Educational Testing Service

Martha L Stocking

Leonard C Swanson

Charles Lewis

A method and system for item exposure control in computer based testing. A set of exposure control parameters representing the probability that an item selected from an item pool is administered to a test taker for each item in the pool is generated through computer based test simulation. The exposure control parameters can be initialized using values less than one that reflect the expected value after the test simulations are complete. The items in the item pool are ordered from most desirable to least desirable to administer based on predefined criteria. The set of exposure control parameters are used to generate operant probabilities associated with each of the items and represent the probability that the item is selected to be administered given that no more desirable item has been administered first. Items are randomly selected to be administered based on their corresponding operant probabilities. Stimuli may be included in the item list, and when randomly selected, the set of items associated with the stimulus are administered. The operant probabilities can be adjusted by item security parameters which are defined based on a desired measure of test security. A system for carrying out the steps includes an exposure control parameter generator, an item selector, and an item administrator.

Method and system for controlling item exposure in computer based testing

A method is disclosed for incorporating into the construction of adaptive tests expert test development practices. The method is an application of a weighted deviations model and an heuristic for automated item selection. Taken into account are the number and complexity of constraints on item selection found in expert test development practice. The method of the present invention incorporates content, overlap, and set constraints on the sequential selection of items as desired properties of the resultant adaptive tests, rather than as strict requirements. Aggregate failures are minimized in the same fashion as in the construction of conventional tests. The extent to which restrictions on item selection are not satisfied is then the result of deficiencies in the item pool, as it is with conventional tests.

Growing community of inventors

Hopewell, NJ, United States of America

Martha L Stocking