Although people still debate how effective AI Detectors are, most of academia has made its stand and concluded AI Detection tools are essential to maintaining academic integrity. Some universities may have different perspectives on the technology, MIT doesn’t believe AI Detectors work, but for the most part, if professors and faculty didn’t employ AI content detection tools, far less works submitted would be human-written text at all.Regardless of policy though, “Do AI Detectors Work” is a question of functionality and different AI writing detectors have different capabilities.
With a 4% false positive rate and millions of essays being handed in, there have to be hundreds of thousands of false accusations being thrown around. For this reason, AI detectors used improperly to pass judgement might be creating more trouble for academia than just letting the AI plagiarism pass might.
And here is how ChatGPT's essay fared against AI Detect's software:
AI Detect works with surgical precision and unparalleled power. Not only delivering a 1% Human score to the AI-generated text but also shedding light on the text's readability and grade level. With this tool, content creators and students alike can learn so much about their text that they may not have known otherwise.
Table of Contents
- How do AI Detectors Work?
- How Reliable Are AI Detectors?
- Is There A Free Version of Turnitin?
- Conclusion
- FAQ
How Do AI Detectors Work?
The AI-generated text from ChatGPT reads robotically for a reason. Large language models like ChatGPT are programmed to write for utility, meaning the writing style uses precise and simple language you can understand. The same writing style is employed whether you need an academic essay or car repair instructions.AI Detectors are trained to measure the perplexity and burstiness of any piece of text and if those two metrics are low, signaling simple, precise word choice, sentence length, and sentence structure, then the AI detector tools will determine the input was AI-written text.If the perplexity and burstiness of the text are high and feature the sort of randomness and complexity that is signature to human writing, then the AI plagiarism checkers will determine the text came from a human writer.In order to bypass an AI detector with AI writing tools, you will need to employ an AI humanizer to take your AI-generated content and exchange the watermarks of AI writing with the watermarks of human writing.The popularity of these AI humanziers and undetectable AI technologies has contributed to the argument that AI Detectors don’t work. If no one is using AI models that don’t write undetectable text, then what does it matter if AI detectors are functional against regular chatbots? Still though, a good AI detector will be able to catch undetectable AI.How Reliable Are AI Detectors?
Lets say you’re an English professor and you want to determine if you’re students are using generative AI to write their essays. So, you pass it through Turnitin and it gives you a high percentage of AI writing present, somewhere around 70%. Can you then unequivocally determine the writer didn’t author the essay themselves?The answer isn’t a clear and concise “yes”. Turnitin has a 4% false positive rate at the sentence level. Stanford also did a study on numerous AI detectors and found a bias against Non-native English speakers. The reason for this is simple. Using what we learned about AI writing style in the previous section, it isn’t hard to see how some student work could end up similar to AI writing.As long the human writing is precise and simple, perhaps by a student that doesn’t understand the nuances of English, then AI's writing style and human writing style can be visualized with a Ven diagram, where in the middle lies the spot where false positives come from.This is why many educational institutions have ethics policy for faculty using AI detectors just as they do with student's using AI writing tools.Is There A Free Version of Turnitin?
Turnitin is an exclusive detection system. In order to access it, you need privelidges bestowed upon an academic status as student or faculty. It’s integrated into both Canvas and Blackboard, the two largest student submission portals, but if just any content creator wants to use it, they won’t have an easy time acquiring access.Not to mention, Turnitin ain’t free and doesn’t have any kind of free trial for you to test. What the regular person needs, whether for student work or content creation, is a free AI detector with just as much functionality as Turnitin, if not more. And trust me, it’s out there.The prestige that comes with mainstream acceptance doesn’t make your tool work no matter how much you wish it did. AIDetect offers an AI content detector with just as much power as Turnitin but it’s made for everyone, offering a free version that’s free of charge and free of judgement.By free of judgement, I mean that AI Detect isn’t made to punish creators that want to use AI writing. People that take advantage of new technologies can use AI detect to test their work against an AI detector before putting it out in the world where people are hungry to take issue with popular content. Professors and students alike are encouraged by AI Detect to use the detector and humanizer to learn the in’s and out’s of AI text.I don't want to just tell you about how well AI Detect works, I want to show you. Here is an essay generated by ChatGPT about AI detectors.