Sat. Apr 27th, 2024

New technology, from the written word in 3400 BC to automated chatbots in the 21st century, has often shaken the foundations of the human routine and provoked public disorder. However, each new technological development often ends up becoming integrated into our lives. The recent release of Chat Generative Pre-Trained Transformer (ChatGPT) has prompted society to contemplate whether complete integration is possible.

Credit: Clara Chae

ChatGPT is an artificial intelligence language model launched by technology company OpenAI in November 2022. The program can engage in conversation with the user and generate complex prose and analysis. Other skills include writing unique essays, replicating writing styles, producing code, and solving mathematical equations in seconds and at no cost for anyone with internet access. Such capabilities have provoked concerns among educational institutions about the future of academic integrity and education as a whole.

Unlike other forms of artificial intelligence, ChatGPT generates different text for each user, even when given the same prompt, making its results difficult to identify. According to OpenAI, the program functions on a reward-based reinforcement learning system, meaning the AI evolves in response to user feedback to improve the text it produces. Thus, each interaction makes ChatGPT’s responses more unique, harder to recognize, and easier to plagiarize.

Additionally, the software does not attribute the authors of its training data in its responses, which creates problems concerning intellectual property. “Who owns these creations?” History Teacher Conrad Pitcher asked. “How do we protect intellectual property in AI?” The program functions solely to produce human-like text, meaning students do not know whose work they are referencing or the credibility of the information it presents.

Mr. Pitcher also said that ChatGPT’s priority is not to produce accurate writing, but rather to “make the writing seem as if it’s human-produced.” The program’s wordsmithing is able to easily make incorrect information seem factual, setting potential traps for students.

In addition to humanities courses, ChatGPT could also affect scientific and mathematical disciplines, especially as it improves. Despite the program’s language proficiency, ChatGPT cannot accurately compute high-level mathematical problems at the moment.

“Many of the things that it doesn’t do very well we already have a tool for, such as Wolfram Alpha,” Mathematics Teacher Samuel Leitermann-Long said. “Some of the questions that ChatGPT solves very well, while showing the set of steps it uses, are reasonably basic word problems. I think that there’s the potential that those two pieces of technology combine in a way that makes them really powerful.”

Despite its potential application in the classroom, ChatGPT may exacerbate economic inequality in schools, impacting the learning process as a whole. According to Mr. Leitermann-Long, ChatGPT is currently open to all internet users in order to collect as much data as possible to improve the program. However, following OpenAI’s switch from a nonprofit to a for-profit organization in 2019, as reported by The Wall Street Journal, the program could become very expensive. The ChatGPT website even states that the program is a “free research preview.”

Mr Leitermann-Long said, “A worry that I often have at Deerfield, but also at a lot of places, is about tools or resources that are available to some of our most privileged students and not to many of our other students. We already know that there are financial disparities in the amount of resources you can bring to bear on your work. It’s possible that [ChatGPT will make] that worse.”

Additionally, prominent artificial intelligence softwares such as ChatGPT may create a dangerous reliance on the technology. Dean of Faculty Ivory Hills said, “[We] have to do this counterfactual thought experiment. Who are the individuals that society trained to be capable of inventing AI? And if you live your whole life with AI, do you develop the skills required to invent AI? Are we hobbling ourselves by becoming too reliant on tools, thus not exercising the part of our brain that would allow us to have invented those tools? Or are we gaining superpowers so that we can invent even better tools in the future? I don’t know.”

Due to such concerns surrounding academic integrity and equity, schools and institutions worldwide are grappling with how to best accommodate ChatGPT. According to The Wall Street Journal, New York City Public Schools banned the program entirely on school networks and devices beginning January 3, 2023. The Chronicle of Higher Education urged more educators to consider more in-class writing assignments, oral assessments, and utilizing GPTZero, a software developed by Princeton University student Edward Tian designed to identify ChatGPT-generated text. The Chronicle also suggested transforming curricula to exploit ChatGPT’s limitations, such as highly subjective and argumentative writing.

However, Dean of Academic Affairs Anne Bruder stated that Deerfield might take a more adaptive approach. “We’re not going to outsmart this technology,” she said. “We’re going to be caught flat-footed if we’re just continually trying to squeeze into the little available spaces where ChatGPT isn’t.” As ChatGPT improves, those spaces likely will shrink – until, as some believe, artificial intelligence will become fully incorporated into education.

Teachers may also be able use ChatGPT’s deficiencies to assess student comprehension. Mr. Leitermann-Long said that ChatGPT’s mathematical mistakes “mirror the misunderstandings that students might have. I could present this and say, ‘This model has fundamentally misunderstood some things. Let’s identify those errors.’”

However, some wonder if the program could ever be suitable for classroom use. Zoë Matias ’23 asked, “If you’re just feeding ChatGPT a task, isn’t it destroying the whole point of formulating your own ideas?”

According to English Teacher Heather Liske, any AI-generated text inherently lacks an essential property of all writing: human personality. She said, “In an algorithmic construction, fingerprints of [the writer] are absolutely invisible, at least to me, and what’s most compelling about any piece of work is that you can hear a particular voice.”

Ms. Liske also expressed one of her primary concerns regarding ChatGPT: devaluing quality and overvaluing quantity. “It seems like an expression of the pressure that people feel to produce. If ultimately, that’s what students care about, then [Deerfield is] not doing [its] job as an institution,” she said. “I want my students to care about the process and care about themselves as writers and care about each other in a classroom community. But if we’re seeing ChatGPT more and more, then clearly, it’s evidence of a student focus on something else that’s not the classroom community, not writing in the company of other writers.”

Other educators believe that ChatGPT could reframe how schools approach education entirely. According to Dr. Bruder, teachers will have to push students to incorporate their own experiences and reflections into their writing in order to evade dishonest use of the software. Concerning Deerfield’s core values of citizenship, face-to-face interactions, connectedness, reflection and balance, pursuit of mastery, and shared experiences, Dr. Bruder said, “They’re not things that can be mastered through AI,” suggesting that Deerfield should not fully embrace the integration of this technology as an educational tool.

Yet the motto, “Deerfield Academy prepares students for leadership in a rapidly changing world,” could contradict the idea that Deerfield should not prioritize the use of artificial intelligence. As Dr. Hills said, “There could come a day or a time, and it could come very soon, where it’s a disservice to our students to not let them know how to use [AI].”

Thus, Deerfield and artificial intelligence may not be fundamentally incompatible. According to Dr. Hills, “We’re going to have to, as a community, engage in a society-wide conversation, trying to remind ourselves of what is important.”