Methodology

From Cognitive Architecture Wiki

Jump to: navigation, search

The task of developing a general cognitive architecture is a daunting task, greatly complicated by the lack of a consensus definition of either intelligence or cognition. Current systems demonstrate nothing like the efflorescence of adaptation evident in human behavior and such a result seems very far away. To make any progress, researchers must set goals much lower and hope to move gradually to the intelligent ideal.


However, this lowering of goals results in great deviations among the AI research community when answering some fundamental questions. What capabilities are most important? What types of environments should be anticipated? What properties should be included? A research team's answers to these questions greatly influence its architecture design, or methodology.


Unfortunately, no answer is completely satisfying or justified. For instance, most researchers consider the ability to plan fundamental to intelligence. However, the subsumption architecture's designers purposely ignore this capability. Both sides provide compelling arguments for their choices, so the final decision is highly subjective.


Thus, the answers to the questions must be regarded as assumptions, made by researchers to constrain the design space. Since such constraints greatly influence the final architecture, it is important to keep them in mind when considering the design decisions made. Use this list to view the methodological assumptions of any given architecture.



Return to the Table of Contents

Personal tools