Goal Formulation based on Communicative Principles 
Kristiina Jokinen* 
Computational Linguistics Laboratory 
Graduate School of Information Science 
Nara Institute of Science and Technology 
8916-5 Takayama, Ikoma, Nara 
630-01 JAPAN 
kris~is.aist-nara.ac.jp 
Abstract 
The paper presents the Constructive 
Dialogue Model as a new approach to 
formulate system goals in intelligent di- 
alogue systems. The departure point 
is in general communicative principles 
which constrain cooperative and coher- 
ent communication. Dialogue partici- 
pants are engaged in a cooperative task 
whereby a model of the joint purpose is 
constructed. Contributions are planned 
as reactions to the changing context, 
and no dialogue grammar is needed. 
Also speech act classification is aban- 
doned, in favour of contextual reasoning 
and rationality considerations. 
1 Introduction 
Two general approaches can be distinguished in 
dialogue management: the structural approach, 
which uses a dialogue grammar to capture regu- 
larities of the dialogue in terms of exchanges and 
moves (Bilange, 1992; Cawsey, 1993; Grosz and 
Sidner, 1986), and the intention-based approach, 
which classifies the speaker's beliefs and intentions 
into speech acts, and uses planning operators to 
describe them (Appel% 1985; Allen and Perrault, 
1980; Bunt et al., 1984). Both regard natural lan- 
guage as purposeful behaviour, but differ in how 
this behaviour is to be described. The former sees 
dialogues as products and compiles participants' 
beliefs and intentions into a predefined dialogue 
structure, whereas the latter focusses on the par- 
ticipants' goals, and hides the structure in the re- 
lations between acts which contain appropriately 
chosen sets of beliefs and intentions as their pre- 
conditions and effects. 
We will not go into detailed evaluation of the 
approaches, see e.g. (Jokinen, 1994), but draw at- 
tention to three aspects of dialogues which have 
*I am grateful to Yuji Matsumoto for providing an 
excellent resem'ch environment during my JSPS Post- 
doctoral Fellowship, and Graham Wilcock for helpful 
discussions. 
not been properly addressed before, although 
widely acknowledged in literature, and important 
in building robust Natural Language interfaces: 
1. dialogue is a collaborative process and its 
structure is recognised by external observa- 
tion, not prescribed as an internal constraint 
of dialogue management (Sacks et al., 1974; 
Clark and Wilkes-Gibbs, 1990), 
2. the speakers' beliefs and intentions in a given 
dialogue situation are various, and cannot all 
be checked when deciding oil the next re- 
sponse (Cherniak, 1986), 
3. communicative acts are part of social activ- 
ity, constrained by normative obligations of 
rational agency (Allwood, 1976). 
We discuss these aspects from the point of view 
of cooperative goal formulation and present the 
Constructive Dialogue Model as a new approach 
to plan system responses. Our departure point 
is in general conmmnicative principles which con- 
strain cooperative and coherent communication, 
and radical steps are taken in two respects: the 
dialogue grammar is abandoned as an infeasible 
way to describe dialogues, and also speech act 
recognition is abandoned as a redundant labelling 
of intention configurations. The first step means 
that the structure is not built according to struc- 
turing rules, but emerges from local coherence as 
the dialogue goes on. The second step means that 
beliefs and intentions are dealt with by reasoning 
about the utterance context and communicative 
constraints instead of speech act types. The deci- 
sion about what to say next falls out as a result of 
the agent complying with the communicative prin- 
ciples which refer to the agent's rationality, sire 
cerity, motivation and consideration. Combined 
with contextual knowledge, they account for the 
acceptability of different alternative responses. 
The paper is organised as follows. The theoret- 
ical framework and its formalisation as the Con- 
structive Dialogue Model are discussed in Section 
2. Section 3 presents how the system's commu- 
nicative goal is determined, and Section 4 provides 
comparision with related work. Finally, conclu- 
sions and filture directions are given in Section 5. 
