Cognitive map

A cognitive map is a simplified representation aiming at to model psychological or intellectual processes. Their field of application is mainly the cognitive Psychologie and the Artificial intelligence through the concept of Agent.

Cognitive sciences are useful in a recurring way of cognitive maps: in front of the complexity of the process making it possible to explain the reasoning and the behaviors, it is indeed practical to pass by simplifying assumptions in the form of models.

A model being however a representation making it possible to apprehend more simply a aspect of a problem, many cognitive maps can cotoyer, each one bringing a particular lighting on a particular aspect.

Cognitive maps in various fields

Aaron Beck proposes, in the field of psychology, a cognitive map which is summarized in three simple degrees:

the first degree is that of the data processing: automatic thoughts, reactions vis-a-vis the stimuli of the environment. It is a not-interpretative degree.
the second degree is that contrary to interpretation of this data processing. It is the field of the cognitive adaptations i.e. inférences, conceptualizations, personalizations.
the third degree, finally, is consisted of the cognitive diagrams latent on which interpretations are based. They strongly depend on the cultural reference frame in which the agent evolves/moves.

An interesting comparison can be carried out between this model and the data-processing concepts of language, source code and achievable. The language is the external diagram with the system on which a source code rests where the actions desired for the application are expressed, i.e., interpretation that one makes of the input-outputs of the program. This source code can be then compiled in achievable which reacts without interpretation, with the intellectual direction of the term, its environment.

More than one hierarchy between the degrees, this metaphor highlights the overlap of the concepts. Same manner that there is no achievable without source code and that there is no source code without a paradigm of language, the reaction rests on a former interpretation of an environment which is apprehended through cultural diagrams external with the agent.

This approach is connected with the architectures in layer ( Layered Architectures ), quite present in the community of the systems multi-agents (the models TouringMachines and Interrap are good examples).

Models BDI

Another important class of cognitive maps in the field of the artificial intelligence is that of the models says BDI ( Beliefs , Desires , Intentions , to see diagram).

These models rest on three principal units:

the beliefs ( Beliefs ) which reflect knowledge that an agent can have on the universe to which it belongs. These beliefs can be as well true as false. One can moreover define a knowledge like a true belief.
the desires ( Desires ), or options , which represents the whole of the opportunities offered to the agent and is generated starting from the beliefs of the agent in a given moment and the longer-term objectives that the agent could be fixed.
the intentions ( Intentions ), finally, which are the options retained by the agent. They lead to an action.

Models BDI were and always today are largely used. Frameworks as Jadex make easier of the implementation of such architectures.

Other approaches for the construction of cognitive maps exist (they base most of the time on the Logique first order). It appears however that very cognitive map rests in an essential way on an interaction deepened with a unit of knowledge.

To interact with a whole of knowledge implies to be interested as a preliminary in the question of the representation of knowledge. In this field, and the same movement which saw the explosion of the quantity of information exchanged through the networks, of major advances took place with the beginning of the year 2000. Tools resulting from the Ingénieurie of knowledge (standard of serialization, ontology S,…) offer powerful and perennial solutions. And if the semantic networks were already the subject of research in the years 1980, they materialized today through technologies like the language OWL for ontologies.

Random links: