AI detectors like ChatGPT and Turnitin are ineffective in catching students using generative AI. A classroom experiment conducted by a teacher showed that 12-year-olds were able to beat these detectors easily. The experiment also revealed that the detectors falsely accused students and targeted non-native English speakers. Educators are seeking innovative solutions to address this problem.
Despite my repeated discussions on this topic through my blog, it is surprising to observe that some teachers still rely on AI detectors, and I find this situation concerning.
💡 The experiment revealed that AI detectors are not effective in identifying students using generative AI, causing frustration among educators.
💡 Retrospective research on handed-in papers is insufficient; conducting real-time experiments with students is crucial to understanding the effectiveness of AI detectors.
💡 Students with limited experience in generative AI were able to bypass the detectors, proving that it doesn’t require much expertise.
💡 The majority of students were able to beat the detectors within 5-10 minutes, showing that they are not a significant challenge.
💡 The current system of accusing students without concrete evidence is flawed and undermines trust in education.
💡 Proposed solutions like Google Documents and Trojan horses are not foolproof and can be easily circumvented by students.
💡 Educators are seeking more innovative ways to address the AI problem in education and use AI for educational benefits.
AI Advisory Boards 