Skip to main content Skip to navigation

Overview of AI Grading of Physics Olympiad Exams

Project Overview

The document explores the integration of generative AI in education, specifically focusing on its application for grading high school physics problems within the Australian Physics Olympiad. It introduces a multi-modal AI grading framework capable of handling diverse question formats, such as numerical, algebraic, graphical, and short answer types. The study addresses various challenges and ethical considerations associated with AI in grading, underscoring the necessity for reliable, explainable, and efficient systems that can reduce teacher workloads while maintaining educational integrity. Ultimately, the findings suggest that generative AI has the potential to enhance grading processes in education, though careful attention must be paid to its implementation to ensure fairness and transparency.

Key Applications

Multi-Modal AI Grading Framework

Context: High school physics exam grading, specifically for the Australian Physics Olympiad

Implementation: A systematic literature review was conducted to identify existing automated grading techniques, followed by the proposal of a multi-modal framework that uses various AI methods for different question types.

Outcomes: The framework aims to reduce teacher workload by automating the grading process and improving grading consistency across diverse question types.

Challenges: Challenges include accurately grading complex algebraic expressions, handling free-form answers, and ensuring the ethical use of AI in education.

Implementation Barriers

Technical Barrier

Difficulty in grading complex types of responses, such as algebraic or graphical answers, due to the variability in correct responses.

Proposed Solutions: Exploration of feature extraction techniques and the use of Large Language Models to capture and translate student responses into a machine-readable format.

Ethical Barrier

Concerns about the ethical implications of using AI in grading, including privacy issues and the potential for bias in automated grading systems.

Proposed Solutions: Aligning AI grading systems with ethical principles such as transparency, privacy, and human oversight.

Project Team

Lachlan McGinness

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Lachlan McGinness

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies