Human-in-the-Loop AI for Cheating Ring Detection
Project Overview
The document explores the integration of generative AI in education, highlighting its transformative potential and key applications, particularly in enhancing online assessment integrity. It introduces an innovative human-in-the-loop AI system aimed at detecting cheating rings during online exams by analyzing keystroke and mouse movement patterns to identify suspicious behavior indicative of professional cheating. This approach addresses significant security concerns related to online assessments while emphasizing the importance of adhering to Responsible AI standards, ensuring that ethical considerations and fairness are prioritized in the detection process. The findings indicate that such AI systems can effectively enhance the reliability of online testing environments, fostering academic integrity and trust in digital education platforms. Overall, the document underscores the role of generative AI not only in improving educational security but also in shaping a more equitable assessment landscape.
Key Applications
Human-in-the-loop AI cheating ring detection system
Context: High-stakes online language assessment for test takers
Implementation: Integrated into the existing online examination platform, utilizing data collection methods and pattern analysis.
Outcomes: Promising performance in detecting cheating rings while maintaining fairness across demographic groups.
Challenges: Potential biases in human proctoring decisions and the need for broader fairness metrics.
Implementation Barriers
Technical
Challenges in ensuring the AI system's fairness and accuracy across different demographic groups.
Proposed Solutions: Future research on broader fairness metrics and continuous adaptation of the system to evolving cheating methods.
Ethical
Risks associated with privacy and potential societal biases in AI systems.
Proposed Solutions: Adhering to Responsible AI standards to protect privacy and mitigate biases.
Project Team
Yong-Siang Shih
Researcher
Manqian Liao
Researcher
Ruidong Liu
Researcher
Mirza Basim Baig
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Yong-Siang Shih, Manqian Liao, Ruidong Liu, Mirza Basim Baig
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