The recent revolution in Large Language Models has increased the accuracy of machine translation and other AI services that can be useful for studying in a foreign language. At the Hopkins-Nanjing Center, an international relations- and area studies-focused graduate program where Chinese and international students take classes in their second language, many students have been taking advantage of new AI tools to assist with their schoolwork.

For some, this has meant assistance with things like reading comprehension and generating ideas for essays; but for others, it has meant fully translating readings into their native language or translating completed assignments into their target language. The rapid development of these new tools and ambiguity surrounding their use at the HNC has sparked discussions about what constitutes fair use of AI as a study tool, and what crosses the line into academic dishonesty.

The SAIS Observer spoke with Chinese and international students about their use of AI in the program. Some asked only to be identified by their first name or using a pseudonym in order to protect their privacy and due to the vague rules around AI at HNC.

HNC is more than a “language program” in that the primary metric for success is not language improvement. With longer and more readings, students sometimes feel they have to choose between learning the language or learning the subject.

“I feel like it is impossible to do everything here,” said one international student. “The way I see it, and in weighing what is important to me, I decided spending my time working on spoken Chinese was more worth it.” She says she typically translates her Chinese readings into English and uses ChatGPT to learn the Chinese keywords. “It allows me to spend my time elsewhere such as speaking with Chinese students and working with my language partner.”

“It all depends on how much time I have,” said John, an international student who says he usually reads the Chinese original text and an English translation side-by-side. He says he reads only the English version about quarter of the time.

Kelany, another international student, said she never uses machine translation to translate entire readings, nor does she use text-based AI-generated article summaries, as she believes these methods interrupt her own process of analysis. She added that she makes an exception for Google Gemini, an AI program released in September, which generates a realistic-sounding podcast discussing a text.

“When I listen to a podcast,” she explained, “I’m thinking about what they said and coming to my own conclusions.”

Beyond raising personal decisions on how to balance learning the target language and learning the content, AI also poses questions about academic honesty.  Students disagree about whether writing an assignment in one’s native language and then machine-translating it to the target language is a form of academic dishonesty.

Ari, an international student, believes it crosses that threshold: “Pretend it’s a human. I think that it’s the easiest way to think about all these questions … If someone’s doing the whole thing for you or translating the whole thing for you, it’s not yours now.”

A Chinese student, using the pseudonym Lanxiao, says there is nothing dishonest about using machine translations, and argued that the source of the ideas expressed, rather than the language used to deliver it, is what matters.

“The food is the content, and the language is just the plate or the bowl,” Lanxiao said. “[AI programs] are created to let people do the creative thing and let the machine do the dirty work.”

Another Chinese student, using the pseudonym Wei, uses machine translation similarly: “I’ll translate [the full text] once and then go back to check it bit by bit,” correcting any deviations from his original meaning, Wei said. Lanxiao expressed that overall, AI massively benefitted her language learning, while Wei thinks his own use of AI has hurt his English learning this semester.

HNC Co-Director Adam Webb, also a professor of political science at the HNC, was not certain about these cases. He says full-text machine translation assignments at least contained “a whiff” of academic dishonesty. He added that “in a bilingual program where everybody’s supposed to be doing academic work directly in both languages there is something which crosses the line if one is just writing something in one language and having it completely translated using AI.”

The HNC Standards for Academic Honesty have been updated in recent years to specify that using AI “to generate written material” constitutes academic dishonesty, though whether translation counts as “generation” remains unclear.

“Over the next semester or year or two we will continue refining this based on what we observe,” Webb noted.

Webb also described ways in which he believes the system self-corrects for students overusing AI, raising the possibility that more advanced AI in future years may be able to detect AI-generated content.

He also echoed sentiments expressed by Hua Tao, a professor of Chinese studies at the HNC and at Nanjing University, that a student’s overreliance on AI for written assignments may indirectly show up in their grades when they have trouble speaking in class about something they are supposed to have researched in the target language.

Hua says he has suspected only one student of over-relying on AI for a presentation, and he is actually more concerned that his students are not using AI as much as they should. Hua showed off his homepage which contains links to at least five language learning models that he uses regularly for various academic purposes including translation. He emphasized that he hopes students will use AI to deepen their understanding of the content of his classes.

Paul Amstrong-Taylor, a professor of economics at the HNC, is enthusiastic that AI-based machine translation has made it easier for his students who are non-native English speakers to turn in their work for his class. Like Hua, he does not mind if his students machine-translate their assignments.

Huang Jie, an assistant professor at Nanjing University’s School of Government who teaches a class on Chinese government and politics at the HNC, said he can accept students using AI for all kinds of purposes, because whether he likes it or not, use of these tools is inevitable. He said that overuse of AI could be disastrous for elementary school students who need to memorize basic knowledge, but that it was an important tool for students at universities, adding that AI could not make students’ judgments on the issues they are researching for them.

To illustrate his point, Huang imagined a scenario in which someone is contemplating marriage with the person they are dating: “I think if you ask AI it will certainly tell you … to consider her interests, her appearance, her family. It will tell you a lot of factors: disappointments, what are your difficulties, what are your strengths… As for the final decision, there’s no way it can make that for you, is there?”

With ambiguous rules and seemingly broad acceptance from professors, students are making their own choices on how to use tools to get the most out of their time at the HNC. Distinct rules about the use of AI and machine translation, written or unwritten, have yet to emerge.

Edited By: Jay Figuerdo and Jordyn Haime