Schools brace for evolution of written word with CHATGPT

When Shelbee Eigenbrode asked an auditorium full of students at Arvada West High School whether they'd heard about CHATGPT, a surge of hands shot into the air accompanied by youthful tittering — and groans from their teachers.

The teens at Jeffco Public Schools' annual student technology convention earlier this month were not as forthcoming when Eigenbrode — a machinelearning architect at Amazon Web Services and the event's keynote speaker — asked what they've used the much-buzzed artificial intelligence program to do.

The main response: shifting eyes and shrugs, although 15-year-old André Smith said he heard about students using CHATGPT to complete their homework for them.

But he hasn't done that himself, he added.

“One teacher told us that it writes like you're in an honors class — you know, like it's good — and not to use it because they'd be able to tell,” Smith said.

CHATGPT, a web-based chatbot that uses artificial intelligence to churn out unique and high-quality text responses to users' questions and prompts, is so powerful that technology and education experts are describing its arrival as a watershed moment for the future of the written word.

The program looks similar to chatbots or search engines where users type in prompts or questions to be answered. Unlike chatbots that spit out standardized, pre-programmed answers or search engines that link to other sources, CHATGPT answers queries by scraping the internet and crafting a unique response of its own, making it largely imperceptible to plagiarism checkers.

English teacher Amber Wilson and her colleagues at Denver's Thomas Jefferson High School recently put the program to the test.

“We asked it what the major causes of World War I were, and it thought for a minute and wrote a beautiful five-paragraph essay in about 15 seconds about the causes of World War I, and it was pretty good,” Wilson said. “It's a powerful tool.”

Educators at K-12 schools through the collegiate level are bracing for students to use CHATGPT to write their essays and do other homework as the latest advancements in AI technology have become more accessible in recent months and more user-friendly than ever before — it's a free program anyone can try.

“We are at an inflection point right now,” said Casey Fiesler, a University of Colorado

Boulder associate professor of information science. “This is going to be huge. This is going to dominate discussions of ethics and policy for quite a while. It's going to change the way we evaluate learning.'”

Colorado universities, school districts and educators — like their peers across the nation and globe — are once again at the forefront of figuring out how schooling should adapt to a new technology. Like the advent of calculators or the proliferation of smartphones, CHATGPT is being heralded as a sign of the ever-changing times, and an opportunity to rethink writing.

Not stifling innovation

Jill Ibeck, the chief information officer at Jeffco Public Schools, said the district will continue to rely on its existing academic dishonesty policies when it comes to students using CHATGPT or other AIS, but that it's important to recognize technological advances are par for the course.

“We have to make sure we aren't stifling innovation,” Ibeck said.

So far, none of the school districts or universities The Denver Post reached out to plan on banning the use of the artificial intelligence program on their networks like New York City schools did last month.

Instead, educators are striking a balance between enforcing existing plagiarism consequences if a student misuses CHATGPT while understanding that the program can be utilized for the advancement of learning, too.

Denver Public Schools officials said the district is working to see how new cutting-edge technologies can be used to support students and what preventative measures can be put in place to curb their misuse in the classroom.

Representatives for the University of Colorado

Boulder said the campus's honor code already covers plagiarism-by-chatbot, but that the university recognizes artificial intelligence can be used in unique and appropriate ways in the classroom. CU Boulder is hosting workshops and panels for faculty on how to adapt their teaching methods for the age of artificial intelligence.

The University of Denver published a blog post explaining the pros and cons of the program and providing resources for faculty to learn more about it.

As CHATGPT has exploded in popularity since its debut in November, human writers have speculated whether the program marks the death of the college essay or the downfall of homework entirely.

Is CHATGPT the undoing of the English class as we know it?

Wilson doesn't think so. She imagines a future where CHATGPT is used as a starting point — an opportunity for students to skip getting stuck in the weeds on outlining or essay structure and, instead, spend more time on developing critical thinking skills.

Maybe students will ask the program to write an essay on Mary Shelley's “Frankenstein,” Wilson said.

CHATGPT puts together a solid skeleton of an essay about the book, but Wilson said its creativity is lacking. (This reporter's editor rejected her use of a Chatgpt-generated opening paragraph for this news article — intended as an example of its capabilities, and fully disclosed — on the grounds that what the program came up with was too boring.)

Wilson said the writing samples she's seen produced by artificial intelligence lack depth, quotations and sourcing — and don't offer much in terms of a writer's voice.

“Why don't we use that as a starting point?” Wilson said.

“Put in those quotations. Is the thesis everything it needs to be? If we think about writing that same way and take away that one hurdle — the basic structure of essay writing — and have them think longer on critical thinking, is that not something that is worth trying?”

Academics at DU and Metropolitan State University of Denver said in these early days of getting a grasp on the technology, there should be a focus on learning and teaching the ethics of artificial intelligence, such as how to be transparent about its usage.

“I think we'll see people leaning into discussions of academic honesty, academic integrity and also tying in professional ethics, so having this more philosophical discussion in addition to thinking about the ways it may play out in terms of assignments,” said Leslie Cramblet Alvarez, director of DU'S Office of Teaching and Learning.

While teachers have had to watch for students' falsified writing since the dawn of homework, Cramblet Alvarez acknowledged that needing to ferret out assignments written by CHATGPT could feel like yet another responsibility added onto educators' overflowing plates.

“This is coming on the heels of three years of really intense change for people, especially in education, so I think there is a little bit of exhaustion,” Cramblet Alvarez said.

“Having to continue to learn and contend with one more thing can be frustrating.”

Wilson stopped assigning essays as homework long ago, she said, opting to test students' writing abilities during class time because she was sick of grading cobbled-together Wikipedia entries written in a voice totally separate from the students' in-class productions.

There are a few programs available that claim to detect whether text was written by an AI, but so far none seem to be 100% effective. Programs such as turnitin.com are already widely used by teachers to scan essays for plagiarism, but CHATGPT responses likely wouldn't get flagged since they're uniquely generated — a statistical prediction of what word should come next, CU Boulder's Fiesler said.

“The students that want to take shortcuts will, and it shows up when they have to do something on their own,” Wilson said.

A whole new world

CHATGPT is not infallible. When Fiesler asked the program to add eight plus five, it insisted the answer was 15 until she corrected it. When asked to make scholarly citations, CHATGPT cited real authors and real academic journals, but fabricated the titles of nonexistent articles, she said.

The potential for spreading misinformation is real, Fiesler said, but also an opportunity to teach media literacy.

“One way to think about teaching students how this might be used as a tool and its limitations is to show a response generated by CHATGPT and ask them to say whether it's correct or incorrect and prove it,” Fiesler said. “Tell me whether this is a good argument or a bad argument. I think that's a very interesting idea.”

Fiesler said to consider CHATGPT as if it were a tool like spellcheck. Would a teacher discourage a student from using a spellchecker? Probably not, Fiesler said — unless the student was taking a spelling test.

“We have to figure out what we would consider a spelling test in this case,” Fiesler said.

Headlines are flying about CHATGPT passing law, business and medical exams and the implications of its ability to accrue knowledge.

Though Fiesler doesn't foresee robot doctors in the near future, she does think CHATGPT will enable human doctors to walk around with a new supercomputer in their pocket to consult.

“Some people are going to say we're just making everyone stupider,” Fiesler said. “It's not that. It's that you're learning other things and allowing people to succeed in the current world we're in.”

Fiesler envisions a future when people will study the art of generating a stellar AI prompt that will produce the best-intended results.

“These are the kind of skills we need to succeed in the modern world as our world changes,” she said.