While plagiarism checks were an inherent part of academics in the last decade, checking papers for AI-generated content will obviously dominate this one. Today, education is already the most prominent market for AI content detectors. Regrettably, this can cause significant problems for ESL (English as a second language) and EFL (English as a foreign language) students.
When great minds of our time warned humankind about the dangers of out-of-control AI, hardly have they been thinking about it damaging international students’ careers. Or when we were cautioned about the possibility of AI-powered tools churning out biased information, barely have we suggested that they can backfire with prejudice toward real people and actually affect them adversely in real-life situations.
The July 2023 study by Stanford University academics led by an assistant professor of biomedical data science James Zou showed that popular AI content detectors outrageously often wrongly indicate human-written papers as AI-generated. Moreover, for the texts crafted by ESL and EFL students, the rate of false positives can exceed 60%. As the study conductors note, this puts the majority of non-native students at risk of false accusations of cheating. How is this even possible when everybody actually expected online tools will make studying much easier for modern students?
The thing is, native speakers communicate and write in English in a very complicated way without even realizing it. However, for ESL and EFL students, the opposite is typically true. Their sentences are simpler, more forthright, more concise, with fewer linguistic flourishes and less variance. The primary target of a non-native student is doing things correctly rather than trying to impress a teacher with ornate rhetoric. That’s where the catch is. AI content detectors consider the lack of linguistic complexity (perplexity, as they call it) one of the main characteristics of the texts generated by large language models (LLM) like ChatGPT, Cohere, or Bard. As a result, they wrongfully target non-native English speakers jeopardizing the academic careers and reputations of millions of ESL/ENF students around the world.
The only thought about how being accused of cheating – even so, unjustly – can damage a student’s honor and mental state is utterly disturbing and calls for action. Because we clearly don’t want the taste of what the future of AI holds to be so bitter.
While AI content detector developers only start considering the measures they can take to fix the situation with false positives, ESL and EFL students need to deal with the risks of AI-generated text usage accusations right now. To this end, there are several sleek ways to beat AI bias – as well as some quite hazy yet effective ones.
1.First and foremost, you can implement one of the basic Aikido principles – use the opponent’s power and attack energy against them. In other words, use the potential AI content generators offer, as the authors of the above-mentioned study suggest themselves. The thing is, they’ve used ChatGPT to improve the vocabulary of papers by ESL and EFL students in order to mimic native-speaker language. This intervention reduced misclassification dramatically, with the average false-positive rate reducing by almost 50%, from 61.3% to 11.6%. So, it certainly looks like you can upload your text into an AI writer tool and ask it to enhance and sophisticate the text as if it was crafted by a native speaker. And voila – the paper’s content perplexity will have increased significantly!
2. Another instance of how non-native speakers can use AI to improve their texts is by asking it to generate the outline based on the provided instructions. Writing a text after such an outline might go much faster and easier while helping the author maintain the correct structure. Finally, running the completed text via an AI-powered tool like Grammarly or Wordtune for editing and proofreading purposes will make it look much more like crafted by a native speaker than an average ESL student ever could
3. As a non-native English speaker, you should also put considerable effort into developing strong research skills, be it single-handedly or with the help of AI tools. Enhancing your research skills will let you gather information from various sources effectively, learn how to critically evaluate and integrate information into your writing, and, thus, improve your writing skills along the way. By demonstrating an understanding of the topic through well-researched and well-structured papers, you can minimize the risk of wrongful cheating and/or plagiarism accusations.
4. Knowing how to properly cite and reference sources is fundamental in academic writing. If you accurately format your paper and correctly cite and reference all the sources, the chances of easily passing any AI text detection check skyrocket, as AI generators just cannot do that. Make sure to give credit to the original authors and sources that have influenced your work. Understand the different citation styles used in your academic field, such as APA, MLA, or Harvard, and adhere to their guidelines.
5. Familiarise yourself with both AI-generated texts and papers crafted by natives. Understanding the characteristics and patterns of AI-generated texts, as well as becoming familiar with their structure and common phrases, will help you avoid unintentionally incorporating AI-styled content into your writing. Fortunately, there’s no lack of online resources that provide both AI content generators and expertly crafted samples; so pick a free essay writing service to your liking, generate a paper on the required topic, and then compare the output with the relevant example from the database of already written works
6. Remember we mentioned that some of the suggested methods might be kind of hazy? Well, here it is. If you’re in a bind and unable to perform time-consuming experiments with AI writing tools or search for samples to use as writing models, you can turn to an experienced native English paper writer for help. The Internet hosts a plethora of specialized services and freelance platforms where you can get your piece crafted within hours by a human expert – and this text will definitely pass virtually any AI content detector. However, keep in mind that such services are typically paid, and you are not supposed to submit the received papers as your own unchanged.
7. Last but not least, don’t neglect the opportunities AI provides for learning and improving your English writing skills. Whether you use one of the above-mentioned tips or not, AI-based tools and chatbots can help non-native English speakers enhance their grammar, spelling, writing style, tone, and vocabulary by tailoring guiding materials to their needs, evaluating and giving feedback, or providing virtually free tutoring.
While the role of AI content detectors in supporting academic integrity is – and obviously will long remain – quite significant, quick measures must be taken to protect ESL and EFL students from the inadequacy such systems bring to the table. In light of the recent findings, it is crucial to quickly develop robust and equitable methods of AI-powered content checking. They shouldn’t be based on a one-size-fits-all approach and reflect a great variety of possible users. On the other hand, teachers in particular and educational institutions in general shouldn’t completely rely on AI content detectors, even though some of them claim to have up to a 98% detection accuracy rate. Until AI is brought back into an acceptable framework and all categories of students are protected from wrongful accusations, isn’t it only right to apply the Blackstone principle? Just to remind, it says, “It’s better that ten guilty persons escape than that one innocent suffer.” Sounds legitimate, doesn’t it?
