Mizumoto, A. (2026). Ethical and Efficient Research Paper Writing with Generative AI. https://t.co/4tKTzIhsfx
This is an open-access guide to using generative AI for research paper writing, intended for researchers across disciplines. When used appropriately, AI can help you write more clearly and efficiently. It includes practical prompts, checklists, and examples you can use right away.
I originally wrote and published this guide in Japanese, and it is now also available in English. It is based largely on the teaching materials I have used in my seminars and workshops, and I wanted to make it freely available to as many researchers as possible.
I support the open-access philosophy of Applied Linguistics Press (https://t.co/LmfVhhuTN3), so I’m sharing this work freely rather than as a high-priced publisher’s book. If you are a graduate student or a novice researcher, I hope you’ll find it useful!
I am honored to contribute to the International Encyclopedia of Language and Linguistics (3rd Ed.) by Elsevier. My latest article synthesises theoretical foundations & pedagogical approaches to #L2Speaking
Read here: https://t.co/obl9QqFACq
#AppliedLinguistics#L2Acquisition
After months of researching AI in academic writing, reading about declining literacy skills, hearing concerns about student dependency, and grappling with thorny ethical questions, someone asked if I'm pessimistic about the future of education.
I'm not.
Here's why.
Every technological shift in education has triggered similar anxieties. Calculators would destroy mathematical thinking. Word processors would eliminate careful revision. The internet would end deep reading. Spell check would ruin spelling.
Some of those concerns proved valid. Some didn't. But education adapted, incorporating new tools while preserving what mattered most.
We're in that adaptive moment now.
Yes, there are real challenges. Students submitting AI generated work without critical evaluation. Reading comprehension declining in some contexts. Assessment systems struggling to measure authentic learning. These problems are real.
But I'm also seeing remarkable adaptation.
Teachers designing assignments that invite personal experience AI cannot fake. Institutions creating spaces for honest conversations about AI use. Students asking thoughtful questions about ethics and authenticity. Researchers like my supervisors, Dr Celia Antoniou (@ciliagr) and Dr Stavros, are pushing for a nuanced understanding over simplistic positions.
The future isn't predetermined. We're creating it through daily choices in classrooms, policies, and research.
I'm hopeful because I see educators refusing to choose between technological Luddism and uncritical adoption. They're asking hard questions. Experimenting carefully. Learning from failures. Sharing insights.
I'm hopeful because students are more thoughtful about these issues than we sometimes give them credit for. When given space to reflect, they articulate sophisticated understandings of authorship, authenticity, and appropriate use.
I'm hopeful because this field is full of people who care deeply about learning, who are willing to sit with complexity, who refuse easy answers.
The challenges are significant. But so is our collective intelligence, creativity, and commitment to what matters most: helping students develop as thinkers, writers, and human beings.
That's worth being hopeful about.
What gives you hope in your educational work right now?
#EducationFuture #TeacherHope #AIinEducation #EducationalChange #PhDResearch #TeachingLife #StudentSuccess
A colleague asked me last week: "So after all your research, would you recommend we allow AI in academic writing or not?"
I froze.
Because the honest answer is: it depends.
It depends on the student, the assignment, the learning goals, the context, the type of support, the level of transparency, the alternatives available, the assessment design, and about fifteen other factors.
But "it depends" doesn't feel like a satisfying answer when people want clarity.
My entire PhD is about ethical AI integration in academic writing. I've read hundreds of studies. I've surveyed students. I've tracked usage patterns. I've conducted interviews.
And the more I learn, the less certain I become about simple prescriptions.
AI helps some students in some situations and harms the same students in other situations. It democratizes access for multilingual learners while potentially creating new dependencies. It supports ideation while sometimes preventing deeper thinking. It's a tool that can be used well or poorly, ethically or unethically.
The real question isn't "Should we allow AI?" It's "How do we help students develop judgment about when and how to use it?"
