(07-31-2024, 10:45 AM)David1477 Wrote:

(07-31-2024, 10:35 AM)ReyMysterioso Wrote: A Stanford study on the issue revealed seven different AI detectors flagged writing by non-native speakers as AI-generated 61 percent of the time.
It's a real problem.
https://themarkup.org/machine-learning/2023/08/14/ai-detection-tools-falsely-accuse-international-students-of-cheating#:~:text=They%20published%20a%20paper%20last,that%20incorrect%20assessment%20was%20unanimous.

So according to Stanford, I have better chances of getting through the AI CHECKER (!) by using AI than writing the assignments myself as a non-native student. Damn, the world is really a mess nowadays. I never expected such stupid issues to be possible.

Unfortunately, nobody came up with a solution yet, right?

(08-03-2024, 11:11 AM)ReyMysterioso Wrote: I personally don't think a solution is possible. AI has now improved enough that there is no definitive way to detect AI used by a smart prompter. Education is just going to have to accept it's a tool everyone will use for writing and research in the same way scientific calculators are accepted tools in math classes and stop trying to fight it. Some will argue "but there's no way to make sure anyone is really learning anything." But I argue that's always been the case. Functionally illiterate students were able to graduate from high school for decades. Really unqualified and undeserving students have been able to get a degree for generations by having other people write their papers and do their homework. It's not that AI broke education. It's been broken for a long time yet so many opportunities require degrees, so we continue doing this performative dance.

(08-05-2024, 07:48 AM)harrywilson Wrote: It sounds frustrating to deal with these AI detection issues, especially when you're putting in genuine effort with your assignment. It's unfortunate that your school only accepts reports from a single tool, particularly when different tools can give varying results.

Here are a few suggestions that might help:

Rewrite in Your Own Words: If possible, try to rewrite sections of your assignment in a more conversational or personalized style while keeping it academic. This can sometimes help in making the text appear more "human-like" to AI detectors.

Use AI Humanizers: Tools designed to humanize text can sometimes adjust the tone and structure in ways that make it less detectable as AI-generated. While you've tried some tools, experimenting with different ones, like MyEssayWriter.ai's AI humanizer (https://www.myessaywriter.ai/content-humanizer), might give you a different result. These tools aim to make text sound more natural and less formulaic.

Manual Revision: Focus on making your writing more varied and personalized. Use more specific examples, personal insights, or unique expressions. This can sometimes help in distinguishing your writing style from what AI tools might produce.

Seek Academic Support: If your school has an academic support center or writing lab, consider asking them for guidance on improving your writing style. They might offer insights into making your work sound more authentically human without compromising academic professionalism.

Check for Common Phrases: Sometimes, AI detectors flag common phrases or transitions. Ensure your writing avoids overused academic phrases and try to vary your sentence structures.

Avoid Over-Reliance on Tools: While AI detection tools can be useful, they aren't perfect. If the school is adamant about using one specific tool, make sure you fully understand how it works and try to align your writing style accordingly.

If none of these methods work, and you're still facing issues, documenting your efforts to address the problem can be important. Even though it hasn't helped in the past, keeping a record of your attempts to solve the issue can be useful if you need to escalate the problem further within the institution.

I hope these suggestions help, and you’re able to resolve the issue without compromising the quality of your work.

(08-07-2024, 07:14 AM)studyingfortests Wrote: As has already been mentioned, it is widely understood that AI detectors commonly return a false positive result on content written by non-native speaker.  This is thought to be related to specific vocabulary and the use of less complex language compared with native speakers.

One suggestion I have is to try originality.ai.  It isn't free, but by a wide margin seems to be the most accurate AI detector out there. It flags content and for each individual phrase, offers a percentage score of how likely it is that the content is AI.

I found that in places where other AI detection tools found my original writing (I am a native speaker) to show a 50-60% likelihood of AI generation, Originality found a 0 likelihood.  (I tested it with AI generated text, too, and it caught that almost 100%.)

Originality charges by the size of the document you submit.  I think a 10 page assignment cost 40 cents or something.  You buy $5 or 10 worth of credits.  I'm still using my original purchase from. a year ago.