Course adaptation to diverse student populations

Student populations (especially Internet audiences) can vary widely in learning characteristics such as preferred learning mode, intelligence, verbal level, and background knowledge. Teachpoint lets you create courses that are polymorphic: that is, students are shown different lessons and exercises, and/or different navigation interfaces are used, according to their individual learning characteristics. For example:

Teachpoint's online analytic tools help you decide where polymorphism is needed. Teachpoint provides authoring tools that facilitate polymorphism: you can create many different versions of a course without proportional cost.

The model for course adaptation

Teachpoint provides a framework for course adaptation that is easy to use in simple cases, but is extremely general and powerful. The underlying model is as follows:

• Each content item has associated content properties of arbitrary type. Some of these (currently reading grade level (RGL) and depth of interest (DOI) , both floating-point quantities) are defined by Teachpoint. Courses may use their own additional properties. The values of content properties are specified by the course developer. Initially they are "best guesses"; they can be adjusted according to analytic data.
• Each student has associated student properties. Again, Teachpoint defines the RGL and DOI properties. The values of these properties for a particular student can be determined in various ways; the simplest of these, demographic-property mapping, is specified by the course developer; see below.
• Given a content item and a student, there is a expected effectiveness function (EEF) that predicts how well the content item will teach (or exercise) that student. The default EEF is the sum of squares difference of the RGL and DOI properties.
• The structure of a Teachpoint course (as expressed in its course document, described later) can contain rules for content selection based on EEF. For example, you can give a list of alternatives, and specify that the alternative with the highest EEF should be shown.

Hence, in the example from the figure above, the given adaptation policy is not specified directly, but rather by assigning different RGL or DOI values to the content items, then defining a demographic-property mapping and course structure that result in the given policy.

This model, thought it seems complex, has the advantage that the issues of how to estimate content properties, how to estimate student properties, and how to structure a course can be dealt with separately.

Demographic property mappings

What determines the RGL of a student? Ultimately, Teachpoint might provide course-independent ways (e.g., reading-comprehension tests) of estimating such properties. For now, there are two main alternatives:

• A course can ask students to estimate their own properties.
• A course can define a mapping from student demographics (age, sex, education level) to property values.

A demographic-property mapping is specified by an XML document that contains a list of rules. Each rule consists of a set of ranges for one or more demographic properties, and a set of student properties. The property values for a particular student are obtained by finding the first rule for which the student matches the demographic constraints.

Next-> The XML documents that comprise a Teachpoint course