The use of ChatGPT for plagiaristic purposes has been the fear of many educators since the AI tool was released in November of 2022. To combat this issue, Princeton University student Edward Tian has created software he believes to be the solution. Tian’s app, GPTZero, was developed in December of 2022 when the coder took it upon himself to create an app that can determine if a body of text has been created by a human or AI. Tian created the app to protect individuals in the midst of a changing technological landscape stating, “I think it is important for people to bet on independent parties to build the safeguards for their technologies.”
The technology works by measuring two factors: perplexity and burstiness. Perplexity refers to the degree of randomness present in a text. Human-created work often uses language in a more chaotic, disorganized way than AI. Hence, the more random a body of text is, the more likely it is to be human-generated. Burstiness is the second factor that refers to the complexity and diversity of sentence and language structure within a text. AI-generated material is much more uniform in style than human-created text. When individuals write, they vary their structure which creates natural “bursts” throughout the writing. AI has not been able to effectively emulate this ability. Thus, using these mechanisms in tandem, GPTZero is able to effectively determine if a text is AI generated with 98% accuracy.
While 98% effective, many have criticized GPTZero for its inability to distinguish texts that are a mix of human and AI-created. Tian emphasizes the limitations of his app stating that he doesn’t want anyone “making definitive decisions” in regard to the app’s ability to predict AI-generated work. Tian simply seeks to preserve the integrity of authentic composition by thwarting ChatGPT’s potential misuse.
Kayley Horton, Tutoring Coordinator
kayley@testprepchicago.com
(312) 848-1266