Data files: GenAI Detection Tools, Adversarial Techniques and Implications for Inclusivity in Higher Education

This data consists of samples of human and AI generated short form content in the form of MS Word files. AI generated samples are subjected to adversarial techniques to evade content detection by GenAI text detectors. Our accompanying paper 'GenAI Detection Tools, Adversarial Techniques and Implications for Inclusivity in Higher Education' discusses the results of the testing. This data can be used to test how effective AI text detectors are at determining whether a sample is human or machine generated. We identify how the application of adversarial techniques reduces the ability of AI text detectors to accurately determine the source of the sample.