598 
2 Constructive Dialogue Model 
2.1 Rational, (:()operative way to react 
Rational agents try to follow the principles of Ideal 
Cooperation (Allwood, 1976) in comimmication: 
(1) assume a joint l)urpose, (2) show cognitive 
consideration (epistemic rationality regarding ap- 
propriate ways to react) and ethical consideration 
(intention to react in a way that does not prevent 
the partner fi'om fiflfilling her goals), and (3) trust, 
that the partner is acting according to the same 
principles. Ideal cooperation does not mean that 
the agents always react in the way the partner in- 
tended to evoke, but rather, it sets the normality 
assumptions for the way the agents would behave 
if no disturbing factors were present. As (Gal- 
liers, 1989) points out, conflict resolution forms 
an important part of human conmmnication, and 
if systems are always ready to adopt the user's 
role, they becolne rigid and unrealistic. However, 
if the conflict becomes so serious that it makes any 
cooperation impossible, communication will break 
down as well. Rational agents thus try to conlnlu- 
nicate so as to conforln to the shared assumptions 
about operationally appropriate and ethically ac- 
ceptable acts in a given situation (Jokinen, 11995). 
Empirical dialogue research has emphasised col- 
laborative nature of dialogues (Sacks et al., 1974; 
Clark and Wilkes-Gibbs, 1990). Also computa- 
tionally oriented dialogue studies show that the 
users express themselves vaguely and continue 
with follow-up questions (Moore and Paris, 1993), 
and our corpus 1 supports the view that even sin> 
ple information seeking dialogues resemble nego- 
tiations rather than straightforward question-ans- 
wer sequences. Based on these obeservations and 
the theoretical assumptions of Ideal Cooperation, 
we distinguished the main factors in rational, co- 
operative dialogue management as follows: 
Surface form Expressive intentions 
Declarative express a belief: 
want (Sp ,know(He ,P) ) 
Interrogative desire for information: 
want (Sp,know(Sp, P) ) 
hnperative desire for action: 
want (Sp,do(ge,P)) 
Exclamative express action: 
want (Sp, do (Sp,P)) 
whereas the information seeker is not ex- 
pected to teach the information provider how 
to look for the information. The roles can be 
further difl'erentiated with respect to social 
factors such as acquaintance of the addressee 
and fornlality of the situation. 
• Communicative obligations. Social, nor- 
mative requirements that concern the agent's 
sincerit'9 (exchange information which is true 
or for which evidence can be provided), mo- 
tivation (exchange information which is re- 
lated to one's goals and strategies), and con- 
sideration (exchange information which the. 
partner is able to deal with). 
• Task. Gives rise to communication. Task 
goals are planned to complete a real workt 
task (rent a car, book ~ flight, repair a pump) 
but because of uneven distribution of knowl- 
edge, the agents usually need to collaborate 
to achieve the goal, and thus formulate com- 
,nunicative goals to obtain missing informa- 
tion, el. (Guilm, 1994). 
B. Commmficative act: 
• Expressive and evocative attitudes. To 
distinguish between the effects of an utter- 
ance and the intentions behind it, Austin's 
concept of illocution is split up into two: ez- 
pvession of the speaker's attitude and evoca- 
tion of a reaction in the partner; perlocution 
corresponds to what is actually achieved by 
the act: the evoked respoT~se, cf. (Allwood, 
1976). Expression may differ fl'om evoca- 
tion (irony, indirectness), aud the evoked 
response fi'om the evocative intentions (the 
agent requests ilfformation that the partner 
cannot or does not want to disclose; the agent 
fa.ils to fi'ighten the partner becmme this has 
guessed the agent's malicious intentions). 
Evocative intentions 
share the belief: 
want (Sp, want (He, know(He, P) ) ) 
provide the desired information: 
want (Sp, want (He, know (Sp, P) ) ) 
(provide) action: 
want (Sp, want (He, do (He, P) ) ) 
attend to the action: 
want (Sp, want (He, do (Sp, P) ) ) 
Figure 1: Conventional association of expressive and evocative intentions 
with surface form, modified fi'om (Allwood, 1992). 
