Does 'AI‑watermarking' mean the party is over for cheating students?
Concerns about students cheating with AI are rife in schools and universities.
Recent developments in AI technology have raised concerns about the potential for widespread academic dishonesty among students. Major AI providers like Anthropic have begun watermarking their text outputs to indicate AI-generated content, but this effort faces significant limitations. The watermark is hidden within the text itself and requires a detection mechanism to be identified, which itself has not yet been released by Anthropic.
Moreover, watermarking can be bypassed using other AI tools, and the technology cannot definitively determine whether the text was human-written, proofread, edited, or created by a different AI system. Even more troubling, the watermark does not account for AI assistance in proofreading, editing, or enhancing the quality of work.
OpenAI has reportedly withheld its watermarking tool due to concerns over potential negative impacts on certain groups, including non-native English speakers. Despite the widespread use of AI among students, unauthorized usage is most prevalent in written assessments completed outside of class. Educators have relied on AI-detection software to identify cheating, yet these tools have faced criticism due to their imprecision, bias, and legal challenges.
A recent survey found that 95% of undergraduate students in the UK use AI, with 94% employing it in their assessments. Universities have increasingly abandoned detection software in light of these issues. Instead, the focus should shift from policing AI usage to ensuring that students have genuinely learned. This involves moving away from "detect and punish" approaches and instead emphasizing inclusive, connected, and secure assessment tasks such as oral presentations, in-class activities, and supervised practical demonstrations.
These methods can help verify that the person completing the assessment is the enrolled student, preparing them for the integration of AI in their future work without relying on potentially unreliable detection tools.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
- Does ‘AI-watermarking’ mean the party is over for cheating students? theconversation.com