The Importance of Architecture

For a long time, my own research has been focused on so called hybrid systems, in particular, systems that integrate the then newly emerging connectionist models and the more traditional symbolic processing models. Such systems arguably combine the strength of connectionist mod- els and symbolic models, thus possessing a wider range of capabilities (Sun and Bookman 1994, Sun 1994). They have been used both in cognitive modeling—in understanding human cognition through developing computational models of cognitive processes, and in building intelligent systems for practical applications.

The general idea behind this body of work on hybrid systems, developing more comprehensive models through integrating a variety of techniques, can be further extended into so called cognitive architectures, that is, cognitive models that are domain-generic and encompass a wide range of cognitive capabilities.

Another perspective on this is that of a progression from engineering to science. While most of the work on hybrid systems (mostly within the field of artificial intelligence) takes an engineering approach, the research on cognitive architectures (mostly within the field of cognitive science) takes a scientific approach—focusing on gathering empirical data and developing models that serve as scientific theories and scientific explanations of the data through an iterative hypothesis-test cycle. The function of a cognitive architecture is to provide an essential framework to facilitate more detailed modeling and exploration of various components and processes of the mind.

Developing cognitive architectures is a difficult challenge. In this article, the importance of, and issues and challenges in, developing cognitive architectures will be discussed, examples of cog- nitive architectures will be given, and future directions will be outlined (Sun 2007b). In the next section, the question of what a cognitive architecture is is answered. In section 3, the importance of cognitive architectures is addressed. In section 4, to further clarify the importance of cogni- tive architectures, a multi-level framework for cognitive modeling is outlined. In section 5, some background regarding the development of cognitive architectures is provided. Then, in section 6, an example cognitive architecture is presented in detail and it applications to cognitive modeling, artificial intelligence, and social simulation described. In section 7, the significant challenges related to developing cognitive architectures are articulated. Finally, section 8 concludes this article.