What types of models have been used to model higher level cognitive processes? Given the limitations of connectionist models, are symbolic models superior? Where do they succeed and where do they fail?

• Serial instead of parallel processing

• Use explicit, meaning-carrying symbolic structures as their basic unit

• Model higher-level cognitive processes based on the types of box-and-arrow models that have been presented throughout the semester

• Hypothesize explicit, localized representations instead of distributed ones.

• Generally hypothesize specific processing stages that execute as a result of sets of rules.

The most successful symbolic models have all used the list structure as their basic unit of information.

There are two primary types of list structures used:

1. Ordered lists: Lists where the order of the information matters, such as the letters of a word or the words of a sentence.
2. Property lists: Lists that contain information about the properties of the object they are associated with, such as (Color = Red, Shape = Round) for the object APPLE.

Finally, the researchers who build models using list structures as the fundamental unit of information generally subscribe to the belief that information in the mind is also represented as list structures.

Some symbolic models, such as SOAR, simply stop at the use of list structures for representing information and have processes for finding specific information by looking through the stored structures. Some models, however, theorize higher levels of organization.

ACT-R: Organizes information into a semantic network similar to the type we talked about back in chapter 7.

This network is somewhat connectionist in nature in that concepts are connected to each other and these connections have differing weights to represent different strengths of association. Information flows via spreading activation, so that once a concept is activated, so are its related concepts. Unlike a connectionist net, though, the network itself does not do any cognitive processing.

Uses a discrimination net (d-net) to organize the chunks (lists) of information it knows about.

A d-net is a tree structure where a particular node in the tree represents a test for some feature of the input list. Leaves of the tree contain the chunks of information already stored.

In EPAM, a list of information is used as the input to the d-net. The net performs tests on the input list that cause it to be sorted down a particular branch of the tree.

Once the point is reached where the list can no longer be sorted, one of two things happens. Either the sorting process has reached a leaf node containing an already known chunk of information, in which case the concept represented by the list has been identified, or else you have a new concept that needs to be added to the list.

All symbolic models conduct their processing based on explicit rules for how information should be manipulated.

These rules can implement specific hypothesized processes, such as means-ends analysis or satisficing, but the rules themselves are relatively static. It is the information that is dynamic.
 
 

The vast majority of symbolic models implement these rules as a production system.

Production systems are based upon two fundamental data structures:

Working memory elements (WME’s):

These are the list structures we discussed above.

Productions:

These are hypothesized to be exactly the same structure as the productions we talked about with reference to proceduralization, and are characterized in exactly the same fashion.

To review, productions are IF-THEN rules with an IF clause of tests that, when satisfied, causes the production to fire, or execute the actions in the THEN clause.

Whereas builders of connectionist models are making claims about their models reflecting how it is that the brain could possibly be performing a particular cognitive process, creators of symbolic models make a different claim.

They are building models that reflect what they feel is the essence of the computation taking place.

Is one more valid than the other?