A. Communicative situation: C. Communicative context: 
• Role. Characterised by global communica- • Expectations. Evocative intentions put 
live rights and obligations of the agents. E.g. 
the information provider is expected to give 
information which is relevant for the task, 
1The corpus was collected by the Wizard-of-Oz 
technique with users trying to find information on car- 
\[tire companies and restaurants in a particular area, 
and is reported in (Nivre, 1992). 
pressure on the agent to react in a particular 
way. Conventional expectations, carried by 
the surface form (Fig. 1), serve as anchoring 
points in reasoning about the partner's com- 
municative goal (Cohen and Levesque, 1990). 
Initiatives. If the agent has initiated a con> 
municative goal, she "has the initiative" and 
599 
also the right to pursue the goal until it is 
achieved or not relevant anymore. She also 
has the right to expect the partner to collab- 
orate or at least not prevent the agent from 
achieving her goal. 
• Unfulfilled goals. If the expressive attitu- 
des of the partner's response match the 
evocative intentions of the agent's contribu- 
tion, the communicative goal of the agent's 
contribution is fulfilled. An unfulfilled goal is 
pushed forward or stored for later processing. 
When the agent has the right to take the ini- 
tiative, a previously unfulfilled goal can be 
taken up. If the goal is still unfulfilled and 
relevant, it is resumed, otherwise dropped. 
• Thematic coherence. A competent agent 
relates the topic of her contribution to what 
has been discussed previously or marks an 
awkward topic shift appropriately; otherwise 
the agent risks being understood. Thematic 
relatedness is based on the types of relation- 
ships which occur in the domain. 
2.2 The CDM System 
The theoretical framework is formalised as an ap- 
proach to dialogue management called the Con- 
structive Dialogue Model, CDM (Jokinen, 1994). 
In CDM, the dialogue is an instrument to ex- 
change new information on a particular topic to 
complete a real world task, and it is managed lo- 
cally by reacting to the changed dialogue context. 
The task division and information flow in a 
CDM system 2 is shown in Fig. 2. The dialogue 
manager operates on the Context Model which is 
a dynamic knowledge base containing facts about 
the agents' goals, expressive and evocative atti- 
tudes, central concepts (topic), and new informa- 
tion. It also has access to three static knowl- 
edge bases: Communicative Principles (knowl- 
edge about rational, cooperative communication), 
Application Model (knowledge about tasks and 
roles), and World Model (general knowledge about 
the entities and their relations in the world). 3 
Dialogue contributions are constructed in three 
phases corresponding to the three main process- 
ing tasl~s. Analysis of the input message results 
in the user's communicative goal, and contains 
four subtasks: determine the explicitness level, in- 
terpret the propositional content, check coherence 
and verify obligations. Evaluation of the user goal 
concerns an appropriate joint purpose and deter- 
mines the next system goal. Response specifies the 
system's communicative goal up to the semantic 
representation using the same subtasks as analysis 
but in a reverse order. Evaluation and response 
form the agent's reaction. 
2The prototype is implemented in SICStus Prolog 
2.1, running under UNIX TM on a Sun SPARCStation. 
3 Linguistic knowledge is encoded in a linguistic lex- 
icon and grammar, and not discussed here. 
Input message 
I Semantic Representation 
I ANALYSE ~ New tnfol 
icituess ~ User CGoal 
iosition 
rence 
~ations 
User CGoal 
LCT 
~ATE 
purpose 
zation 
System CG al \] 
OND 
~ations 
rence 
osition 
Central 
Concept 1 
User 
Attitudes 
New- 
Info2 
System 
CGoal 
System 
Attitudes 
C 
O 
N 
T 
E 
X 
T 
M 
O 
D 
E 
L 
icitness __, Central 
I Semantic Representation 
Output message 
Figure 2: Information flow in the CDM system. 
