dialogue corpus which makes it hard to

implement statistical methods. However, as

demonstrates our analysis, it is difficult to determine

the values of coordinates only using texts, without

the possibility to take into account speech features

and what is more, nonverbal means of

communication.

Telemarketing calls are institutional dialogues

and this fact in a big way determines communicative

space – most of the coordinates typically have

neutral values (0). Still, there are some interesting

exceptions.

In the following examples, A is a sales clerk and

B is a customer. All the values of coordinates of

communicative space can be either +1, 0 or -1.

Transcription marks used in the examples are given

in (Hennoste et al., 2008). Let us only point out that

square brackets are used for overlapping speech and

capital letters for a louder segment.

3.2.2.1 Dominance

Headers of transcripts of dialogue recordings can

indicate whether the communication participants are

ICAART 2018 - 10th International Conference on Agents and Artiﬁcial Intelligence

584

in dominance-subordination relation or not.

However, headers of the telemarketing calls do not

include such relation – the participants are ‘equal’.

Instead, we can use another cue – overlapping

speech where a participant starts his/her turn before

the partner has finished, thereby indicating

dominance (value +1 in Example 1; the customer

takes initiative before the clerk has finished his turn,

overlapping speech is in square brackets).

Example 1

A: nii=et=te ei pea ´vajalikuks inimeste ´arendamist

nendes ´vald[kondades.]

Also you (plural) do not consider it necessary to educate

your people in these [domains]?

B: [tändap=ee] ma ´kordan me saame ´ise selle teemaga

´hakkama.

[It means], I repeat, we can make it ourselves.

3.2.2.2 Communicative Distance

In institutional dialogues like our telemarketing

calls, the participants typically are strangers and

their communicative distance is whether neutral (0)

or long (+1). In Estonian, the 2

person plural of

verbs and pronouns can be used to indicate a long or

neutral distance; the singular form indicates a short

distance. In Example 1, A indicates a neutral

distance (0) by using a plural form of the pronoun

you. B’s overlapping speech indicates a long

communicative distance (value +1), in addition to

dominance (value +1).

3.2.2.3 Cooperation

In our analyzed dialogues, the sales clerks always

communicate cooperatively as determined by their

social role. Similarly, the customers typically

express cooperation. Still, the customers are

antagonistic in a few of dialogues where they are

driving at a negative decision (do not take a course).

One cue to recognize cooperation of a sales clerk is

the dialogue act used as a reaction to the customer’s

counterargument (in our dialogues, a clerk always

uses DA ‘accept’, instead of ‘reject’). However,

when accepting a counterargument the clerk does

not abandon his communicative goal – after that he

presents his new arguments for taking a course by

the customer (Example 2, DA tags in bold).

Example 2

B: aga näiteks pro´jektijuhtimist teil=ei ´ole, mida mina

otsin tegelikult ´prae[gu.]

But you don’t have project management what I’m looking

for. Assertion

A: [mm]mmq noo::: jah.

Uhuh, yes. Limited accept

kui keegi=on=meie käest projektijuhtimist ´küsinud,

.hhh[h sii]s tegelikult

But if someone would order project management then we

actually… Assertion

B: [mhmh]

Uhuh. Neutral continuer

A: meil=on=need ´vahendid ´olemas.

… we are able to teach it. Assertion

3.2.2.4 Politeness

A linguistic cue for recognizing the politeness in

institutional dialogues is the 2

person plural of

verbs and pronouns. In addition, there are DAs of

greeting and leave-taking in the beginning and at the

end of negotiation. In most cases, the participants

also thank each other at the end of conversation.

Politeness can also be expressed by using of some

emotion words and expressions, Example 3 (value

+1; you in plural; nice):

Example 3

A: soovin teile `meeldivat `õhtu jätku

I wish you (plural) a nice continuation of the evening

[ja kuulmi]seni

and hear you

3.2.2.5 Personality

In institutional dialogues, the value of personality is

typically neutral (0). Still, there are some exceptions,

especially in the cases if customers have made a

negative decision and are about to finish negotiation.

Some cues to stress personality (value +1) are using

the 1

person of pronouns, loud speech, DA for

expressing a counterargument (Example 4, loud

segment in capital letters):

Example 4

B: ´ühesõnaga (0.3) ´M:INULE ei ole ´vaja tulla ´õpetama

firma ´juhtimist?

You can’t come to teach ME neither how to manage the

company

.hh ega seda kudas ´mina pean müüjat=vel (.) ´õpetama.

nor how I have to teach my shop assistant.

