Automatic Knowledge Generation:

For any adaptive instructional system to operate effectively, it needs to include some means by which the system builds a model of each individual student using its services. The software components which accomplish this are typically referred to as automatic knowledge generation engines (AKGs). There are, in fact, two AKGs incorporated into MediaMatrix. The first of these, user tracking, is common to many computer-based training and instructional/tutoring systems. The second, the sematic network mirroring system, is totally unique to MediaMatrix and is the most powerful element underlying its design and services.

All questions incorporated into MediaMatrix's tutorials, regardless of question type or content, share a common evaluational goal. That common goal is to ascertain the degree to which a student has established a specific association between conceptually relevant terms that help to define a semantic network. Every question is coded for its inclusion of at least two conceptually related terms.

To illustrate what this means, we may refer to the last sentence in the above paragraph (the one in italics) as a working case study in question construction and coding, as well as how these processes help to establish a data base for semantic network mirroring. The sentence might be used to generate the following question:

Every ________ is coded for its inclusion of at least two ___________.

"Question" fills the first blank and "conceptually related terms" fills the second. If the question were changed to incorporate only a single blank, it would either say:

Every question is coded for its inclusion of at least two _______.

or alternatively:

Every ________ is coded for its inclusion of at least two conceptually related terms.

In either case, the question may be coded on the basis of its evaluation of a student's knowledge of whether the word or phrase "question" and "conceptually related terms" are somehow related!

The question type which supplies to most informational prompting for the student is the multiple choice question. In other words, a question composed in fill-in-the-blank form is given with a list of alternative words or phrases which might apply. If answered correctly, we have at least some evidence that the student associates, in the case above, "question" and "conceptually related terms."

To begin to fade the heavy use of informational prompting in this question, a slightly more difficult variation might by the simple use of the question in fill-blank form, but where no selection prompting is given as to the possible correct associated phrase to fill the blank. This is thus the second-level question type used by MediaMatrix to gather data on whether the student associates "question" with "conceptually related terms."

A second type of fading is one which fades the context-defining words of the sentence itself. That is, we can eliminate all but the most important element of the sentence (the element that it was coded upon!), which in the above case is the term "question." To make a question of this term, we can use a paired-associate style of presentation wherein we ask:

In your opinion, are the terms or phrases:

"question"

and

"conceptually related terms"

Related / Not Related

It should be obvious by now that the answer should be "Related"!

Ultimately, a fourth type of question may be generated wherein the near-maximum amount of fading of any question-interpretational prompting. This type of question is Verplanck's (199x) "association" test item. In the example above, we might supply the following as an evaluation item:

Question:

a. ________________ b. _________________

c. ________________ d. _________________

Where correct associated terms might include: "conceptually related terms"; "multiple choice"; "fill-blank"; "paired associates "; "prompted association"; and even "fade prompts."

From answers given to any of these types of questioning, we can begin to construct a mirror image of the student's ever-emerging semantic network of terms, with some degree of confidence that he/she has established some knowledge about the form and purpose of questions in MediaMatrix. Of course, other questions coded for associations among markedly different terms will help us form a relatively complete picture of what is known by the student and what is misunderstood. This is the knowledge generation engine at work, and its intelligent application allows for the delivery of adaptive guidance, adaptive tutoring, and adaptive assessment -- all of which become highly integrated services in MediaMatrix's approach to adaptive instruction.

Next Topic... Custom Publishing