The Context Model is represented as a parti- 
tioned Prolog database and the predicates have 
an extra argument referring to the contribution 
whose processing introduced them. In the attitude 
language the predicates know, want and do repre- 
sent belief, intention and action, respectively, s 
refers to the system and u to the user. Communi- 
cative Principles are reasoning rules of the fort0: 
if cntxtFactl .... , cntxtFactN 
then cntxtFactM+1,...,cntxtYactK. 
The World Model uses neo-Davidsonian event 
representation, and the Application Model provi- 
des mappings from World Model concepts to task 
and role related facts. 
3 Cooperative Goal Formulation 
In CDM, joint purpose represents the communica- 
tive strategy that an agent has chosen in a particu- 
lar situation to collaborate with her partner. It is 
determined by evaluating the partner's goal with 
respect to the communicative context: expecta- 
tions, initiatives, unfulfilled goals and coherence. 
Assigning binary values to these aspects, we get 
2 4 = 16 joint purposes, summarised in Fig. 3. The 
600 
goals 
fulfilled 
unfulfilled 
initiative 
speaker 
partner 
speaker 
partner 
central concept 
related 
related 
unrelated 
related 
unrelated 
unrelated 
expected response 
finish/start 
finish/sp-eeify 
~p-new 
new-request 
baekto 
repeat-new 
follow-up-old 
new-q~--- 
nml-expected response l\] 
continue/start 
objee~ 
~gelse 
new-indir-request 
subquestion 
~\]-eet 
~ed i 
Figure 3: Possible joint purposes if the contextual factors are assigned binary values. 
reasoning rules are as follows (examples of the al- 
ternatives can be found in (aokinen, 1994)): 
1. The agent has fulfilled goals only, and 
the initiative: Finish the dialogue or start a 
new one depending on the pending task goals 
(finish/start/continue/obj ect/spe elf y). 
Maintain the initiative if the response is re- 
lated, give the initiative if unrelated. 
2. The agent has fulfilled goals only, but 
no initiative: Adopt the partner's goal. 
Maintain the initiative if the response is ex- 
pected (follow-up-new,new-request), take 
the initiative if the response is non-expected 
(somethingelse ,new- indir-request). 
3. The agent has unfnlfilled goals, and 
the initiative: Adopt the partner's goal if 
the response is thematically related (backto, 
subquestion), persist with the own goal if 
unrelated (repeat-new,object). Maintain 
the initiative if the response is expected, give 
the initiative if non-expected. 
4. The agent has unfulfilled goals, but no 
initiative: Adopt the partner's goal. Main- 
tain the initiative if the response is themat- 
ically related (follow-up-old,continue), 
take the initiative if unrelated (new- 
question, uotrelated). 
The joint purpose describes coinnmnieative in- 
tentions in a context where no speaker obligations 
or considerations hold. In order to attend the re- 
quirements of a particular communicative situa- 
tion, the joint purpose needs to be specified with 
respect to the agent's role, task and communica- 
tive obligations. 
Specification of the joint purpose via the Ap- 
plication Model captures the cognitive consider= 
ation of Ideal Cooperation: the agent plans her 
response to be operationally appropriate in the 
current situation. The result is a communicative 
goal (c-goal), a set of communicative intentions in- 
stantiated according to the current task and role. 
The c-goal is then filtered through communica- 
tive obligations which impleinent the ethical consi- 
deration of Ideal Cooperation: the agent's com- 
municative competence shows in the ways she can 
realise the same c-goal in various situations. Some 
communicative obligations are listed in Fig. 4. 
Sincerity: "do I know this or can provide evidence?" 
1. Everything that the speaker asserts or implies is 
true unless otherwise explicitly stated. 
Motivation: "can I say this?" 
1. Everything that the speaker wants to know or 
wants the partner to do is motivated except if 
the speaker cmmot take the initiative on it. 
2. Everything that addresses what the partner 
wanted to know or wanted the speaker to do is 
motivated, except if the speaker emmot disclose 
the information or do the act. 
