Using Misunderstanding and Discussion in Dialog as a Knowledge Acquisition or Enhancement Procecss - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 80

New Constants Learning Strategy. Learning predicates remains theoretical, however, learning constants
satisfying these predicates allow the agent to apply its knowledge to the real world wherein it evolves. So
a curious 'student' agent will be able, after having learned new predicates, to ask the 'teacher' whether
it knows constants satisfying these predicates.
For example, if the 'student' learns that it should keep away from hostile agents ( hostile ( x )→
keepaway ( x )), then it should be useful for it to know such agents. For example, hostile ( Carl ), hostile ( Kevin ),
… By the way, it should be useful to know values associated to some constants. For example, if the
agent learns that the gravity value on Earth is g , then it should be interested to know the value of this
constant. Thus by proceeding to precision inquiry regarding this value, the 'student' will be able, using
the required processing skills, to use it for solving gravity problems.
Strategy For Understanding Why Some Formulas Imply Others. After learning any implication
( P → Q ), the 'student' may wonder whether ( P ) directly implies ( Q ) or if it was a result of a series of
implications. This is the typical case of an explanation request dealing with an implication. By realizing
such a request to the 'teacher', the 'student' gets the opportunity to increase the amount of its KB data
while increasing its connexity.
Symmetry Between Dialog And KB Management In Discussions Without Issues. As one can see,
dialog, modeled with Functional Roles and the student's knowledge base, present an harmonious rela-
tionship. Dialog reflects actions performed to modify the KB. The KB graph denotes the result of these
actions and its state, controlled and checked by the student's personal goals, may lead to launching a
new utterance/inquiry and therefore increasing the dialog by one step. This complete symmetry can
nevertheless be broken by a conflicting piece of information introduced by the dialog. Discrepancies in
updating the KB occur, and the two next sections handle the way dialog can be used in order to repair
such a dammage and proceed further in the lesson.
Misunderstanding Management Strategy Through Discussion: KB Connexity Preservation Com-
mon Goal. Misunderstanding, in our framework, occurs when the student is unable to link a given
predicate to its KB graph. There is also another case of misunderstanding, on behalf of the teacher who
does not understand why a student links two pieces of knowledge in which the teacher see no relation-
ship at all. But in this work, we have restricted misunderstanding cases to those related to the student's
management of its KB.
We suggest here an appropriate common strategy: Solving a misunderstanding problem by choos-
ing adequate questions and answers. We have adopted a technique inspired from the socratic teaching
method.
For each predicate p i to be taught, the 'teacher' knows another one p j linked with p i by an implication
or an equivalence F . Therefore, to ensure that the 'student' understands p i thanks to p j , it will have to
ask the 'student' if the latter knows p j . If the 'student' knows it, then the 'teacher' only has to give F to
the 'student'. Otherwise, the 'teacher' will find another formula that explains p j and so on.
The dialog of the Figure 4 is an example of such a misunderstanding and its solving, based on a the
mathematical example seen in the previous section.