Does ‘AI-watermarking’ mean the party is over for cheating students?
If this new development – designed to show if a text is AI-generated – sounds too good to be true, that’s because it is.
Concerns about students cheating with AI are widespread in schools and universities. Lecturers have even been worried that students may be "lobotomised by AI" due to the spread of AI-detection tools, which are not always reliable. Recently, New South Wales banned take-home assignments for Year 12 students in an attempt to curb AI use in assessments.
In response, generative AI developers have introduced a method known as "AI watermarking" to identify AI-generated text. Anthropic, a major AI system provider, announced that Claude, one of their five AI tools, would include a hidden label or watermark indicating the text is AI-generated. This watermark will also be embedded in the file itself for Claude-generated files such as Word and PowerPoint documents.
The primary purpose of this watermark is to comply with the European Union's new AI act, which requires providers of AI systems that create synthetic audio, video, images, or text to identify the outputs as AI-generated. Other tech companies like OpenAI have also developed text watermarking tools, but have not made them publicly available.
However, this watermarking solution is not foolproof. Similar to AI detection tools, watermarking can be bypassed by using other AI tools. Moreover, the watermark does not guarantee whether the text was human-written, and it may not work on small samples of writing or tell whether a different AI system wrote the text. Despite these limitations, the concept of watermarking raises questions about the effectiveness of cheating prevention measures in educational settings.
Oral exams have been making a comeback as an alternative to catch AI cheating, but they also have their own drawbacks. Some educators argue that focusing on policing AI usage instead of assessing what students have learned can be counterproductive. Instead, a shift should be made towards creating assessments that are fair, inclusive, and suited to the current AI-driven society.
Higher education regulators have suggested various alternatives to the "detect and punish" approach, such as in-person assessments, connected assessments throughout a degree program, and secure assessments that verify the identity of the student completing the work.
Written by urgent.news from The Conversation AU's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
