Constraints on the Processor

Summary prepared by Group #3:

Zubin Tiku, Mark Domino, and Litza Stark

The purpose of studying the constraints on the processor is to determine its initial states. The three areas of study to consider are input/output, memory storage and access, and cognition (computation).

The processing power (neurons) of the human brain is heavily skewed towards input as opposed to output. Input is vital as a filtering device; our minds absorb and disregard more than what they note. For example, at a cocktail party, we may notice our name in another conversation even if we're not paying conscious attention to the conversation.

Our senses are energy detection mechanisms. Taste and smell are chemical detectors, touch and hearing mechanical detectors, and vision an electromagnetic detector. Each of the senses has a finite range, although we are biased more towards some forms of input (vision) than towards others (smell).

The organization of the mind reflects the organization of the world; the location of a memory is related to its content.

That is, the brain is organized into topogrpahic maps, where the structure of the information to be coded is preserved in the organization of the information as it is coded; memory in organic computers is content-addressable.

The two major topics to discuss in terms of memory are storage and access. Popular concepts of memory state that there are three types: sensory, short term, and long term. Sensory memory is an extremely brief storage of sensation as perceived by sensory receptors. Short term memory is "working memory," consisting of seven ±1 chunks of information. Debate ensues over whether long term memory is episodic (a sequenced memory of a specific event), semantic (information in abstract form), or both.

There are a number of processes which can affect storage and access, such as rehearsal, consolidation, and activation (see Stillings et al.). Another phenomenon that affects retrieval of information is the primacy vs. recency effect. This means that given a sequence of information, people tend to remember the first and last elements more readily than the middle. The early information is remembered because there is more rehearsal time, whereas the later information is remembered because it is retained in short term memory. This is an example of similar ecological effects with different algorithmic sources.

Two forms of encoded memory are declarative and procedural, which can be thought of as "knowledge that" and "knowledge how," respectively. Any skill being learned starts as declarative knowledge, and becomes procedural once it is learned entirely.

Cognition could be defined as the manipulation of computable representation (mental content). Mental content can be expressed in terms of logic, concepts, and images. Logic is the computation of abstract propositional structures from which conclusions are deduced. Inferential knowledge is affected by information not necessary to solve the problem (it is cognitively penetrable). Concepts could be described in two ways: as fuzzy prototypes (a "best instance") or as necessary and sufficient definition (limits). Images are straightforward; they seem to be faithful to the real world (distances in the mind and in the physical world are proportional, for example).

Processes of cognition and mental operations are manipulations of mental content. There are two types of processes: automatic, which are fast and non-deliberate (reading, movement); and controlled, which are slow and more deliberate (learning to read).

The goal of examining these cognitive mechanisms is to determine the core universal mind. What is our he factory-installed equipment? The limitations on our intrinsic mental content are the limitations on our core universal mind.

Experiments performed on children lead to ideas that we start with initial knowledge of objects, numbers, geometry, and people. We seem to begin with a basic physical knowledge of the world; that objects have continuity, cohesion, and contact. Shape and form are more important than secondary attributes like texture and color. We also start with a knowledge of increments and decrements, as well as basic concepts of Euclidean geometry, such as two lines are capable of intersecting. Lastly, we know that people can affect each other at a distance, unlike other objects.

The paper by Gallistel on comparative representation and processing uses animal learning and behavior to gain insight into human cognition. The experiments performed upon barn owls showed clear-cut examples of critical periods in development. He also cited experiments that supported the idea of domain-specific knowledge. Interchangeable with the that concept is privileged response, the notion that organisms have innate tendencies towards certain cognitive skills. The initial knowledge within animals is typically understated and consist of minimal expressions.

Discovering the nature of the processor is a continuing process, each experiment revealing more complexity. The debate continues concerning the nature and content of cognitive processes, and the nature of initial knowledge.

Summary of Spelke and Gallistel