If some artificial intelligence programs were able to write texts convincing enough to make a Google engineer believe they had feelings, it was only a matter of time before someone started to use these tools for other, darker purposes, like cheating in class. According to some students on the Reddit forums, they’ve added A’s since they started using OpenAI to write their class assignments.
And of course, this raises an obvious technological debate.
cheating at school. Many students criticize that writing articles and essays for college is anything but encouraging. Not only must they present an original and well-argued thesis, but they must also support the texts and structure the thought accordingly. Others, however, prefer to go faster. A few weeks ago, Reddit user innovate_rye shared that his college professors hadn’t realized in a long time that all of their articles were written by an AI. “We had to write five good and bad things about biotech. I messaged the AI asking exactly that and it generated a response that gave me a 10 for the exercise.”
Without AI, this student explains that it would take hours whereas now it only takes about 20 minutes. He brags about it in this Vice article like this: “Being able to do it faster and more efficiently seems like a skill to me.”
How it works? In OpenAI, the student simply enters text from one or more sources into the program to initiate the process. The program then generates content using a set of parameters on a topic, which can then be customized to the author’s specifications. With a little practice, a student can use it to write their paper in minutes. Moreover, this system is not considered “plagiarism” because there is no previous work that the student has copied, and therefore no original that the detectors can recognize.
It all works through NLP, or Natural Language Processing, which is a branch of AI that deals with understanding and generating human language. Using a machine learning model, it is able to predict which word should appear next in a sentence, based on a set of training data. This is called “Transformer” technology, i.e. a kind of improved version of text message auto-completion.
A trend. In fact, professionals have been using automated article writing software for months to write blog posts, digital advertisements, product descriptions, and other marketing copy. Now, we see more and more students using this same technology to brainstorm and write essays with high-quality, well-written, well-argued content. There are even several AIs marketed for it.
Since OpenAI introduced the latest Application Programming Interface (API) for its widely used language model, GPT-3, students have begun to input messages written in OpenAI Playground and similar programs to generate sufficiently convincing to convince teachers.
Debate. It’s always the same: Is it cheating to use an AI for this? A student on Reddit defended the following: “I have the knowledge, I have the lived experience, I am a good student, I go to all the tutorials and lectures and read everything we have to read, but I felt like I was penalized for not writing eloquently. And I don’t think that’s the right thing to do. However, many professors think the opposite, that the system gives them an advantage unfair by using systems designed to create marketing content and not for academic work.
Many educators believe that using these programs violates academic integrity. Georgetown University professor Lise Howard explained in this Slate article, “I think it’s unethical and an academic violation to use AI to write paragraphs, because the work academic is about original writing.” The issue has already been addressed in university integrity codes. For example, American University states, “All documents and materials submitted must be the student’s original work unless the source is cited.”