3.2.2.6 Modality

As expected, the value of modality is neutral (0) in

most of the dialogues. In a few of cases, clerks are

especially friendly (+1). In some other dialogues a

customer who has already made a negative decision

and does not want to continue communication,

expresses hostility (value -1, cf. Examples 1 and 4).

The cues are overlapping and loud speech.

Therefore, big values of dominance and personality

can also imply hostility.

Modelling Attitudes of Dialogue Participants - Reasoning and Communicative Space

585

3.2.2.7 Intensity

Typically, the value of intensity is neutral (0) as

expected for institutional dialogues. An exception is

a dialogue where the customer gives a lot of

counterarguments against the course offered by the

clerk (Example 4). Here, personality and hostility

expressed by counterarguments also imply

vehemence. The most important cue is loud speech.

When communicating, both sales clerks and

customers are restricted on their social roles because

both of them are official persons who represent their

institutions. A sales clerk having the communicative

goal to sell training courses of his company has to

keep the fixed communication point which can be

represented as (0, +1/0, 0, 0, 0, 0, 0), and the fixed

communicative tactics (persuasion – stressing the

usefulness of the courses). A customer has more

freedom – she can present a lot of counterarguments

to the clerk’s proposal being sometimes

confrontational, vehement and even hostile

(Examples 1 and 4).

4 IMPLEMENTATION

A limited version of our dialogue model is

implemented as a simple DS which interacts with a

user in written Estonian. Information-state dialogue

manager is used in the implementation (Traum and

Larsson, 2003).

There are two parts of an information state –

private (information accessible only for one

participant) and shared (information accessible for

both participants). For example, the private part of

an information state of the initiator A (where A’s

communicative goal is “B will do D”) consists of the

following information slots:

• current partner model (vector w

of A’s

attitudes about B’s attitudes in relation to the

action D)

• current location of A in communicative space

(A’s attitudes in relation to B, that is, the values

on the scales of dominance-subordination,

cooperation-antagonism, etc.)

• communicative tactics t

which A has chosen

for influencing B

• reasoning procedure r

which A is trying to

trigger in B (and bring to a positive decision)

• stack of (sub-)goals under consideration. In the

beginning, A puts the initial goal (“B will do

D”) into the stack

• set of dialogue acts: proposal, assertions

(arguments) for increasing or decreasing the

weights of different aspects of D for B (that is,

for changing B’s attitudes in relation to D), etc.

• set of utterances for verbalizing the dialogue

acts.

The shared part of an information state contains

world knowledge, language knowledge, set of

reasoning procedures R={r

,…,r

}, set of

communicative tactics T={t

, t

, …, t

}, and the

dialogue history – the utterances together with the

participants’ signs and dialogue acts: p

],…, p

] where p

=A; p

, p

,… are

whether A or B; u

, u

,… are utterances and d

, d

,…

are DAs.

Update rules will be used by a participant for

moving from one information state to another. There

are different categories of update rules both for

generating and interpreting of turns.

The computer plays A’s role and the user B’s

role. A’s communicative goal is “B will do D”, and

B’s goal is “do not do D” (Koit, 2017). The

computer has ready-made sentences (assertions) for

expressing of arguments, i.e., for stressing or

downgrading the values of different aspects of the

proposed action, which depend on its user model.

The user (B) can put in free texts. Communicative

space is not involved in the current implementation,

thus, the attitudes of participants in relation to each

other are not yet taken into account.

Starting a dialogue, A determines a partner model

, fixes its communicative strategy and chooses

the communicative tactics which it will follow, that

is, the computer respectively determines a reasoning

procedure which it will try to trigger in B’s mind. A

applies the reasoning procedure in its partner model,

in order to ‘put itself’ into B’s role and to choose

suitable arguments when convincing B to decide to

do D. Supposedly, the models w

and w

are

different when a dialogue starts but they are

approaching each to another during negotiation, as

influenced by the presented arguments and

counterarguments. Still, the user B is not obliged

(but can) to follow neither certain communicative

tactics nor reasoning procedures. (S)he is also not

obliged to fix his/her attitudes in relation to D by

composing a model w

. However, A does not

‘know’ B’s attitudes (the values of the coordinates of

the supposed vector w

), it only can choose

arguments on the basis of B’s counterarguments.

Respectively, A is making changes in its partner

model w

during a dialogue.

ICAART 2018 - 10th International Conference on Agents and Artiﬁcial Intelligence

586

5 DISCUSSION

Our dialogue model considers two kinds of attitudes

of a dialogue participant: (1) his/her attitudes in

relation to an action (which is the object of

negotiation), and (2) his/her attitudes related to a

communication partner. Both kinds of attitudes are

changing in dialogue as influenced by behavior and

arguments of the communication partner.

