Matsuda05c - Matsuda, N., Cohen, W. W., & Koedinger, K....

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
41 Building Cognitive Tutors with Programming by Demonstration Noboru Matsuda, William W. Cohen, Kenneth R. Koedinger School of Computer Science, Carnegie Mellon University 5000 Forbes Ave., Pittsburgh PA 15213 {mazda,wcohen,koedinger}@cs.cmu.edu Abstract : The aim of this study is to incorporate the technique of programming by demonstration (PBD) into an authoring tool for Cognitive Tutors. The pri- mary motivation of using PBD is to facilitate the authoring of Cognitive Tutors by educators, rather than AI programmers. That is, instead of asking authors to build a cognitive model representing a task to be taught, a machine-learning agent – called the Simulated Student – observes the author performing the target task and induces production rules that replicate the author’s performance. FOIL is used to learn conditions appearing in the production rules. An evaluation in an example domain of algebra equation solving shows that observing 10 prob- lems solved in 44 steps induced 9 correct and 1 wrong production rules. Two of the correctly induced rules were overly general hence produced redundant solu- tions. 1 Introduction This study considers the application of programming by demonstration (PBD) to an unusual task domain: constructing intelligent tutoring systems, or Cognitive Tutors. Cognitive Tutors are known to be an effective means of tutoring students in various topics including algebra, chemistry, and physics [1]. However, building a Cognitive Tutor is difficult, as it requires knowledge of the subject matter, a good understanding of the prior abilities of the students who will use the system, and extensive program- ming skills. Our goal is to allow educators – i.e., people with knowledge of the sub- ject matter and students’ abilities, but little programming skills – to build Cognitive Tutors. To accomplish this, we wish to construct a system in which an author can construct a GUI for a Cognitive Tutor, and then use this GUI to present examples of how the human students should solve problems. A PBD learning system, called the Simulated Student , will then generalize these examples and build a set of production rules for solving problems in the task domain. It is clearly desirable for the Simulated Student to learn a cognitive model that can be easily communicated with the authors. It is also clearly desirable for the Simulated Student to learn the “correct” generalizations (the ones intended by the authors) as quickly as possible. Less obviously, it is also useful for the Simulated Student to pro- duce generalizations that are incorrect, but consistent with those that a human student might produce: such incorrect but plausible generalizations are often incorporated in the Cognitive Tutors to model possible human errors. There have been a number of studies reported so far to integrate PBD into an au-
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 6

Matsuda05c - Matsuda, N., Cohen, W. W., & Koedinger, K....

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online