Our plagiarism detector uses "predictive text", i.e. what statistically is the next word that may follow (think predictive text when texting on a phone, or predictive searches when typing into a google search bar). If our detector finds a post to follow this statistical pattern too closely - it is flagged. Our data indicates we have a .04% miscalculation (false positive) - so if a student is being flagged, there is a 99.96% they are using an AI source. With that said, if it was their own work being flagged, this is due to the student's lack of specificity when writing. In a sense, their writing structure is too "robotic" and generic. A student can always go back and edit their post to be more thorough and detailed in their explanation! Dec. 24, 2024