3. Everything that is related to CC is motivated if 
not already known. 
4. Everything that informs of inconsistency is mo- 
tivated if not already known. 
Consideration: "may I say this?" 
1. If the partner's goal cammt be fulfilled (presup- 
positions are false, facts contradictory, no infor- 
mation exists), it is considerate to inform why 
(explain, compensate, initiate repair). 
2. If the response would repeat previous informa- 
tion, it is considerate to leave this implicit unless 
the information is assigned a special emphasis. 
3. If the partner's response is unrelated, it is con- 
siderate to inform of the irrelevance, given that 
the speaker has unfulfilled goals. 
4. \[f the partner did not request a piece of related 
information, it is considerate to include this ex- 
plicitly in the response, given that the speaker 
intends to close the topic. 
Figure 4: Some communicative obligations. 
3.1 Example 
Consider the following sample dialogue where tile 
system's task is to provide service information to 
the user: 
Uh I need a car. 
$1: Do yon want to buy or rent one? 
Ui: Rent. 
$2: Where? 
U3: In Bolton. 
$3: OK. ttere are the car hire companies 
in Bolton: .... 
The analysis of the first user contribution U1 is 
given in Fig. 5. The content of the user's c-goal 
is inferred from the World Model which says that 
'needing a car' can be interpreted as 'wanting to 
have a cl~r'. 
601 
NEW INPO: needE(n,u,c), user(u), car(c) 
USER C-GOAL: want(u, want(s, know(s, 
\[want~e E(h, u, e )\]) )) 
CENTRAL CONCEPT: needE(n,u,c) 
EXPRESSIVE ATTITUDES: 
intention: user intend that system know P: 
want(u, know(s,\[needE(n,u, c ), user(u), car(c)\])) 
assumptions: user know that system not know P: 
know(u, not know(s,\[needE(n,u,d, user(u),car(e)\])) 
EVOCATIVE ATTITUDES: 
intention: user intend that system intend that 
system know P: 
want(u, want#, know(s, 
\[.eed E(n, u, c ), user(u ), ear(c)\]))) 
want(u, want(s, know(s,\[wantHave E(h,u,c )\], 
user(u ),~a,'(~ )\]) ) ) 
Figure 5: Context Model after the user contribution 
1 need a car. The constants n,u,c,h identify instanti- 
ated concepts. 
In the beginning of the dialogue the system has 
no unfulfilled goals, and its role as an obedient in- 
formation provider does not allow it to have the 
initiative. Moreover, any contribution is trivially 
unrelated to the previous topic, since no previ- 
ous topic exists. According to the Joint Purpose 
rule (2), the user's c-goal is thus adopted, and the 
system also takes the initiative, since the user con- 
tribution is non-expected (an information seeker 
is expected to start with a question or a request). 
The joint purpose becomes new-indir-request with 
"user wants to have a car" as the content, i.e. the 
communicative strategy is to share the user's want 
to have a car, and check if this want can be satis- 
fied within the Application Model. 
The system cannot provide the user with a car, 
but it can provide information about the services 
that, enable the user to have a car. Application 
Model lists car hire companies and car garages as 
possible services, so the communicative goal is for- 
mulated as to know which is the preferred service. 
The services are associated with renting or buying 
cars, thus the disjunction is realised as 5'1. 
The system responses $2 and 5"3 are based on 
the same strategy baclcto: the system 'goes back' 
to adopt the user's previous unfulfilled goal and 
tries to satisfy this in the updated context. 4 How- 
ever, they carry different c-goals due to different 
specification in the Application Model: $2 aims 
at narrowing down the database search, 5,3 com- 
pletes the original task. Finally, the communica- 
tive obligation Consideration (4) requires that the 
application service (car hire company) and loca- 
tion (Bolton) are explicitly expressed in $3 before 
the list of services. 
4The user response fulfills expectations and is the- 
maritally related, and the system has the initiative 
and unfulfilled goals, at least one based on the origi- 
nal task to provide information. 
