Distinguishing AI-Generated Work from Student Writing: Exploring Factors and Tools for Professors

Absztrakt (kivonat)

To guarantee students' academic integrity throughout higher education, universities have developed practices to ensure academic honesty. However, with the recent spread and commercialization of generative AI chatbots, students now have the chance to use ChatGPT and other generative AI tools in their education. These tools have the potential to enhance a student’s education, but they can also be overutilized and turned into a crutch for students who rely solely on it. This leaves the responsibility of detecting and distinguishing inauthentic student work to current professors. A responsibility which is becoming increasingly more difficult as AI advances. University professors have varying methods for detecting plagiarism and AI-generated text, but there are other factors that affect accuracy in detection. This paper covers university professors' accuracy and ability to detect generative AI writing, employing a quantitative research method through an online survey sent to higher education professors. The study includes factors that affect accuracy in detecting AI texts such as the professors’ personal details (age, gender, years of teaching experience, and department that they work in) and exposure and understanding of AI (professors’ familiarity with AI and its tools as well as confidence in the ability to recognize AI writing). Additional aspects explored are the existing tools for AI detection and their accuracy, the use of evasive soft prompts to alter generative AI’s ability to mimic human writing styles, and the levels of difficulty in detecting AI writing between ChatGPT3.5 vs ChatGPT4.0.