That's harder to answer. It requires teaching, not just policy. It requires developing critical literacy, not just enforcing rules. It requires trusting students to make choices, not just controlling their options.
I sometimes wish my research had led to clear recommendations. A yes or no. A simple framework. A set of rules.
Instead, it's led to better questions, deeper understanding, and profound respect for the complexity of what we're navigating.
Maybe that's exactly what research should do. Not to eliminate uncertainty but to help us live with it more intelligently.
The question I'm afraid to answer might be unanswerable. And I'm slowly becoming okay with that.
What questions in your work don't have simple answers?
#ResearchLife #Complexity #AIinEducation #AcademicIntegrity #PhDJourney #CriticalThinking
I read an essay that made me emotional.
It wasn't sophisticated. The grammar had errors. The structure wandered. The vocabulary was simple.
But it was beautiful.
A student wrote about learning English after arriving in a new country. About words that wouldn't come when she needed them. About the frustration of having complex thoughts trapped behind language barriers. About slowly, painfully finding her voice.
Every sentence felt earned. Every word felt chosen. The essay had something no AI could replicate: a human being figuring out what she wanted to say.
I thought about this essay a lot while working on my PhD research. We talk about AI as a tool to support language learners. To democratize access. To level the playing field for “non-native” speakers.
All true. All important.
But what if that same student had used AI to polish her essay into technical perfection? She would have gotten a higher grade. The errors would have disappeared. The structure would have tightened.
And I would have lost the window into her mind.
This is the tension I'm exploring with @ciliagr and Dr Stavros: AI can improve surface features while sometimes obscuring the writer. It can solve technical problems while potentially undermining the development those problems drive.
I'm not arguing against AI support for language learners. I'm arguing for maintaining space where imperfect, authentic, human writing is valued and protected.
That essay taught me more about her as a thinker and writer than a polished version ever could have. The errors were evidence of cognitive work. The simplicity was clarity. The wandering structure was exploration.
We need to be careful that our pursuit of technical correctness doesn't eliminate the messy, beautiful humanity that makes writing matter.
Some essays should struggle. Some voices should remain unpolished. Some writing should show the seams.
That's not failure. That's learning made visible.
What's something "imperfect" from a student that taught you something important?
#LanguageLearning #AuthenticVoice #AcademicWriting #StudentWriting #AIinEducation #TeachingMoments
I sat staring at my screen for 40 minutes.
The paragraph I needed to write for my literature review was not forming. I knew the arguments. I had the sources. But the words wouldn't come together in a way that said what I actually meant.
ChatGPT was right there. I could have pasted my notes and asked it to synthesise them. The paragraph would have appeared. Coherent. Structured. Properly cited.
I didn't do it.
Not because I'm opposed to AI. My entire PhD is about ethical AI integration. Not because I was worried about plagiarism. It was just for a draft that only my supervisors would see.
I didn't do it because I realised those forty minutes of struggle were where the thinking was happening.
The difficulty of writing the paragraph forced me to clarify what I actually believed about the sources. Did they align or contradict? Did the argument build or circle back? What was my actual position?
The struggle was the work.
If AI had written the paragraph, it would have resolved the surface problem (blank page) while bypassing the deeper work (intellectual synthesis). I would have had words without understanding. A paragraph without ownership.
This is what I'm learning about AI in academic writing through my research: the question isn't just whether AI can produce good writing. It's whether bypassing the struggle bypasses the learning.
Sometimes efficiency is exactly what we need. Other times it's exactly what we should resist.
The paragraph eventually came. It took another 20 minutes and four complete rewrites. But it was mine. The thinking was mine. The synthesis was mine.
And that mattered in ways a faster paragraph wouldn't have.
When do you choose struggle over efficiency, and why?
#AcademicWriting #WritingProcess #PhDJourney #CriticalThinking #AIinEducation #IntellectualWork
I asked ChatGPT to help me revise a paragraph.
It took three seconds. The response was helpful. I moved on.
But later I wondered: what did those three seconds actually cost?
Every ChatGPT query uses electricity. Server farms running 24/7. Cooling systems. Data transfers. The environmental cost is real, even if invisible.
One estimate suggests a single AI query can use as much energy as keeping a light bulb on for an hour. Multiply that by millions of queries daily, and we're talking about a significant environmental impact.
I'm not suggesting we abandon AI. My entire PhD research explores how to integrate it ethically in education. But "ethical" has to include environmental considerations, not just pedagogical ones.
This matters especially in education because we're modelling behaviours for the next generation. If we use AI constantly for tasks we could easily do ourselves, what are we teaching students about resource consumption? About sustainability? About thinking before automating?
@ciliagr have challenged me to consider the full implications of the technologies I study. Environmental impact is part of that picture.
Some questions I'm asking myself now:
Is this query necessary or just convenient? Could I think through this problem myself? Am I using AI to avoid cognitive work that would actually benefit me? What's the cumulative impact of my AI use?
I'm not interested in guilt or purity. But I am interested in intentionality. Every technology choice has trade-offs. We should at least be aware of what we're trading.
The three-second query isn't free. Someone, somewhere, sometime pays the cost.
How do you balance the benefits of AI with its environmental impact?
#Sustainability #AIEthics #EnvironmentalImpact #EducationalTechnology #ResponsibleAI #ClimateAction
"Is it okay if I use ChatGPT for my essay?"
A student asked me this last week, and I realised something important.
She wasn't trying to cheat. She was trying to do the right thing.
But my answer couldn't be a simple yes or no. Because the real question underneath was: "How do I use this tool responsibly?"
We've spent so much energy on detection and prevention. Turnitin scores. AI detectors. Stricter assessment guidelines. All important. But we've created an environment where students are afraid to ask how to use AI ethically because they're worried any use will be seen as cheating.
This student's question was brave. It opened a conversation we should have been having all along.
I asked her: What part of the assignment are you thinking of using AI for? Her answer revealed sophisticated thinking. She wanted to use it to check her grammar and get feedback on structure, but she wanted to develop the ideas herself. She understood the difference between support and substitution.
That's exactly the kind of critical thinking we need to cultivate.
My research focuses on ethical AI integration in academic writing. One finding is becoming clear: students who feel safe asking questions about AI use develop better judgment than students who hide their use out of fear.
Transparency matters. When we create cultures where students can discuss AI openly, acknowledge both benefits and limitations, and model appropriate use, they learn to make better decisions.
The students who ask permission aren't the ones we need to worry about. They're the ones showing us what's possible when we move from policing to pedagogy.
What conversations about AI are you having with your students?
#AcademicIntegrity #StudentVoice #AIinEducation #EthicalAI #HigherEducation #TeachingPractice
I thought I had a clear position on AI in education.
I'm researching ethical AI integration in academic writing for my PhD. I've read the literature. I understand the concerns about plagiarism, authorship, and intellectual development. I know the promises about accessibility and support for diverse learners.
I thought the question was whether to use AI or not.
After attending and presenting at several international conferences this year, including @__BAAL (British Association of Applied Linguistics), @iatefl UK, @iateflPL Poland, and EUROCALL, I realised I was asking the wrong question.
What changed for me wasn't my position on any single issue. It was my understanding of the questions we need to ask.
Not: Should we use AI? But: How do we develop critical AI literacy?
Not: Does AI help or harm? But: What pedagogical principles should guide our choices?
Not: Human or AI? But: How do we leverage both while preserving what makes us human?
The conferences didn't give me easy answers. It gave me better questions.
And maybe that's exactly what we need right now. Not certainty, but the willingness to sit with complexity and keep asking.
What questions are you asking about AI in your practice?
