In Short:
Over the past year, students have submitted more than 22 million papers that may have used generative AI, with 11 percent potentially containing AI-written language. Turnitin, owned by Advance, released this data on their new AI writing detection tool. ChatGPT’s launch raised concerns about the use of AI in student writing, with some teachers struggling to detect AI use and potential bias in detection tools.
Students’ Use of Generative AI in Academic Writing
According to new data released by plagiarism detection company Turnitin, students have submitted more than 22 million papers that may have utilized generative AI in the past year.
Impact of Turnitin’s AI Writing Detection Tool
A year ago, Turnitin introduced an AI writing detection tool trained on student papers and other AI-generated texts. Over 200 million papers have been reviewed by the detector, predominantly from high school and college students. It was found that 11 percent of papers potentially contained AI-written content, with 3 percent flagged for having 80 percent or more AI writing. Turnitin assures a false positive rate of less than 1 percent when examining full documents.
Challenges and Concerns
While tools like ChatGPT offer quick information synthesis, concerns about generative AI’s accuracy and biases have been raised. Despite risks, students use AI chatbots for research and as ghostwriters, with traces even appearing in published academic writing. Teachers face difficulties in detecting AI use and some are incorporating generative AI into grading processes.
Evolution of AI Detection
Detecting generative AI poses challenges as it creates original text, leading to nuanced usage by students. Aside from large language models like ChatGPT, word spinners are used to rewrite text and may evade detection. Turnitin’s detector now includes capabilities to identify word spinners and work modified by tools like Grammarly’s generative AI assistance.
Concerns of Bias and Accuracy
Detection tools may exhibit bias, with English language learners at higher risk of false positives. Turnitin’s detector has been trained on diverse writing samples to improve accuracy. Studies have shown Turnitin to be among the most accurate AI language detectors in comparing undergraduate and AI-generated papers.
