Academic Fraud and Artificial Intelligence
Plagiarism
One of the immediate concerns regarding academic fraud is the use of Large Language Models (LLMs) that generate convincing and coherent text that students and researchers can pass off as original work. The growth and inclusion of generative text into academic articles has been widespread, particularly in the computer science literature1 with estimates ranging as high as 17%2.
Examples are widely shared in social media3 and are becoming increasingly common.
What is acceptable use?
Many journals like Nature4, Science6, Elsevier5, and IEEE 7 have similar requirements for using Generative AI in academic articles. They all allow the use of LLMs for copy-editing tasks like grammar checking. Another commonality in these guidelines is the disclosure of using of LLMs or other AI, with some journals requiring the specific AI system to be included in the article.
Just as a reminder: LLMs have been used for copy-editing purpose for the content of this workshop.
Some of the journals prohibit the listing or attributing authorship of an article by a LLM model46.
A general recommendation is to be overly cautious and disclose any use of Generative AI in the creation of academic work unless prohibited by the journal.
Use of LLMs in Education
In this blob post8, Alice Evans provides the following guidance for educators:
- Cultivating AI Literacy by designing classroom activities that use LLMs like Claude to help answer and explain questions on the class material.
- Design Assessments that exceed current AI functionality by assuming that students will use AI but that require students to have a deeper understanding of the source material and use deadlines to limit the time a student has to complete the assessment.
A further note of caution from a recent paper9 about a study that demonstrated that allowing students to use GPT 4.0 in their course improved their performance but when the LLM is no longer available, the students perform worse than students who have never used a LLM.