4 Discussion and related work 
In Section 1 we pointed out three important as- 
pects of dialogues which have been insufficiently 
accounted for in the earlier approaches to dialogue 
management. In CDM, these aspects form the ba- 
sis of the system's functionality: dialogues are re- 
garded as collaborative activities, planned locally 
in the changed context as reactions to the previ- 
ous contributions and governed by the rationality 
principles of Ideal Cooperation. The logical omni- 
science assumption is tackled by partitioning the 
Context Model and focussing on specific knowl- 
edge with the hel f ) of thematic coherence; also ra- 
tionality considerations constrain reasoning. 
By adhering to general communicative princi- 
ples, CDM provides a new and uniform way to 
treat various phenomena that have been sepa- 
rately studied in previous research: goal formula- 
tion, coherence and cooperativeness. Communica- 
tive principles fimetion on the following levels: 
1. Determination of the joint purpose: 
reasoning about a communicative strategy in 
the context (expectations, initiatives, unflfl- 
filled goals, thematic coherence) 
2. Selection of the communicative goal: 
filtering the joint purpose with respect to the 
agent's role and task. 
3. Realisation of the goal: specifying the 
goal in regard to the communicative obliga- 
tions sincerity, motivation and consideration. 
However, we also use insights from the huge 
body of research that exists on dialogue man- 
agement and natural language planning. For in- 
stance, the negotiative nature of dialogues is em- 
phasised in (Moore and Paris, 1993) who show 
how rhetorical knowledge can be combined with 
the knowledge about the speaker's intentions and 
communicative goals so that the system can un- 
derstand follow-up questions or justify its expla- 
nations. Our work differs from this in that 
we study general requirements of communication 
rather than rhetorical relations and their augmen- 
tation with speaker intentions, to determine @- 
propriate responses. It is possible to modify our 
joint purpose algorithm with information about 
rhetorical relations so as to check expectations in 
regard to argmnentation, or to include rhetorical 
knowledge in the obligations used when reason- 
ing about multisentential contributions, but as our 
primary goal has been to specify communicative 
principles and use them in the formalisation of the 
cooperative and rational nature of dialogues, this 
kind of extension is left for future. 
(Guinn, 1994) presents a model of mixed- 
initative negotiation as collaborative problem 
solving. His Missing Axiom approach demonstra- 
tes collaboration and communication between two 
agents wl~o possess complementary knowledge: if 
the agent's information is not sufficient to allow 
602 
completion of the proof the agent is set to do, 
the agent attempts to provide the missing axioms 
through interaction. This is similar to our basic as- 
sumption of how domain tasks give rise to eonlinu- 
nication. The differences lie again in our einphasis 
on 'Rational and Cooperative Communication' as 
opposed to 'Interaction as a FMlure to Prove'. 
In abandoning dialogue grammar and speech 
act classification, we agree with the common view 
currently held among researches: dialogue struc- 
ture is constructed on the basis of the partici- 
pants' beliefs and intentions, and speech act types 
are at most convenient abbreviations for a set 
of attitudes held by the speakers, but do not 
constitute an explanation of the dialogue (Co- 
hen and Levesque, 1990; Galliers, 1989). We Mso 
use contextual knowledge extensively, and connect 
intention-based approaches to practical dialogue 
management: rationality and cooperation are not 
only tied to the agent's beliefs and intentions of 
the desidered next state of the world, but also to 
the wider social context in which the communica- 
tion takes place. 
5 Conclusion and future directions 
This paper has presented a new way to formulate 
system goals in intelligent dialogue systems. It ad- 
vocates a view-point where the system's fnnction- 
ality is iml)roved by relating the dialogue situation 
to communication in general. The constraints of 
rational, cooperative communication p,:ovide the 
framework in which to deal with contributions: 
communicators have a joint purpose, they obey 
communicative obligations and they trust that the 
partner behaves so that these constraints are tiff- 
filled. Dialogues are dynamic constructions, and 
contributions are locally planned and realised so 
that the communicative requirements of the dia- 
logue us a whole are respected. 
