A framework is presented that models some important aspects of cooperative dialogue. The framework is based on certain beliefs of the participants and on pragmatic rules concerning the communicative acts that can be performed during the dialogue. It is claimed that, although beliefs and intentions are important modelling concepts in Artificial Intelligence, their role in conversation is over-estimated. In the type of dialogue considered, participants have no special expertise about the subject matter and the communication channel is an ideal channel. Participants have no access to a domain of discourse and behave maximally cooperative. Propositional information can be exchanged that is extended with the illocutionary force indicators 'statement1 or 'question'. In two cases, dialogue control acts can be performed: a. if the agent does not know the information asked for or b. if the agent detects inconsistent beliefs. Three types of beliefs are considered: a. the agent's belief about the subject matter, b. the agent's belief about mutual knowledge of both agents, and c. the agent's belief about what his partner does not know. Every dialogue starts with a question by one of the participants. A dialogue is finished either when the questioner knows the answer, or if both agents agree on the fact that the answer cannot be found, or if inconsistencies emerge. Examples are presented that show the consequences of different response strategies of the participants to answer the initial question.
Cite as: Beun, R.-J. (1991) A framework for cooperative dialogues. Proc. 2nd VENACO Workshop - The Structure of Multimodal Dialogue, 119-126
@inproceedings{beun91_smmd,
author={Robbert-Jan Beun},
title={{A framework for cooperative dialogues}},
year=1991,
booktitle={Proc. 2nd VENACO Workshop - The Structure of Multimodal Dialogue},
pages={119--126}
}