[Paper Review] Contra generative AI detection in higher education assessments

Cesare Giulio Ardito|arXiv (Cornell University)|Dec 8, 2023

Artificial Intelligence in Healthcare and Education4 citations

TL;DR

This paper critiques the use of generative AI detection tools in higher education, arguing they are ineffective, ethically problematic, and misaligned with modern educational realities. It advocates for replacing detection mechanisms with pedagogically sound assessment redesign that embraces AI use while preserving academic integrity and authenticity.

ABSTRACT

This paper presents a critical analysis of generative Artificial Intelligence (AI) detection tools in higher education assessments. The rapid advancement and widespread adoption of generative AI, particularly in education, necessitates a reevaluation of traditional academic integrity mechanisms. We explore the effectiveness, vulnerabilities, and ethical implications of AI detection tools in the context of preserving academic integrity. Our study synthesises insights from various case studies, newspaper articles, and student testimonies to scrutinise the practical and philosophical challenges associated with AI detection. We argue that the reliance on detection mechanisms is misaligned with the educational landscape, where AI plays an increasingly widespread role. This paper advocates for a strategic shift towards robust assessment methods and educational policies that embrace generative AI usage while ensuring academic integrity and authenticity in assessments.

Motivation & Objective

• To examine the effectiveness and limitations of current generative AI detection tools in higher education settings.
• To identify ethical and practical vulnerabilities in AI detection mechanisms used for academic integrity enforcement.
• To challenge the assumption that detecting AI-generated work is a viable or desirable strategy in modern assessment.
• To propose a shift from detection-focused policies to assessment design that integrates generative AI responsibly.
• To advocate for educational policies that prioritize authenticity and learning over surveillance and punitive detection.

Proposed method

• Synthesizes insights from case studies, media reports, and student testimonies to evaluate real-world impacts of AI detection.
• Analyzes the philosophical and pedagogical contradictions of relying on detection to maintain academic integrity.
• Examines the technical and ethical flaws in AI detection systems, including false positives and privacy concerns.
• Proposes alternative assessment frameworks that embed generative AI as a legitimate learning tool rather than a threat.
• Recommends institutional policy reforms that prioritize transparency, equity, and educational value over monitoring and control.
• Uses a critical discourse analysis approach to deconstruct the assumptions underlying AI detection in academic contexts.

Experimental results

Research questions

• RQ1 How effective are current generative AI detection tools in identifying AI-generated work in higher education assessments?
• RQ2 What are the ethical and pedagogical consequences of relying on AI detection for academic integrity?
• RQ3 Why is the current detection paradigm misaligned with the evolving role of generative AI in education?
• RQ4 What alternative assessment models can support academic integrity while embracing generative AI use?
• RQ5 How can educational institutions redesign assessments to value authenticity and learning over surveillance and detection?

Key findings

• AI detection tools in higher education are prone to high false positive rates, leading to unfair academic penalties for students.
• The use of detection tools undermines trust between students and educators and may discourage authentic academic engagement.
• Detection mechanisms often lack transparency and are based on flawed assumptions about writing style and originality.
• The paper finds that detection is not a sustainable or ethical solution for maintaining academic integrity in the age of generative AI.
• Redesigning assessments to integrate AI as a legitimate tool leads to more authentic, equitable, and pedagogically sound outcomes.
• Institutional policies should prioritize educational value and student learning over surveillance and punitive detection practices.