A frequent question we get from educators is “What is Respondus doing about ChatGPT?” We spoke previously on this topic here, but want to expand on those thoughts.
Detecting the use of AI generated text is relatively straightforward with long-format essays. Indeed, plagiarism detection systems such as Turnitin can detect AI generated text with high levels of probability when the text is sufficiently long. But detection works poorly with short texts, such as responses to essay questions on a test.
We used ChatGPT (the AI bot itself) to explain why this is so:
Detecting the use of ChatGPT on short texts of under 300 words can be difficult… Short texts may not contain enough features or clues for detection software to analyze, such as word frequency, punctuation patterns, or sentence length.
With further prodding, the AI bot explains that detection is difficult because ChatGPT isn’t good at writing short essays in the first place.
Because of the complexity of the deep learning model and the nature of the training data, ChatGPT may not perform as well with short texts… [T]he model is more likely to generate generic or less informative responses for short texts, as it may not have enough specific information to draw upon.
Since AI generated text cannot be accurately detected with short-answer essays, it’s essential to prevent students from accessing ChatGPT during the assessment. And that’s what LockDown Browser and Respondus Monitor do. During online tests, students cannot use ChatGPT, Google, Chegg or thousands of others of resources on the internet (unless the instructor’s settings allow it). Other areas of the student’s computer and learning system are also locked down during the assessment.
For online assessments delivered in classrooms or testing centers where a human proctor is present, LockDown Browser is all that’s needed to prevent students from accessing outside resources on their computer. For online exams delivered remotely (where a human proctor isn’t in the room), the Respondus Monitor proctoring system prevents cheating on the computer itself and deters students from accessing other devices during the exam.
ChatGPT is new, but the tools to prevent its use during online tests have been around for years.