DAWSON'S NOTES ON GREEN (1996) CHAPTER 1
"INTRODUCTION" BY DAVID GREEN

GENERAL COMMENTS

One theme to relate this chapter to the lecture material is the notion that the mind is so complicated that we need to use different kinds of vocabularies, or different levels of analysis, to explain how it works. Green introduces a particular view of three levels that are required, and makes the case that one cannot reduce these levels into a single one (e.g., neuroscience). This is a major theme throughout the course. I disagree with Green's account of the "behavioral level", preferring something that is quite different (computational). But this disagreement shouldn't stand in the way of the general point that Green wants to establish in this introductory chapter. Pay attention to the discussion of evolutionary theory later in this chapter, in the context of levels of analysis.

Another theme to relate this chapter to the lecture material occurs in the section on "The Mind As A Representational System". In the lecture, I argue that in order to make sense of the metaphor cognition is information processing, we have to get some understanding of what information processing is. Indeed, this is a major undertaking in the course. This section in the chapter is a first pass at an account of information processing .

MARGIN NOTES

The goal of cognitive science is to understand the mind scientifically. What does this mean? To answer this question, Green contrasts scientific thinking with everyday thinking. "What matters to us as social agents is whether our knowledge of the social and physical world, and our methods for handling various problems, are sufficient to allow us to meet our practical concerns. We learn, or infer, the relationship between concepts and observations" (p. 2). Everyday thinking focuses on the particular, and does not focus on the general. In contrast, scientific thinking aims to capture generalizations: "In thinking scientifically about the mind we aim not only to generate explanations that allow us to predict observable behavior but we also want these accounts to be general and to use the fewest explanatory concepts possible" (p. 4). (NB: We will see that the need to capture generalizations is one of the primary motivations for adopting the tri-level hypothesis that will be detailed throughout the course).

The scientific study of the mind is a controversial undertaking which may or may not be tractable. Many special techniques have been developed to study the mind scientifically. (NB: Think back to your intro and cognitive courses for examples of such techniques!) These techniques have provided insights, so Green sees "no fundamental objections to the scientific study of mind" (p. 5).

The Discipline Of Cognitive Science

What is cognitive science? What is the scope of its problems? What is its set of practices? "We define the scope of cognitive science as the interdisciplinary scientific study of mind. Its practices and knowledge derive from those of the primary contributing disciplines, which are computer science, linguistics, neuroscience, psychology, cognitive neuropsychology, and philosophy. It seeks to understand how the mind works in terms of processes operating on representations. Mind, and hece the basis of intelligent action in the world, is viewed in terms of computations or information processes" (p. 5).

Cognitive science links three levels of description: 1) Behavioral, 2) Cognitive, and 3) Biological. (NB: One dominant theme of the course will be a re-working of this approach. We will talk about the computational, algorithmic, and implementational levels. The latter 2 are essentially the same as numbers 2 and 3 from Green. However, our notion of the computational level is quite different from Green's notion of the behavioral level!). A key method in cognitive science is the building of computer simulations. But, cognitive science is not the same as research in artificial intelligence, because AI is not really concerned about how people think.

"How in a nutshell can we capture the approach of cognitive science? We treat the mind as a machine whose workings we are trying to understand" (p. 7). But, cognitive science is not reductionism. "Reduction is not an appropriate goal because the mind is a representational system -- in order to understand its workings we need to understand what it is trying to do. This is why we say that three levels of description are required for a complete causal account."

The Mind As A Representational System

What does "mental representation" mean? We manipulate symbols -- not physical things. This notion of symbol manipulation requires functionalism. "Central to this approach is the idea that we can abstract from the way a symbol is represented in the brain. The precise physical nature of symbols is irrelevant to their role in governing behavior" (p. 9). (NB: The reason for this is that it is the key to capturing meaningful generalizations -- "it is the high-level program or the functional description which is the key to understanding the performance of the machine".)

There are many different types of mental representations. These include such things as images, propositions, mental models -- and many other symbol systems that you would have already seen in your cognitive psychology course. Another approach to mental representation that we will be exposed to a little bit in this course is connectionism.

Strategies For Cognitive Science

How should cognitive science proceed? Green hints at several related strategies: 1) Build models of interesting phenomena. 2) Build a cognitive architecture. 3) Synthetic psychology -- e.g., see what complex behaviors emerge once you build a simple robot and embed it in an interesting environment.

Importantly, computational level understanding is crucial to all of these strategies. "In order to understand the mind, we need to understand the nature of the problem(s) it is trying to solve.  In developing explicit accounts of mind, Marr urged that we need to formulate a view about the nature of the task to be performed. We can then go on to specify a set of explicit procedures (i.e., algorithms) that could be used to achieve the task. We can also explore the nature of the brain systems involved (that it is the implementation of these procedures)" (p. 15). (NB: How does understanding the nature of the problem being solved relate to Green's behavioral level?)

There is an evolutionary spin that can be added to the computational approach too, in order to build links with evolutionary biology. "We can ask: what kinds of problem did the mind evolve to solve?" (p. 15). "Such a view also suggests a single and universal design to the human bind, despite baring levels of current technological sophistication" because we've had 2 million years of evolving as hunter gatherers. This is one more argument that can be used to support the claim that there is a single cognitive architecture (an issue that we will face when we cover Chapter 3 of the book).

The basic idea of the evolutionary approach is that cognitive systems are useful. In other words, cognitive systems accomplish some purpose that is of benefit to an organism. For example, Shepard's view that "perception has been optimized to make the best possible inferences about the world, given the perceptual data" (p. 17). Similarly, there is Anderson's view of memory and categorization in the context of rationality, in which "the cognitive system operates at all times to optimize the adaptation of the behavior of the organism." Others argue that cognitive systems maximize the relevance of information that is processed.

But what is optimization? -- something that gives the best results given resource limitations ("bounded rationality"). The evolutionary approach leads to the view that "the human mind is likely to be modular, that is to comprise a set of systems specialized for different purposes." As we will see, much of cognitive science involves decomposing the mind into these functional parts.

The Methodology Of Cognitive Science

"At the core of science is the process of formulating theory. Theories rarely emerge full-blown, rather they develop from certain ideas or conjectures over time." Part of theory building will involve experiments, to identify interesting kinds of behavior, and to test competing theories. Theories will also be developed in the context of creating working computer simulations. Importantly, simulation work is comparative. "We suggest that the comparative testing of computational models is a second critical aspect of an adequate methodology for cognitive science" (p. 19).

The Nature Of This Textbook

Importantly, the goal of the book is not to provide facts to memorize. Instead, it encourages the reader in "understanding the debates and styles of argumentation and explanation; learning about the methods; and exploring new ways of thinking about the facts."

Pearl Street | Green's "Cognitive Science" Home Page | Dawson Home Page |