While writing a recent paper, we found ourselves constantly concerned about the use of certain terminology that could lead to the text being seen as being generated by AI. And when I mentioned this to my colleagues and students, it became clear that they too had similar concerns when writing papers and essays. A new wave of paranoia seems to be sweeping through higher education as everyone becomes amateur experts in AI detection.
In discussions about the writing of students and colleagues, I often hear phrases like “I can tell it was written by an AI.” A recent conversation that gained attention on social media revolved around certain vocabulary that was considered overly common in content generated by ChatGPT. For example, “cultivate,” “dig into,” “within the realm,” “endeavour,” “excite,” and even “.” Other adjectives identified as AI's favorites included “admirable,” “innovative,” “meticulous,” “complex,” “remarkable,” and “multiple.” “Uses” and so on.
But before we appoint ourselves as judge and jury, we should be very careful about whether our colleagues and students really wrote what is being attributed to them. One problem is that her human AI detective, like any human-programmed mechanical detective, has limitations and biases, especially when it comes to non-native English. This can make a non-native speaker paranoid about fully participating in academic discussions for fear that her research will be seen as generated by her AI. This can stifle children's creativity and prevent them from developing an authentic voice, which is an essential component of effective writing and critical thinking.
Second, assuming that specific words or phrases indicate the use of AI excludes the possibility of a wide range of expressions and overlooks the diversity of native English usage around the world. For example, “delve” is often used in former British colonies such as Nigeria. This necessarily means that Nigerian students and writers are more likely to use nigeria in academic discourse than other English speakers, and are disproportionately likely to be accused of using A.I. It means that.
Campus Resource Collection: Introducing AI Transformers like ChatGPT. What's next?
The third problem actually stems from the fact that detecting AI-generated content is not as simple as counting the frequency of certain telltale words or phrases. This is not how AI works. Things like ChatGPT are trained on massive datasets from human writing across genres and contexts. Therefore, the AI model does not develop its own dialect. They just regurgitate the language given to them. So, I use words like “search'' because words like “search'' appear relatively frequently in existing literature.
This is another reason why we are foolish to believe that we are experts in AI detection. In any case, all of these proof words are out there, and the list can vary from one “expert” to another. From a writer's perspective, this is a goalpost in terms of what useful phrases to avoid if you don't want to risk your writing being flagged as being generated by her AI. means that it is always moving.
Then there's the question of who feels entitled to judge whom. Determining who has the power to detect and whose work is subject to detection involves power relations across the world. But even if someone feels they have the right to judge, human judgment in this area has shortcomings, so in reality, the epistemic authority to determine whether someone else has used AI means no one has.
Police operations without the necessary expertise can lead to false accusations against students and writers. Those who are incapable of contesting such accusations may suffer disciplinary and reputational consequences as a result. The paranoia caused by such a crackdown is such that students and writers often use familiar terminology to avoid accusations that their work was generated by AI, even when that term best expresses what they want to convey. You may be forced to stop using it.
Given the legitimate academic potential of generative AI, we must also be wary of creating an atmosphere of universal condemnation and suspicion that promotes its use underground. While some journals do not allow its use, more advanced journals (some highly prestigious) already allow the use of AI assistants for grammar and spell checking, and even data analysis. doing.
In this age of AI, academic integrity must be redefined and the purpose of education reevaluated. Perhaps a focus on teaching students and staff to use generative AI responsibly, rather than trying to catch students and staff using generative AI without regard to the ethics of intent. Maybe there is a need.
After all, scouring papers and essays for words that seem suspiciously related to generative AI won't stop people from using them. And if they are doing so transparently and responsibly, should we even try to stop them?
Lillian N. Schofield is a senior lecturer in not-for-profit management at Queen Mary University of London, and Xue Chow is an expert in entrepreneurship and innovation.
