What has happened? They say in Futurism that in 2024, the Australian Catholic University accused about 6,000 students of academic misconduct. At least 90% of cases were related to the use of AI for cheating. What is striking is that the university itself used an AI to issue these accusations, many of which were erroneous.
Why it is important. It is one more example that AI is still not reliable. We see it constantly with erroneous results and hallucinations. The Australian university is not the only one that has relied on AI to accuse its students, it is a fairly common practice and there have been other similar cases. The reality is that AI text detectors are also AI and, at least for now, they are imperfect.
Turnitin. It is a plagiarism detection software whose first version was released in 1997 and is widely used in universities and educational centers. In 2023 it added a tool to detect texts created with AI and it is the one they used at the Australian Catholic University. The company itself says in its user guide that the AI detector is not always accurate and should not be used as the only source when accusing a student. However, according to ABC Australia, the university used it as the only evidence when issuing his misconduct records.
The university version. Allegations regarding AI use included AI-generated works, fabricated references (hallucinations), and the use of AI tools to cite and translate content. The university says at least a quarter of all allegations were dropped after an investigation. They also rejected those in which the only proof was the AI itself and in March of this year they stopped using that software.
The dilemma. The emergence of AI tools poses challenges in the education sector. There are voices that advocate their ban, while others defend integration and promoting good practices. UNESCO published a guide for the use of generative AI in education in which it establishes standards and obligations, such as privacy protection, age limits and an approach that guarantees ethical and safe use of these tools.
Image | Turnitin
In WorldOfSoftware | A teacher corrected a final exam done with ChatGPT, but another AI evaluated it differently and exposed the dilemma
