Journalism students 'copy' from AI's work.

TP - I chose the commentary genre for the students' final exam in Journalism Writing Skills. Commentary in newspapers naturally requires citing multiple sources. Therefore, students were allowed to use computers connected to the internet to complete the exam. Because of this, the exam room wasn't just occupied by me and the students…

To get students interested in the issue of pollution, I gave them the assignment: “As a resident of a city with one of the world’s worst air quality problems, what can you do? Write an 800-1,000 word commentary for a youth newspaper, warning about the worsening air pollution situation.” I included the note: “You may use published books and articles as source material” and “Absolutely no artificial intelligence applications are allowed in this assignment.”

Journalist Nguyen Manh Ha is pictured with renowned singers Khanh Ly and My Linh backstage at a music show. Photo: Provided by the interviewee.

Of course, this assignment, to be completed in 150 minutes, is not easy for first-year journalism students. But I didn't demand that the output be a publishable article; my main goal was for them to express their thoughts and desires regarding the environmental situation they are currently experiencing. The candidates all looked tense while working on the assignment, but none of them used up all the allotted time.

When I used S – another artificial intelligence application – to check the level of AI usage, A's essay was concluded to have "0% of the text likely generated by AI." B's essay was deemed to have 18% AI-generated text, with two suspected "AI plagiarism" paragraphs highlighted in yellow. Therefore, the conclusion of S, which clearly shows a margin of error, cannot be relied upon. Furthermore, students can easily use other applications or add their own personalization to the prompt before submitting it to the application, making it impossible for the grader to fully control the process.

When grading the papers, I was initially pleased because the quality was significantly better than the midterm exam six weeks earlier. I was sure my teaching had been effective. Some papers closely followed the prompt. For example, student A's introduction was very natural: "Every morning I wake up, pull back the curtains, look outside and see a hazy gray covering the city. It's not morning fog, but smog. I open the air quality app and am not surprised to see that the AQI (air quality index) continues to be at a hazardous level."

The writing style throughout the entire piece is concise and consistent. To confidently use the first-person pronoun "I" and present a viewpoint on a societal issue that's currently of great concern, especially for someone who just graduated from high school, is quite impressive. Then I see that same tone in B's piece: “This morning, like every morning, I opened the AirVisual app and saw the AQI in Hanoi had risen to over 230, a level that warns of 'very harmful to health'. This is nothing new to me. The sky was gray, the fog thick, but it wasn't fog, it was PM2.5 fine dust. I put on my mask, stepped out onto the street, squeezed through the stream of exhausting motorbikes, and suddenly thought: What am I breathing in? And why… am I getting used to it?”

Both A and B suggest that people install air quality monitoring apps. Neither calls for reforestation or planting more trees in the city, but only for planting indoors or on balconies. And then C insists on only greening balconies, which makes me suspicious.

At this point, I tried giving my exam question to ChatGPT to work on. In less than three minutes, I had nearly 900 words in bullet points. The "sample essay" began like this: "Waking up on a winter morning in Hanoi, I pulled back the curtains and saw the whole city shrouded in a thick layer of fog. No, it wasn't the romantic morning mist in poetry, but fine dust—an invisible enemy silently attacking the lungs of millions of urban dwellers." It's clear that A only needed to make a few minor adjustments to create their own opening.

And here, a cluster of solutions suggested by ChatGPT in just one sentence: "Don't burn trash, plant more trees on your balcony, use energy-saving appliances, and reduce electricity consumption…". B shows great skill in "applying" AI, interweaving: "You can absolutely walk or bike for short distances, limit the use of air conditioning when unnecessary, plant more trees on your balcony or rooftop, and choose a low-consumption lifestyle." And carefully changing "at" to "in". In the past, two exam papers that were somewhat identical would be considered plagiarism. Nowadays, it's more likely because the candidates used the same AI software.

As a visiting lecturer, I called the vice dean to inquire about the university's handling of this situation. She expressed her frustration, as graduation theses are now often written by AI. Then she asked me how news organizations handle reporters using AI to write articles… We'll have to wait until the news organization discovers it, right?

In any case, it's time for newsrooms to define the boundaries of collaboration with AI in reporters' work. Currently, it remains a matter of personal decision and choice, because the topic fed into the AI ​​will then only be known to the reporter.

Nguyen Manh Ha

Source: https://tienphong.vn/sinh-vien-bao-chi-nhin-bai-ai-post1752094.tpo