Skip to main content Skip to navigation

The Imitation Game: Detecting Human and AI-Generated Texts in the Era of ChatGPT and BARD

Project Overview

The document examines the impact of generative AI, particularly large language models (LLMs) such as GPT and BARD, on the education sector, emphasizing both its applications and challenges. It raises significant concerns regarding academic integrity, as students may misuse AI-generated content for their assignments. To address these issues, the research presents a novel dataset comprising both human-written and AI-generated texts across diverse genres, assessing various machine learning models' effectiveness in differentiating between the two. The findings reveal that while LLMs can generate text that closely mimics human writing, there are viable detection methods available to identify AI-produced content. This capability is essential for preserving academic standards and integrity in educational environments. Overall, the document underscores the dual-edged nature of generative AI in education, highlighting its potential benefits while also advocating for strategies to mitigate risks associated with its misuse.

Key Applications

Detection of AI-generated vs human-written texts

Context: Educational institutions and contexts where academic integrity is a concern

Implementation: Utilization of machine learning models to classify texts from a novel dataset comprising human and AI-generated texts

Outcomes: Effective differentiation between human and LLM-generated texts, improving academic integrity checks

Challenges: Difficulty in detecting AI-generated texts, especially from GPT in creative writing tasks

Implementation Barriers

Ethical

Concerns over academic integrity due to students potentially using AI tools to generate assignments.

Proposed Solutions: Development of effective detection tools and discussions on ethical guidelines for AI usage in education.

Technical

Challenges in accurately differentiating between human-generated and AI-generated texts, particularly with sophisticated models.

Proposed Solutions: Continued research and refinement of machine learning models to enhance detection capabilities.

Project Team

Kadhim Hayawi

Researcher

Sakib Shahriar

Researcher

Sujith Samuel Mathew

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Kadhim Hayawi, Sakib Shahriar, Sujith Samuel Mathew

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