Current interests concern the extension of the 
communicative principles into different activities 
and agent roles. This contributes to the generality 
of the model by spelling out specific requirements 
of different communicative situations. It also en- 
ables us to study strategic planning and how dif- 
ferent roles affect the obligations that the agents 
want to obey (e.g. in conflict situations). Work is 
now in progress to cover other types of task dia- 
logues, and to enhance the impleinentation. 
References 
J. F. Allen and C. R. Perrault. 1980. Analyzing 
intention in utterances. Artificial Intelligence, 
15:143 178. 
J. Allwood. 1976. Linguistic Communication as 
Action an, d Cooperation. Department of Lin- 
guistics, University of GSteborg. Gothenburg 
Monographs in Linguistics 2. 
J. Allwood. 1992. On dialogue cohesion. Tech- 
nical Report Gothenburg Papers in Theoretical 
Linguistics 65, University of Gothenburg. 
D. Appelt. 1985. Planning Natural Language Ut- 
terances. Cambridge University Press, Cain- 
bridge. 
E. Bilange. 1992. Dialogue personne- 
machine. Mod~lisation et r&disation informa- 
tiquc. Ilerm~s, Paris. 
It. C. Bunt, R. J. Beun, F. J. H. Dols, J. A. van der 
Linden, and G. O. thoe Sehwartzenl)erg. 1984. 
The TENDUM dialogue system and its theo 
retical basis. Technical Report 19, IPO. 
A. Cawsey. 1993. Explanation and Interaction. 
~lT~e Computer Generation of Explanatory Dia- 
logues. The MIT Press, Cambridge, MA. 
C. Cherniak. 1986. Minimal Rationality. The 
M1T Press. Cambridge, MA. 
it. It. Clark and D. Wilkes-Gibbs. 1990. Refer- 
ring as a collaborative process. In P. R. Cohen, 
J. Morgan, and M. E. Pollack, editors, Inten- 
tions In Communication, pages 463-493. The 
MIT Press. Cambridge, MA. 
P. R. Cohen and It. J. Levesque. 1990. Rational 
interaction as the basis for communication. In 
P. R. Cohen, J. Morgan, and M. E. Pollack, ed- 
itors, Intentions in Communication, pages 221 
255. The MIT Press. Cambridge, MA. 
J. R. Galliers. 1989. A theoreticM framework 
for computer models of cooperative diMogue, 
acknowledging multi-agent conflict. Technical 
Report 172, University of Carat)ridge. 
B. a. Grosz and C. L. Sidner. 1986. Attention, in- 
tentions, and the structure of discourse. Com- 
putational Linguistics, 12(3): 175-204. 
C. I. Guinn. 1994. Mcta-Dialogue Behaviors: Im- 
proving the Efficiency of Human-Machine Dia- 
logue. A Computational Model of Variable Ini- 
tiative and Negotiation in Collaborative Prob- 
lem-Solving. Ph.D. thesis, Duke University. 
K. Jokinen. 1994. Response Planning in Informa- 
tion-Seeking Dialogues. Ph.D. thesis, UMIST. 
K. Jokinen. 1995. Rational agency. In M. Fehling, 
editor, Rational Agency: Concepts, Theories, 
Models, and Applications, pages 89--93. Pro- 
ceedings of The AAAI-95 Fall Symposium. 
J. D. Moore and C. L. Paris. 1993. Planning 
text for advisory diMogues: Capturing inten- 
tional and rhetorical information. Computa- 
tional Linguistics, 19(4):651-694. 
J. Nivre. (Ed.)(1992). Corpus collection an'd anal- 
ysis. Technical Report D2.1, PLUS deliverable. 
It. Sacks, E. A. Schegloff, and G. Jefferson. 
1974. A simplest systematies for the organiza- 
tion of turn-taking in conversation. Language, 
50(4):696-735. 
603 