When reasoning about doing an action, a subject

is weighing different aspects of the action (its

pleasantness, usefulness, etc.) which are included

into his/her model of motivational sphere. We

evaluate these aspects by giving them discrete

numerical values on the scale from 0 to 10. Still,

people do not operate with numbers in a reasoning

process. Instead, they rather use words of a natural

language. For example, the pleasantness of an action

can be evaluated by such words and expressions as

excellent, very pleasant, not so pleasant, etc.

Further, when reasoning, people do not operate with

exact values of the aspects of an action but they

rather make ‘fuzzy calculations’, for example, they

suppose/believe that doing an action is much more

pleasant than unpleasant and therefore they wish to

do the action. A problem is that the aspects of

actions considered here are not fully independent.

For example, harmful consequences of an action as a

rule, are unpleasant (while unpleasant will not

always be harmful). In addition, if a reasoning object

is different (not doing an action like in our case)

then the attitudes of a reasoning subject can be

characterized by a different set of aspects.

We represent the relations of communication

participants that influence the communication

process and its results as dimensions of

communicative space. As said, we use the values +1,

0 and -1 for the coordinates. Still, it is possible to

divide all the scales to a bigger number of values

(for example, from 0 to 10 like in the case of the

aspects of actions). Likewise, it is possible to

operate with continuous scales instead of discrete

values (cf. Mesiarová-Zemánková, 2016). It is also

possible to use words of a natural language for the

values. For example, the modality of communication

can be friendly, ironic, hostile, etc. The problem

remains how to determine objective criteria and

apply them when dividing the scales. However,

annotation of the points of communicative space in

written dialogues is difficult and subjective already

with three different values (+1, 0, -1).

A serious problem is that the dimensions are not

fully independent (like in the case of aspects of

actions). For example, dominance usually implies a

longer communicative distance. A longer

communicative distance can imply a smaller value

on the scale of personality, etc. as demonstrated in

the examples (Section 3.2).

Further empirical research is needed in order to

determine the list of dimensions of communicative

space, their relations and values on different scales

(which can be different). Linguistic cues can be used

for recognizing of values of some coordinates

(Section 3.2). For example, if a participant uses the

person singular form of pronouns in Estonian

then (s)he is indicating a short communicative

distance (-1) and a big value on the personality scale

(+1). Emotion words and expressions help to

recognize the values of some coordinates, for

example, please and thank indicate politeness. The

comments of transcribers in transcriptions of spoken

dialogues help to determine the modality of

communication (for example, the comment

((friendly)) indicates the value +1). The dialogue act

tags contribute to recognizing of some coordinates.

For example, conventional (ritual) DAs of greeting

and thanking express politeness. Opinion mining

(Liu, 2015) could be used to automatically annotate

communication points. Still, the small size of the

Estonian dialogue corpus does not yet allow

implementing statistical or machine learning

methods.

How to use the concept of communicative space

in human-computer systems? Let us point out two

general research areas. Firstly, the systems which

model human communication, not participating in

communication but analyzing human dialogues on

the expert level, e.g. analyzing communication

protocols, reconstructing locations and movements

of participants and making conclusions. The second

direction is development of DSs which interact with

people in a natural language and perform certain

roles. An interesting and useful kind of DSs rapidly

developing in the last years are embodied

conversational agents (Harthold et al., 2013;

Ravenet et al., 2015; Dermouche, 2016). Such a

conversational agent behaves like a human thereby

expressing a suitable emotional attitude.

6 CONCLUSIONS

We introduce a model of dialogue concentrating on

modelling of attitudes of dialogue participants. We

consider two kinds of attitudes expressed by

participants in negotiation about doing an action: (1)

related to the action, and (2) related to a

communication partner. We represent the first kind of

Modelling Attitudes of Dialogue Participants - Reasoning and Communicative Space

587

attitudes as coordinates of a vector of motivational

sphere of a participant (who is reasoning about doing

an action). The second kind of attitudes is represented

by using the concept of communicative space – a

mental space where communication takes place.

In order to explain the concept of communicative

space we analyze a sub-corpus of human-human

telemarketing calls of the Estonian dialogue corpus.

We have implemented the model of negotiation

as a simple dialogue system where the computer

plays A’s and the user B’s role. So far, the

implementation does not include communicative

space. This needs deeper investigations and remains

for the further work.

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).

REFERENCES

Bratman, M. E., 1999. Intention, Plans, and Practical

Reason. In CSLI Publications.