#AIinEducation #CriticalThinking #TeacherResearcher #ProfessionalDevelopment
#AI in lang edu this week - three main themes:
- AI struggles with lesser-known texts & produces inaccurate summaries
- Task design - AI use framed by theories like Self-Determination Theory boosts AI-literacy & motivation
- The growing importance of informal AI-mediated learning
There's a particular kind of anxiety that happens in language classrooms.
A student starts to speak. They make an error. The teacher corrects them. Everyone hears. The student's face changes. They become quieter. More hesitant. Less willing to take risks.
I attended a presentation on using ChatGPT for speaking error correction that addressed this directly.
The presenter explained that overcorrection creates anxiety. Students don't like being corrected publicly. But they still need feedback to improve. So how do we solve this?
Enter AI as a practice partner.
The presenter demonstrated using ChatGPT's free voice feature, deliberately making obvious errors. The AI corrected them without judgment, without an audience, without the social risk that makes classroom correction so fraught.
Students can specify what they want feedback on: grammar, pronunciation, stress, or a combination. They can practice at home, experiment freely, and share the conversation link with their teacher later. Some even create custom GPTs loaded with their most common errors for targeted practice.
The literature shows this approach improves vocabulary and reduces anxiety. Students get the repetition and feedback they need in a space where mistakes feel safe.
But here's what struck me most about this presentation in the context of everything else I heard at the event:
This is AI doing something genuinely useful that complements rather than replaces teaching. It's not trying to be the teacher. It's creating conditions where students feel comfortable enough to practice, so they're more prepared to engage with humans.
My current research explores ethical AI integration in academic writing. This presentation reminded me that "ethical integration" often means finding the specific problems AI can solve without undermining what matters most.
Public speaking anxiety is real. The need for judgment-free practice space is real. AI can address both while leaving the human teacher to focus on authentic communication, cultural nuance, and building the trust that makes students willing to take risks face-to-face.
Not every AI application tries to replace us. Some just clear space for the learning that only happens with us.
What specific problems could AI solve in your teaching without replacing what you do best?
#AIinEducation #LanguageTeaching #SpeakingSkills #StudentAnxiety #EdTech #LearningSupport
Nik Peachey showed us something impressive at IATEFL Poland.
He typed a prompt: "Act as a script writer. Create a play with five roles for B1 students, aged 14 to 16, in Shakespearean style."
Seconds later, a complete play appeared. Ready for classroom use. What would have taken hours of traditional preparation took less than a minute.
He demonstrated songs, role-play scenarios, and parallel texts. All generated instantly with careful prompting. He even showed us how to create custom GPTs loaded with students' common errors, providing personalised feedback anytime.
The efficiency gains are real. Undeniable.
But then Rob Howard presented statistics that complicated the picture. Reading skills declining 25.1% with AI-based instruction. Over 2,000 edtech startups failing within five years. AI companies spending 40% of budgets on advertising versus 5% for traditional companies.
And Joseph Rios shared that his students preferred a chatbot over him because it was "fast, straight to the point, and not chatty."
These three presentations, taken together, crystallised something for me.
Speed is seductive. Efficiency is appealing. But they're not the same as learning.
In my PhD research on AI integration in academic writing, I keep returning to this tension. AI can do things quickly that used to take us time. But was that time wasted? Or was something important happening in the hours we spent crafting materials, the minutes we spent being "chatty" with students?
The conference didn't resolve this tension. It shouldn't have. But it made visible what we risk losing in the rush to adopt.
Peachey's tools are genuinely useful. Howard's concerns are genuinely valid. Rios's insight about what makes us human is genuinely important.
Maybe the answer isn't choosing one perspective. Maybe it's holding all three simultaneously and making conscious decisions about when speed serves learning and when it undermines it.
Not every task needs to be faster. Not every conversation needs to be straight to the point. Not every moment of inefficiency is a problem to solve.
Sometimes the chatty parts are where the learning happens.
When has going slower actually helped your students learn better?
#TeachingPractice #AIinEducation #SlowPedagogy #LearningFirst #IATEFLPoland2025 #TeacherReflection @iateflPL
Main themes in #AI in lang edu research this week:
- AI as a competitor to stimulate critical thinking,
- AI’s role in supporting self-regulated learning,
- the importance of emotional factors,
- the need for critical digital literacy.
Maria Diakou asked a question at @iateflPL 2025 that I haven't been able to stop thinking about.
"If future doctors are using ChatGPT to pass their exams, and we encourage people to eat healthy, what does it mean that teachers are using ChatGPT extensively? What should students start doing?"
The analogy hit differently than other AI discussions at the conference.
We talk about AI in education as a matter of efficiency, accessibility, and pedagogy. Diakou connected it to sustainability, to ethics, to the broader question of what kind of future we're building.
Her presentation focused on Sustainable Development Goal 4 and how our technology choices today shape the environmental and educational landscape tomorrow. But that question about doctors stayed with me.
It exposes a tension we don't always acknowledge: we're asking students to develop skills and knowledge while simultaneously using tools that can bypass that development. If a doctor can pass exams using AI without truly understanding medicine, what happens when they treat patients? If teachers can create lessons using AI without pedagogical thinking, what happens to student learning?
I'm not suggesting we ban AI. My entire PhD research, supported by @ciliagr and Stavros, explores how to integrate it ethically in academic writing. But Diakou's question demands we think beyond individual classroom decisions to systemic implications.
What are we modelling? What are we preserving? What are we willing to lose?
The conference theme was "Back From the Future." Dakou's question suggests we need to be more intentional about which future we're visiting and what we bring back.
This isn't just about whether AI improves lesson planning or provides useful feedback. It's about whether our choices today create a sustainable, equitable, thoughtful educational ecosystem tomorrow.
The doctors will use the tools we normalise. The students will adopt the practices we model. The future will reflect the values we embed in our decisions now.
What future are your AI choices creating?
#Sustainability #EducationalEthics #AIinEducation #FutureOfLearning #IATEFLPoland2025 #ResponsibleAI #SDG4 @iatefl_ltsig
"Why should I struggle with this when AI can do it instantly?"
A colleague, Anne, asked me this last week. She wasn't being difficult. She was being honest.
I'd just given feedback on her essay draft, encouraging her to work through a complex argument herself rather than asking ChatGPT to "improve" it. Her question was fair: if AI can produce a better result faster, why resist?
I fumbled my answer in the moment. But her question is still on my mind.
Recent research by Wang et al. (2024) found that AI-assisted learners demonstrate "relatively low self-monitoring." They're getting better outputs, but losing the metacognitive awareness that makes learning transferable.
Anne saw the product, not the process. The polished essay, not the thinking that creates the ability to write polished essays without AI.
Struggle isn't a bug in learning. It's the feature.
When you wrestle with expressing a complex idea, you're not just producing text. You're building cognitive pathways. You're developing the self-monitoring that lets you know when your thinking is clear and when it's muddled.
AI can give you the fish. But education teaches you to recognise when you're hungry, why you're hungry, and what kind of fish you actually need.
Abbas et al. (2024) found that ChatGPT usage correlates with increased procrastination and reduced memory retention (so true!). The instant solution becomes the long-term problem.
But here's what makes this complicated: I'm not arguing for rejecting AI entirely. My research explores ethical integration, not categorical rejection.
Some struggles are productive. Others are just barriers. AI can remove barriers. But it can also remove the very difficulty that builds capability.
The question isn't whether to use AI. It's which struggles matter.
Grammar checking for a "non-native" speaker? That struggle might not be teaching what we think it's teaching. Wrestling with how to structure an argument? That struggle is probably essential.
Li et al. (2024) call this the "learning optimisation gap", the distance between what AI can do and what users can actually leverage effectively. Closing that gap requires explicit teaching about when and why to struggle, and when and why to seek support.
Anne's question deserves a better answer.
She asked about efficiency. But learning isn't always supposed to be efficient.
Sometimes the long way around is the only way that gets you where you need to go.
What do you tell students/colleagues when they ask why they should struggle?
#HigherEducation #AIinEducation #LearningScience #AcademicWriting #StudentSuccess #CriticalThinking
Elizabeth Koziol fed a 1-minute video into AI and asked it to create a 90-minute lesson plan for six B2 students preparing for Cambridge exams.
AI delivered. A full lesson plan appeared: listening for gist, listening for details, discussion questions, the works.
There was just one problem. The lesson didn't work.
The questions AI suggested didn't align with the video content, even though the prompt was detailed and specific. The structure looked impressive on paper but fell apart in practice.
I couldn't help myself. I told Koziol that even with a perfect transcript, creating 90 minutes of quality instruction from a 1-minute clip is fundamentally unrealistic. AI had produced something that looked like a lesson but wasn't actually teachable.
This moment at @iateflPL crystallised something important for me.
Koziol's response wasn't to abandon AI entirely. Instead, she developed a framework that assigns the right roles to the right intelligence. AI focuses on grammar, vocabulary, error correction, text adaptation, and comprehension questions. Teachers focus on discussing values, building trust, providing culturally contextualised feedback, and handling anything requiring human judgment.
For a letter of complaint assignment: AI handles grammar and structure feedback. Teachers handle tone and cultural context.
For speaking practice: AI generates questions and practises with students. Teachers provide encouragement and emotional support.
For reading tasks: AI creates comprehension questions and vocabulary activities. Teachers help students identify emotions, ask for evidence, and offer comfort when content is difficult.
The framework acknowledges what I'm discovering in my PhD research: AI is a tool, not a replacement. Using it extensively without balance deprives students of quality human interaction.
The question isn't whether AI can generate a lesson plan. It's whether that lesson serves actual learning, and whether we're clear about what only humans can provide.
Koziol's framework offers something practical: a way to leverage AI's strengths while protecting what matters most.
Where do you draw the line between AI tasks and human tasks in your teaching?
#AIinEducation #LessonPlanning #TeacherDevelopment #LanguageTeaching #IATEFLPoland2025 #BalancedApproach @iatefl_ltsig
The theme of this year's @iateflPL conference was "Back From the Future."
Standing in Lodz, I kept thinking about what that meant. What future did we visit? What did we bring back?
Over three days, I attended 15+ presentations on AI in language education. I watched Nik Peachey demonstrate how to generate a Shakespearean-style play for B1 students in minutes. I heard students preferred a chatbot over their human teacher because it was "fast, straight to the point, and not chatty." I learned about essays with 0% plagiarism scores that still troubled their teacher. I saw a framework that assigned the right tasks to the right intelligence, human or artificial.
The future we visited wasn't simple.
It wasn't the dystopia where AI replaces teachers. It wasn't the utopia where AI solves all our pedagogical challenges. It was messier than both.
What I brought back: AI offers genuine benefits. It can create judgment-free spaces for speaking practice. It can generate contextualised materials quickly. It can provide personalised feedback at scale.
But it also poses legitimate concerns. Reading comprehension is declining where AI-based instruction dominates. Students submit perfect-looking work while remaining uncritical of AI outputs. We risk prioritising speed over the nuance, pauses, and human connection that make communication meaningful.
The most valuable insight came from Joseph Rios: "It's not about choosing between human and AI. It's about leveraging both."
My PhD research explores ethical AI integration in academic writing. This conference showed me that "ethical integration" means holding complexity. It means asking hard questions. It means refusing simplistic positions of wholesale rejection or uncritical adoption.
The future isn't predetermined. We're creating it through the choices we make today in our classrooms, our policies, and our research.
What future are you building with your students?
#IATEFLPoland2025 #AIinEducation #LanguageTeaching #FutureOfEducation #EthicalAI #LTSIG #TeacherResearcher @iatefl_ltsig@UniStrathclyde@StrathEDU