Beyond Writing Assistance: Generative AI, Multilingual Meaning-Making, and Authorial Agency in EMI Critical Writing Classrooms
This qualitative study of 26 Indonesian undergraduate students reveals that in an EMI critical writing course, generative AI functions not merely as a writing tool but as a sophisticated epistemic mediator that supports multilingual meaning-making, argument construction, and the critical enactment of authorial agency, leading to the proposal of a Human–AI Epistemic Mediation Framework.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you're trying to build a massive, intricate Lego castle, but you've been handed a instruction manual written in a language you only sort of know, and the pieces are all jumbled up. That's what writing a serious academic essay in English feels like for many students in Indonesia, where English isn't their first language. They have big, complex ideas in their heads (in Indonesian or local dialects), but they need to build those ideas into a solid structure using English bricks.
For a long time, people thought Artificial Intelligence (like the chatbots we all know) was just a super-fast robot brick-layer. You'd ask it to build the castle, and it would spit out a finished tower. But a new study by researchers Pasca Kalisa and Muhammad Zaki Pahrul Hadi suggests that's not what's happening at all. Instead of being a robot builder, these students are using AI as a super-smart, multilingual co-pilot that helps them figure out how to build the castle in the first place.
Here is the story of what they found, based on the experiences of 26 university students over one semester.
The Four-Step Dance with the Robot
The researchers watched these students write two types of essays: "hortatory" (trying to convince people to do something) and "analytical" (breaking down how things work). They didn't just look at the final essays; they looked at the messy middle part—the chats with the AI, the drafts, and the students' own notes. They found the students weren't just copying and pasting. They were doing a four-step dance:
1. The "Wait, What Does That Mean?" Phase (Conceptual Clarification)
Before writing a single sentence, the students used the AI to untangle their brains. If a topic was about "fuel crises" or "fertilizer prices," the students would ask the AI, "Explain this to me like I'm five."
- The Analogy: Think of the AI as a translator for ideas, not just words. One student said the AI helped them understand how a fuel crisis connects to fertilizer prices. The students weren't asking the AI to write the essay; they were asking it to help them understand the topic so they could build their own argument. It was like using a flashlight to find the right Lego pieces in a dark room before you start building.
2. The "Language Switcheroo" Phase (Multilingual Meaning-Making)
This is where it gets really cool. The students didn't just think in English. They thought in a mix of Indonesian and English. They would write their messy, brilliant thoughts in Indonesian, then ask the AI, "Okay, how do I say this fancy idea in academic English?"
- The Analogy: Imagine you have a secret code (Indonesian) that holds all your deep thoughts. The AI acts like a magical bridge that lets you walk across that code to the English side without losing the meaning. The students weren't just translating words; they were using the AI to keep their original ideas safe while dressing them up in the formal suit required for school. The study suggests this "translanguaging" (mixing languages) is actually a superpower for learning, not a crutch.
3. The "Let's Make It Stronger" Phase (Argument Construction)
Once they had their ideas and their English words, they used the AI to build the walls of their castle. They asked the AI, "Is my logic sound?" or "What evidence do I need?"
- The Analogy: The AI was like a construction foreman. If a student said, "This situation makes farmers poor," the AI might suggest, "Maybe say 'In the long term, this creates agricultural instability' to sound more scientific." The students didn't just copy that. They took the suggestion, thought about it, and decided, "Yes, that's better," or "No, that's too fancy, I'll keep my version." They were actively building the argument, with the AI handing them better tools.
4. The "I'm the Boss" Phase (Critical Evaluation & Agency)
This is the most important part. The students didn't blindly trust the robot. They checked everything. They knew the AI could lie, make up facts, or give generic answers.
- The Analogy: The students treated the AI like a very knowledgeable but slightly unreliable intern. They would ask the intern for ideas, but then they'd double-check the intern's work. One student noted, "AI gave me a general response and faking the data," so they went and found the real data themselves. They kept the "Authorial Agency," which is a fancy way of saying, "I am still the author." They decided what to keep, what to change, and what to throw in the trash.
What This Study Says It Is NOT
It's important to know what this study doesn't say. The researchers are very clear: This is not a story about AI writing essays for students.
- It's not a "Spoon-Feeding" Machine: The study explicitly argues against the idea that students are just using AI to get easy answers or to have the robot do the thinking for them. The students were doing the heavy lifting of thinking, reasoning, and deciding.
- It's not just about Grammar: While the AI did help fix some grammar, the study suggests that was a side effect. The main event was helping students understand complex topics and build logical arguments.
- It's not a Magic Bullet for Everyone: The study was done in one specific university in Indonesia with 26 students. The researchers admit they can't say this works exactly the same way for native English speakers or in every single school in the world. They are suggesting a possibility based on what they saw, not claiming they have solved the problem of writing forever.
The Big Picture: The "Human-AI Co-Pilot" Map
Based on these findings, the researchers proposed a new map called the Human–AI Epistemic Mediation Framework.
Think of "Epistemic" as a fancy word for "how we know what we know." This framework suggests that when a student uses AI, they aren't just typing and getting text. They are going through a process where:
- They use AI to understand the concept.
- They use AI to translate that concept across languages.
- They use AI to build the argument.
- They use their own brain to critically judge and own the final result.
The study suggests that in this setup, the AI isn't the writer; it's a partner in the thinking process. The students are the captains of the ship, and the AI is the radar and the map. The ship moves forward because the captain is steering, not because the map is driving.
The Bottom Line
The researchers found that 26 students used AI not to cheat, but to think deeper. They used it to clarify confusing ideas, mix their native language with English to find the right words, and build stronger arguments, all while keeping their own "authorial voice" intact.
The study suggests that if schools stop treating AI as a "copy-paste" tool and start treating it as a "thinking partner," students might actually learn more about how to think critically and argue their points, rather than just learning how to type faster. But remember, this is a suggestion based on one group of students in one place. It's a promising new way to look at things, but it's not a finished, proven fact for every school on Earth yet. It's a really interesting clue that the future of writing might be a team sport between humans and robots.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.