Improving Automated Feedback-BuildingAGenericRule-FeedbackGenerator

Date: 2007-09-27

Time: 11:45

Room: BBL room 471

Speaker: Eric Bouwers

Title: Improving Automated Feedback

Building a Generic Rule-Feedback Generator (thesis defense)

Abstract

The importance of feedback in learning a procedural skill cannot be ignored. It is therefore surprising to see that many of the existing educational tools aimed at learning these skills fail to provide feedback on a level above “this is incorrect”. Those tools that do manage to provide higher quality feedback are the result of several years of research. Transferring this research to different domains is not interesting research material, and therefore not done by the academic world. On the other hand, commercial implementers do not seem to want to spend large amounts of resources on improving the feedback part of their software. This is probably because the trade-off between investing in research and the expected profit is of no commercial interest.

We aim at solving this problem by combining both existing as well as new research into one generic framework. Instantiating this framework on a particular domain results in a configurable tool that can provide feedback after each step in solving an exercise. The quality of this feedback ranges from a simple ‘incorrect’ to a custom-made feedback message, depending on the level and amount of configuration that is used. As an example, we instantiate the framework on three different domains and evaluate the feedback with several domain-specialists. The result is a solid framework that can serve as a basis for future research as well as a prototype for further development.

slides.pdf: Slides of the presentation

-- EricBouwers - 28 Sep 2007