An Artificial Intelligence-Based Program for Automated Grading of Student Work in Balancing Chemical Equations

TCE

The Chemical Educator

ISSN: 1430-4171 (electronic version)

Table of Contents

Abstract Volume 8 Issue 4 (2003) pp 271-279

An Artificial Intelligence-Based Program for Automated Grading of Student Work in Balancing Chemical Equations

Benny G. Johnson* and Dale A. Holder

Quantum Simulations, Inc., 5275 Sardis Road, Murrysville, PA 15668, johnson@quantumsimulations.com
Received June 12, 2003. Accepted July 3, 2003.

Published online: 17 July 2003

Abstract. We report the development of the first artificial intelligence (AI) grading system in chemistry. Building on prior work on intelligent tutoring systems for chemistry, an automated grading program for student work in balancing chemical equations was created. Characteristics of the system include the ability to specify the problems to be graded to the system rather than being restricted to a fixed database of prestored assessment items, derivation of grades by analysis of the student’s own work, and the ability to analyze and grade unanticipated student errors. In direct comparison with human-graded results from a large sample of authentic student work, the AI grader exhibited considerably better performance in terms of accuracy, robustness, and reliability, and thus it is capable of assisting teachers dependably in performing grading tasks. This study clearly establishes the feasibility of the new AI-based approach, and prospects are excellent for further development of advanced AI assessment techniques. The potential benefits of robust, dependable AI-based software assessment tools are considerable for helping teachers increase their effectiveness and achieve the greatest return with their time and resources.

Key Words: Computers in Chemistry; general chemistry; artificial intelligence; assessment; grading; balancing chemical equations; computer assisted instruction; general chemistry; introductory/high school chemistry

(*) Corresponding author. (E-mail: johnson@quantumsimulations.com)

Article in PDF format (502 KB) HTML format

Supporting Materials: