Can AI-Generated Academic Work Really Go Undetected by Educators?

ChatGPT has already infiltrated everyone's lives and it has been just two years since its release. There is a lot of work that is being done by AI, but people are also making ChatGPT do their homework, assignments and other written work. This is a serious problem, because students are using it to get their degree even though they haven't learnt anything at all. Many media and software engineers, are going to their related fields without having any understanding about what they studied. It seems that students are using ChatGPT for work, but many teachers are unable to detect AI content. Many schools have also allowed the use of AI but disallowed that use of tools that can correctly identify AI.

A study from the UK found that teachers are unable to detect any academic work generated by AI. The researchers of the study faked some students’ profiles and sent some basic AI generated content to teachers. The researchers found that almost 94% of the submitted work generated by AI was undetected by the teachers. Many studies have proved that humans themselves cannot identify AI work and the difference of linguistics used by AI and humans.

Another study from American universities in Vietnam found that AI detectors are mostly right when it comes to AI detection as compared to humans. The study said that Turnitin identified 91% of the AI generated papers correctly, while only 54.5% of the papers could be identified by faculty members. The study also used prompt engineering to make AI content as difficult to detect as possible but still AI content detectors could identify 91% of AI generated papers, and humans only 55%. But the crisis is that many schools are not using AI detectors.

• Also read: Who Owns the Most Generative AI Patents? China and the US Compete

So, not using AI detectors for AI detection and using humans instead results in AI work being missed most of the time. The UK study also found that the work that was created by AI scored higher than human created work. 83.4% of the work created by AI got a higher score than human written work. In short, if a student is using basic AI prompts for their work, they are 83% more likely to score higher and only 6% chance of getting caught.

• Related: Read This Before It's Too Late: A Guide to Highest Paying Academic Fields

But it seems like professors and teachers do not care, as well as schools that don't have any sanctions for using AI for work. A recent BBC report showed that a student was caught using AI after being detected by an AI detector but the panel cleared her as there was no solid evidence against her. Students are using AI to cheat and improve their grades, but teachers have no way to identify it. Many online classes are also being run without AI detectors which can have long term effects. But schools seem reluctant to not use AI detectors, probably because of additional work and high costs.

Image: DIW-Aigen

Read next:

• New Research Shows Generative AI can Help Teachers in Saving their Time and Making them More Productive

• FTC Bars Data Brokers From Selling Americans’ Sensitive Location Tracking Data
by Arooj Ahmed via Digital Information World