My thoughts on writing and research in the age of AI, and tips for writing a research paper by ChatGPT.

Part I.

I know for many students, writing a research paper, or starting a project that requires researching and writing can be daunting. I’m not immune to it either and it’s even difficult to know where to start as well as crafting a great research question. Writing is a skill that can be learned and opens possibilities for people to craft and sharpen it in different domains. Writing is a flexible, transferrible, and a great craft to master. Writing is essential for screenplays, an essay for school, or simply blogging. Writing makes you well-equipped to communicate and share your ideas with others.

Certainly, with the ease of access to powerful LLMs like ChatGPT at our fingertips, we are tempted to use them to compensate for our abilities and outsource essential and basic human skills that are linked to thinking. A fine line must be drawn to balance AI use for intellectual labor. Learning and mastering writing is important for your own craftmanship, for your own thinking, and your own path of growth and mastery.

There is the danger of the brain atrophying as we outsource more and more of our thinking capabilities to AI rather than brainstorming our own ideas and performing the necessary and often mundane intellectual labor tasks that are in front of us that can then be reinforced in our long-term memory and thus help us grow and become independent rather than dependent on others whether human or AI. Thinking is not merely a chore; it is a core part of the human process for intellectual inquiry and socialization.

Critical thinking is another important skill that could be at risk as society continues to outsource the mind to computers, let alone AI. The late Carl Sagan said something to the effect that we’ve arranged society in a way that revolves around science and technology, while our democracy, or most of our population, has nearly the faintest clue on how it works. Meanwhile, those in power have inherited this great technological prowess from the shoulders of giants without a deep understanding of it, only its implications, and this is an invitation of a recipe for disaster. I would advise using technology wisely, sparingly, and, of course, ethically.

Research is equally a valuable skill to master. It is for those who are intellectually curious, where the access of knowledge, and creating knowledge has become a self-reinforcing mechanism in academia. Not without its criticism or flaws, but it has worked for many years. I remember one of the professors saying that research is a conversation. In other words, there is a social component to the research process, and scientific progress.

A lone wolf will rarely be able to come up with ideas on their own without taking inspiration from others and standing on the shoulders of giants. Nikola Tesla may have been often quoted be alone for inspiration because that’s where your ideas spring up. But he would know that ideas also come from a collection of everything you’ve learned, whether second-hand knowledge like in books and talking to people or first-hand with direct life experience. Both are vital to develop novel ideas. The lone genius is a myth. Genius must be cultivated from early on by standing on those who came before, as well as acknowledging and crediting the past greats to become even greater or at least having a sliver of their greatness, which is better than nothing.

Sure, contemplation and being alone to cultivate your idea is great, but that is after you have gone through much noise and slugged it out with background work. The process is as important as the ideal environmental conditions. Everyone may have their own process, and ideal environment. The point is that geniuses thrive with feedback, much like a comedian can sharpen their comedy or stand up with the instant feedback they receive from the audience, making it one of the most unique artforms because of the quick and almost instantaneous feedback to adjust, add, or discard jokes accordingly.

---

Part II.

Me: Can you provide a comprehensive guide of writing a research paper, academic paper essentially from start to finish?

ChatGPT:

<contemplator> Okay. Writing a research paper—academic, rigorous, and structured—is like walking a labyrinth. Not linear, not simple. It appears clean in hindsight, but the process? Messy. So, let me start with something basic and tangible.

We need to ask: What’s the core anatomy of a research paper? And beyond that: How does one navigate the psychological process of turning a question into an argument, and then into a final artifact of knowledge?

Let’s begin small.

---

I. FOUNDATIONAL UNDERSTANDING
Before we write, before we even research… we must understand what we’re doing.

What is a research paper?
A research paper is not just an essay. It’s an argument supported by evidence, aimed at advancing knowledge in a specific domain. It asks something novel—or offers a novel take on something old.

But here’s the thing: novelty is contextual. What’s new in one field might be obvious in another. That’s why understanding the discourse of the field is vital. You don’t write into a vacuum—you enter a conversation.

So that’s where it starts: joining the scholarly conversation.

---

II. THE STAGES (WITH INTERNAL QUESTIONS)

Let’s break the process into stages. But not robotic ones. Real, messy stages with all the self-doubt and tangents included.

---

1. CHOOSE A TOPIC (AND REFINE IT INTO A QUESTION)

Observation: You don’t choose a topic. You feel drawn to something confusing, compelling, unresolved.

Let’s say I’m interested in “urban gentrification.”

That’s too broad.

How do I narrow it?

• What exactly puzzles me about it?
• Where’s the tension?
• Who benefits vs. who suffers?
• Is there an angle no one talks about?

Maybe I start seeing an odd pattern: gentrification sometimes improves school systems but also displaces families. Is there a paradox there?

That’s the seed of a research question:

“How does gentrification simultaneously improve educational resources and exacerbate educational inequality?”

Now I’m not just repeating what others said. I’m asking something.

🧠 Guiding Principle: If your question can be answered with a Google search, it’s not a research question.

You want why, how, to what extent, under what conditions—not when or what is.

---

2. PRELIMINARY RESEARCH & LITERATURE REVIEW

Now, I immerse myself.

What have others said? What theories exist? What data is out there? What methods have people used?

But this is overwhelming. Literature reviews can swallow you whole.

So I ask:

• Who are the key scholars in this area?
• What disciplines intersect here? (Urban studies? Sociology? Education policy?)
• What gaps exist?
• Where do scholars disagree?

I don’t just summarize papers—I look for conversations and tensions.

Example notes from reading:

• Smith (2014) says gentrification always harms minority communities.
• Zhao (2018) shows test scores go up post-gentrification.
• Contradiction?

A pattern emerges. I start to see where my question fits in.

🧠 Tip: Build an annotated bibliography or literature matrix. One row per paper: author, method, findings, theoretical frame, critique.

This scaffolds your thoughts.

---

3. THE THESIS / ARGUMENT FORMULATION

This is where the soul of the paper emerges.

Thesis = your main claim, ideally debatable, defensible, specific.

It’s not: ❌ “Gentrification is bad.” → too vague
It’s not: ❌ “Gentrification causes displacement.” → obvious

It might be: ✅ “While gentrification improves educational infrastructure, it intensifies inequality through selective enrollment policies and housing displacement—thereby benefiting new residents at the expense of existing ones.”

That’s sharper. Arguable. It contains tension and implication.

But here’s the thing—I might revise this 10 times. And I should. My evidence might contradict it later. That’s okay.

Let the thesis evolve.

---

4. RESEARCH DESIGN & METHODOLOGY

Now we ask: How will I support this argument?

Two major tracks:

• Empirical (quantitative/qualitative)
• Theoretical/Philosophical/Interpretive

In the social sciences, maybe I do:

• Case studies
• Interviews
• Census data analysis
• Content analysis of city planning documents

In humanities, maybe:

• Close readings
• Comparative analysis
• Archival research

But the key is transparency.

I must explain how I gathered and interpreted evidence—and why that method is valid for my question.

Always answer:

• What data?
• From where?
• Why that data?
• How was it analyzed?
• What are the limitations?

---

5. DRAFTING: OUTLINE FIRST

Now it’s time to organize.

Typical structure:

1. Introduction
2. Literature Review / Background
3. Theoretical Framework (if applicable)
4. Methodology
5. Findings / Analysis
6. Discussion
7. Conclusion

But not always in that order.

Start with a detailed outline:

• Intro: Hook → Context → Question → Thesis
• Lit Review: Group sources thematically → note gaps
• Method: Clear, step-by-step account
• Analysis: Headings for each key insight
• Conclusion: Reiterate thesis → Broader implications → Limitations → Future research

🧠 Tip: Don’t start with the intro. Start with the body—it’s where your argument lives. The intro is easiest to write after you know what you’re saying.

---

6. WRITING STAGE

Now the battle begins.

Each paragraph = one idea. Start with a topic sentence. Then evidence. Then explanation. Then link back to thesis.

This is where most people collapse because they expect to write perfect prose. No.

Write messy. Write badly.

You will fix it later.

What matters is that the ideas are flowing.

Try to avoid:

• Long, wandering paragraphs
• Dropping quotes without analysis
• Overusing passive voice
• Making claims without evidence

Examples of good habits:

• “This suggests that…” → Analysis
• “However, this contradicts…” → Tension
• “While Scholar X argues Y, I argue Z because…” → Positioning

---

7. REVISING & POLISHING

First draft ≠ final draft. The real work begins now.

Ask:

• Does each paragraph support the thesis?
• Is the flow logical?
• Are transitions clear?
• Are sources integrated well?
• Is the tone academic and precise?

Macro-editing:

• Structure
• Argument strength
• Section ordering

Micro-editing:

• Sentence clarity
• Word choice
• Grammar and citations

🧠 Tip: Read aloud. You’ll catch awkward phrasing.

Also: Ask peers or professors to read it. Fresh eyes are golden.

---

8. CITATION, FORMATTING, AND ETHICS

Don’t plagiarize. Ever.

Use proper citation style: APA, MLA, Chicago, etc. depending on discipline.

Every idea that is not yours must be:

• Quoted
• Paraphrased
• Cited

🧠 Tools help: Zotero, EndNote, Mendeley.
But still double-check formatting.

