Automated tools for distance learning

Noelle LaVoie, Joseph Psotka, Karen E. Lochbaum,
and Charles Krupnick

Abstract

The US Army War College conducted an extensive asynchronous threaded discussion on US national security policy and the future of NATO with 20 sections and approximately 300 participants. The collaborative discussion generated 6 MB of high quality text interchanges that simulated the actual US government interagency process of developing national security policy and strategy. Automatic text understanding tools based on Latent Semantic Analysis (a machine learning software system that understands and represents text in ways similar to humans) were used successfully to assess the quality of groups' and individuals' contributions, and to separate administrative comments from content contributions. This presentation will discuss the results of this exercise, which showed how effective and powerful software agents can be built to moderate online discussions, alert instructors to breakdowns in group communications, monitor content in the discussions, advise and mentor e-learners, and perform embedded assessment of group and individual products.

Full Paper (HTML)