Burgoon, J. K. B., Buller, D., Hale, J. L., De Turck, M. A.,

1984. Relational Messages Associated with Nonverbal

Behaviors. In Human Communication Research, 10(3):

351–378. DOI: 10.1111/j.1468-2958.1984.tb00023.x

Carney, D. R., Hall, J. A., Smith LeBeau, L., 2005. Beliefs

about the nonverbal expression of social power. In

Journal of Nonverbal Behavior, 29:105–123.

Carofiglio, V., De Carolis, B., Mazzotta, I., Novielli, N.,

Pizzutilo, S., 2009. Towards a Socially Intelligent

ECA. In IxD&A 5–6: 99–106.

D’Andrade, R., 1987. A Folk Model of the Mind. In

Cultural models of Language and thought, 112–148.

London: Cambridge University Press.

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

theory of mind debate. Oxford, Cambridge,

Massachusetts: Blackwell.

Dermouche, S., 2016. Computational Model for

Interpersonal Attitude Expression. In ICMI’16, 554–

558, Tokyo, Japan. DOI: 10.1145/2993148.2997619

Fu, Q., 2014. Typical Non-verbal Behaviour of Different

Interpersonal Stances during Police Interrogations. In

Research Topics, Human Media Interaction,

University of Twente.

Hartholt, A., Traum, D., Marsella, S.C., Shapiro, A.,

Stratou, G., Leuski, A., Morency, L.P., Gratch J.,

2013. All together now: Introducing the Virtual

Human toolkit. In Proc. of the 13th International

Conference on Intelligent Virtual Agents, IVA,

Edinburgh, UK, 368–381. Springer Verlag. DOI:

10.1007/978-3-642-40415-3

Healey, P. G. T., White, G., Eshghik, A., Reeves, A. J.,

Light, A., 2008. Communication Spaces. In Journal

Computer Supported Cooperative Work, 17:2–3, 169–

193. http://dl.acm.org/citation.cfm?id=1363341

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, 2025–2032,

Marrakech, Morocco, ELRA. www.lrec-conf.org/pro-

ceedings/¬lrec2008

Hess, U., Adams, R. B., Kleck, R., 2005. Who may frown

and who should smile? Dominance, affiliation, and the

display of happiness and anger. In Cognition and

Emotion, 19(4):515–536.

Knapp, M. L., Hall, J. A., 2009. Nonverbal Communica-

tion in Human Interaction. Wadsworth Publishing.

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

argumentation in agreement negotiation processes. In

Argument & Computation, 5, 209−236, DOI: 10.1080/

19462166.2014.915233

Koit, M., 2015. Communicative Strategy in a Formal

Model of Dispute. In Proc. of ICAART, 489−496,

Lisbon, Portugal. DOI: 10.5220/0005274904890496

Koit, M., 2017. Reasoning and Communicative Strategies

in a Model of Argument-Based Negotiation. In Proc.

of INISTA, 6 p., Gdynia, Poland.

Leuski, A., Traum, D., 2011. NPCEditor: Creating virtual

human dialogue using information retrieval techni-

ques. In AI Magazine, 32(2), 42–56.

Liu, B., 2015. Sentiment Analysis. Mining opinions, senti-

ments, and emotions. Cambridge University Press.

Mesiarová-Zemánková, A., 2016. Sensitivity analysis of

fuzzy rule-based classification systems by means of

the Lipschitz condition. In Soft Computing, 103–113.

DOI: 10.1007/s00500-015-1744-z

Õim, H., 1996. Naïve Theories and Communicative

Competence: Reasoning in Communication. In

Estonian in the Changing World, 211–231. University

of Tartu Press.

Ravenet, B., Ochs, M., Pelachaud, C., 2012. A computa-

tional model of social attitude effects on the nonverbal

behavior for a relational agent. In Proc. of WACAI,

Grenoble, France.

Ravenet, B., Cafaro, A., Biancardi, B., Ochs, M.,

Pelachaud, C., 2015. Conversational Behavior

Reflecting Interpersonal Attitudes in Small Group

Interactions. In Intelligent Virtual Agents, 14 p. DOI:

10.1007/978-3-319-21996-7

Scherer, K. R., 2005. What are emotions? And how can

they be measured? In Social Science Information.

http://journals.sagepub.com/doi/abs/¬10.1177/-053901

8405058216

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

Conversation Analysis. Boston: Wiley-Blackwell.

Traum, D., Larsson, S., 2003. The information state

approach to dialogue management. In Current and

New Directions in Discourse and Dialogue, 325–353,

Kluwer.

ICAART 2018 - 10th International Conference on Agents and Artiﬁcial Intelligence

588