Empirical Studies of Procedural Flaws, Impasses, and Repairs in Procedural Skills.

Abstract

This report represents the results of several extensive empirical studies. 925 students who were in the process of learning subtraction were tested using highly diagnostic tests developed by DEBUGGY. Some students were retested two days later to measure the short-term stability of bugs, and others were retested several months later to study long-term stability. All tests were analyzed by DEBUGGY and by several expert diagnosticians in order to assess DEBUGGY'S diagnostic abilities. It was found that DEBUGGY was as good as or better than human diagnosticians at discovering bugs that explain a student's errors. However, a third of the students who committed errors could not be modelled with bugs and slips. Moreover, bugs were found in general to be unstable rather than stable, both in the short term and the long term. These findings challenged the belief that bugs and slips alone could account for procedural error data. Repair Theory was orginally developed as a generative theory of bugs, one that predicted which bugs would exist for a given procedural skill. However, it also predicted that certain kinds of non-bug, non-slip performance would exist, both in the static analyses and the stability data.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 05, 1982
Accession Number
ADA113791

Entities

People

  • Kurt VanLehn

Organizations

  • University of Pittsburgh

Tags

Communities of Interest

  • Biomedical
  • Human Systems

DTIC Thesaurus Topics

  • Artificial Intelligence
  • California
  • Cognition
  • Computer Programs
  • Computer Science
  • Computers
  • Databases
  • Military Research
  • Motivation
  • Navy
  • Operations Research
  • Psychology
  • Social Sciences
  • Students
  • Training
  • United States
  • Uss Carl Vinson

Readers

  • Artificial Intelligence