Information Technology Reference
In-Depth Information
sitional logic, or in first order logic with Modus Ponens as a basic axiom. It relies on the implication
transitivity.
Induction
is a more advanced process, that can be represented by the term 'inferencing' used
here above. It is multifaceted and might as well include
generalization
and some
syllogisms
, predictive
reasoning and bayesian (statistical) induction (Holland, Holyoak, Nisbett, & P.Thagard, 1989). As a whole
induction is a process that tries to augment the knowledge base by extending the truth value from an
assumption to another by other means than transitivity. In this chapter, we will use inductive reasoning
as a set of axioms defined in (Manna, 1974). It represents generalization and predictive reasoning.
Abduc-
tion
is the process through which similarity could be assessed or falsified: Does a fact belong to a given
law (general knowledge)? This simple question triggers the abductive process, which is an inferential
process (Josephson & Josephson, 1994), thus making abduction a possible 'subcase' of induction, seen
as a large set of different types of reasoning.
Unification
is a simple successful abduction. However, in
most of the cases abduction is associated to conflict: A given knowledge is often either contradicted by
an incoming fact, of this incoming fact cannot be attached to the current knowledge base. How does the
agent restore the broken chain or extend it to include the next fact? It needs to create a new knowledge
item (abductive induction) or to relax constraint to attach the fact to a broader knowledge. All these
issues are tackled in this chapter.
Cognitive Agents are considered as able of planning and decision making. Therefore, one has to
include goals and intentions as elements of cognition (Pollack, 1998). Reaching a goal needs reasoning.
Strategies, which are procedures derived through reasoning, consider the goal as a consequence and the
possible set of actions (the plan) as a part of the knowledge base.
Learning is a process that tackles both acquisition (capturing inputs) and reasoning (including those
inputs into the agent knowledge base, and deriving further knowledge out of them). Therefore learning
seems to be a very representative process of cognition. In AI literature, learning is not much associated
to communication. One may find several types of learning methods for symbolic agents like reinforce-
ment learning, supervised learning, neural networks and bayesian learning models that are very far from
our domain. This type of learning prepares agents for typical situations, whereas, a natural situation in
which dialog influences knowledge acquisition, has a great chance to be unique and not very predictable
(Ravenscroft & Pilkington, 2000).
Learning Through Communication
Human beings as natural cognitive agents favor dialog as their main source knowledge revision. Each
agent tends to consider any fellow agent as a possible knowledge source. The source is 'triggered' through
questioning, and information is acquired, from the answer. This question-answer process is launched to
meet a given goal, proper to the asking agent, and for which he/she has planned a set of actions (Pollack,
1998). There are several possible cases. Either the asking agent is looking for a very precise information,
and therefore, its locutor agent might not possibly provide information, therefore he/she will undertake
a research in order to find the appropriate interlocutor. Thus the plan is not tightly associated to dialog
which stabilizes a given pair of agents, but to the proper interlocutor recognition through a very brief
question-answer session. The other case is when the agent needs knowledge. (Williams, 2004) has
described an application in which agents augment a common ontology, as well as their own by sharing
knowledge and learning from each other. The author has sketched an overall communication framework
to do so. It is not properly a dialog situation, since several agents may communicate simultaneously
however, its interest relies in the fact that it designates communication as a mean of cognitive change.
Search WWH ::
Custom Search