Completed: Recombinant Knowledge Generation for Planning, Training and Assessment Using Latent Semantic Analysis

Principal Investigator: R.J. Sternberg
Co-Principal Investigator: Jennifer Hedlund

Description: One of the advantages of traditional multiple-choice tests is that they are easy to score. But the advantage of these tests if also their disadvantage: They are easy to score because they have "right" and "wrong" answers, whereas many problems in life are not nearly so black and white. Thus essay tests have a distinct value as do multiple-choice tests. Ideally, one would want to use converging operations to measure skills in a variety of different ways. In this study, we will be applying Landaulet's LSA model to the scoring of essays derived from our work on tacit knowledge in military leadership. We will be examining whether the computer program associated with LSA can provide an enhanced and more objective level of scoring for our assessment instruments.

Funding Agency: U.S. Army Research Institute
IRB Approval by: Not Applicable
Funding Period: 1999 - 2002
Award Amount: Approximately $170,000