Artificial Intelligence or Artificially Intelligent?

Christmas came early for students around the globe in the form of a new, publicly-accessible technology: ChatGPT. 

ChatGPT is an internet chatbot developed by OpenAI and launched to the public on Nov. 21. It is designed to simulate a natural human conversation, doing so through the use of GPT-3.5. GPT-3.5 is a language model trained on large amounts of human-produced writing. While it does not have the ability to search the internet for the most up-to-date information, it gets most questions correct. 

Ask ChatGPT to write an essay on a well-known piece of literature or explain a mathematical concept and it will happily yield. While this is a blessing for students, it presents a problem for teachers and administrators. 

Similar to SpanishDict and PhotoMath, ChatGPT can take away the process of problem solving. If students are always able to have questions answered immediately, they do not get to think through problems and will not develop critical problem solving skills. Rather than go through the process of digesting information crucial to understanding, students can simply ask ChatGPT questions from a homework sheet or guided reading. For most students, this is an easy way to get points without spending hours studying. 

However, there is no replacement for actually doing schoolwork, and a student’s temporary facade will eventually falter as new concepts build on prior understanding. This facade of understanding ultimately works to the detriment of the student. 

Some argue that human creativity cannot be replicated by AI. While ChatGPT may not be able to replace human creativity or reliably answer complex, multi-part prompts, it can still be abused in the high school setting. 

Why? 

High school students aren’t all that creative in completing schoolwork in the first place. 

Because students are drawing information from the same lectures and discussions as their peers, they are apt to recycle phrases and ideas similar to other students and previous assignments. Most high school level courses (and even general level college courses) don’t require students to generate new ideas, except a few APs or advanced honors courses. 

Senior Julianne Binto was a student in microbiology this semester. The lab portion of the lab culminated in a week-long lab, followed by an in-depth write up that required students to explain almost all procedures and mechanisms learned in the course. 

Binto believes using ChatGPT’s writing on the write-up would definitely be cheating. “I would find it very unfair if a classmate used ChatGPT to do this assignment and still received the same grade,” she said. “For them, the assignment likely took minutes and required no critical thinking skills. They did not have to review all the previous course material to properly answer the questions, a task which takes a considerable amount of time and effort.”

Audrey Holland, instructor of the microbiology course, reflected on the importance of completing work authentically in her class. “Completing hands-on learning activities requires you to showcase a skill and then use the content knowledge from class to explain the results of tests you have performed; at least in microbiology,” she said. “Without the application of content knowledge from class, the hands-on learning and application loses some of its power and purpose.” 

Even in classes that seem to require students to generate unique ideas, like language arts or social sciences, it’s rare that a student comes to an idea that no one else has ever thought of before. Conclusions students draw about patterns in human behavior have probably already been named and studied in depth. This makes it hard to know if the student organically generated their ideas, or if ChatGPT pulled from previously written publications. 

Binto attributes the lack of originality to time restrictions and the plethora of ideas already available on the internet. “There simply isn’t enough time to be creative,” she said. “Websites such as Sparknotes provide summaries and analysis of novels, which means students don’t necessarily have to do the analyzing themselves. Even if they don’t directly copy ideas from Sparknotes, ideas from Sparknotes are likely to inadvertently show up in students’ essays.”

If humans cannot distinguish between AI and original student work in the majority of cases, teachers and administrators (ironically) need to recruit advanced technology to detect AI writing. 

Many online tools like Chegg, Quizlet, CourseHero and SparkNotes already have a plethora of information for students to use to shortcut assignments; however, with these sites, the text is published and protected by copyright laws. To copy and paste directly from these sites is obvious plagiarism (and even copyright infringement). In order to avoid detection, students spend at least a small portion of time revising and rewording the information. With ChatGPT, not only can answers to the same prompt be unique, copyright laws and plagiarism policies for AI are lacking in definitive guidance.

While copyright laws surrounding AI writing can be confusing due to the fact that the machine did not put in “intellectual labor” into creating the work, which is a traditional marker of copyright protection, plagiarism is a little more straightforward: if you didn’t organically come up with the idea, you must cite your source, even it that’s a machine. 

It can be hard to detect AI writing through traditional plagiarism-scanning services like TurnItIn; however, the company claims that it already has products capable of detecting the authenticity of student work and has more products specifically designed to detect AI writing in its research and development labs. However, many people have reported that ChatGPT has generated responses that pass through TurnItIn undetected. 

AI writing will only get better and better at replicating human writing, casting doubt on if authentication services will ever gain the upper hand or if they will always be a step behind, as they are today. 

Already, OpenAI and other companies and individuals are working to solve the issue of AI plagiarism. Princeton student Edward Tian created GPTZero, which can detect AI writing, over the span of his holiday break. The application scores text based on its complexity and on how randomly it is written. 

Students should expect the coding behind GPTZero to be made into a product available for administrators and teachers to purchase shortly. OpenAI, the creators of ChatGPT, have expressed discomfort about the potential use of their tool for plagiarism and are working to create certain markers in ChatGPT’s writing that would be unnoticeable to readers, but would be easily picked up by programs searching for those characteristics. 

Regardless of how they are identified, students submitting AI writing under their own name will face academic consequences. AI writing policies will most likely be similar or the same as plagiarism policies, although these policies will vary from school to school and possibly teacher to teacher. 

Holland says she would follow her usual plagiarism policy, starting with a one-on-one conversation about the work. “Once I start asking about the writing process over the course of the assignment and asking them to explain how they arrived at their particular answers, it usually becomes clear when work is not original,” she said. “In my experience, students who are questioned one on one usually end up identifying the work they submitted as plagiarized  and opting to submit an authentic paper for a portion of the credit back.” 

ChatGPT is the latest chapter in an ongoing battle between students and teachers. Students wanting to avoid actually completing assigned schoolwork have found ways to do so for years, whether that be by copying another student’s answers, searching the internet, or recycling old work. What makes ChatGPT different is that the technology to catch cheaters is not accessible to educators yet; students have the upper hand. 

To remedy this, OpenAI should accelerate their development of AI writing markers. Academic integrity may not be the primary concern of OpenAI, but it was irresponsible of them to release ChatGPT knowing that no technology yet existed to address this glaring problem. 

Programs like GPTZero will make their way to the education technology market, but in the meantime, it is up to students to value the long-term benefits of authentic learning over the short-term benefits of boosted grades.