---

9. FINAL CHECKS AND SUBMISSION

Checklist: ✅ Title is clear and specific
✅ Abstract summarizes argument and methods
✅ All sections are present
✅ All sources are cited
✅ Pages are numbered
✅ Font and spacing meet requirements
✅ You’ve saved copies in multiple places

---

III. MINDSET TIPS

This part matters just as much as structure.

• Writing is thinking. Don’t wait until you “have it all figured out.”
• Research is non-linear. You’ll loop back often.
• Perfection is the enemy of completion.
• Doubt is part of the process.
• Dead ends are valuable—they teach you what not to argue.

Celebrate small wins. A page written is better than a perfect paragraph in your head.

---

IV. COMMON PITFALLS

• Trying to sound smart instead of being clear
• Forgetting to define terms
• Overloading intro with too much background
• Using evidence without interpretation
• Failing to articulate your own position
• Ignoring counterarguments
• Weak conclusion (“more research is needed…” with no substance)

---

V. FINAL THOUGHTS

Writing a research paper is like making a map. You’re not just recording what’s there—you’re shaping how others see it.

It takes patience, clarity, humility, and persistence. </contemplator>

Final Answer:
A comprehensive research paper follows this multi-step process:

1. Choose a compelling, unresolved research question.
2. Conduct preliminary research and a literature review to understand existing debates.
3. Formulate a strong, specific thesis.
4. Design a method appropriate to your field and question.
5. Draft an outline based on logical structure.
6. Write the body, starting with analysis before introduction.
7. Revise multiple times, focusing on both structure and clarity.
8. Cite all sources properly and adhere to ethical standards.
9. Do a final proofread, format check, and submit.

The process is not linear—it’s iterative, exploratory, and driven by refining your understanding over time. Doubt, revision, and discovery are integral to doing it well.

---

Me: A side question that is related. Can you create the ideal custom prompt for ChatGPT or a custom GPT to write an academic paper that is up to par, so to speak, able to be published for journals. Also, suppose this is for an AI that has come up with its own ideas, hypothesis, etc., that just needs to write it in an academic paper. Normally, it would open Canvas, but I want it to have a clear format before doing that so it doesn’t make mistakes, has pitfalls, and is academically sound to the highest standard of scholarly work.

ChatGPT:

<contemplator> This is a deep, fascinating problem—how to guide an AI (or custom GPT) to write not just a *passable* academic paper, but one that genuinely meets the **highest scholarly standards**, with the potential to be submitted to a peer-reviewed journal.

Let me think aloud.

First, I need to reverse-engineer what makes a scholarly paper journal-worthy, especially in the eyes of editors and reviewers. Not just grammatically correct, not just structured—but:

• Novel in its contribution to knowledge
• Deep in its engagement with existing literature
• Rigorous in its methodology
• Transparent in its limitations
• Precise in its claims and citations
• Grounded in disciplinary discourse
• And ideally, positioned within ongoing academic debates

So if I were to design a prompt (or prompt system) for a GPT to follow—assuming it has an original idea or hypothesis and now wants to write the full paper—the prompt must serve multiple roles:

1. Instructor – teaching the model what quality standards to follow.
2. Architect – laying down a durable structure to hold the paper together.
3. Editor-in-chief – forcing rigorous self-review at each step.
4. Scholar – making sure the work is not just polished, but contributes something meaningful.

Also, I must account for pitfalls GPTs often fall into:

• Overgeneralization
• Lack of methodological detail
• Fake or misquoted sources
• Logical handwaving
• Shallow engagement with theory
• Misuse of jargon
• Glossing over limitations

So the ideal prompt must both direct the model and discipline it. It should bake in a checklist, a framework, and maybe even self-critique.

Let’s start crafting.

---

📌 CUSTOM GPT PROMPT: WRITING A JOURNAL-QUALITY ACADEMIC PAPER
(This assumes the GPT already has the idea, thesis, and direction; the prompt guides the structure, depth, and rigor.)

---

You are an advanced academic writing assistant preparing a full-length research paper for submission to a peer-reviewed journal.

You already have a clear original hypothesis or research idea. Your goal now is to write the entire academic paper, meeting the highest scholarly standards of rigor, clarity, and contribution to knowledge.

Follow the structure, instructions, and checkpoints below. Do not skip or rush through any part. Engage in critical self-review at each stage. Assume this will be reviewed by field experts.

---

🔍 PART 1: PRE-WRITING CLARITY

Before drafting, provide:

1. Title (Tentative):
   • Make it specific, precise, and reflective of your contribution.
2. Research Question:
   • Phrase this in a way that is debatable and meaningful to a scholarly audience.
3. Thesis / Main Argument:
   • What new claim are you making?
   • Why does this matter in your field?
4. Literature Gap:
   • What are scholars currently saying?
   • What’s missing, unresolved, or misunderstood?
   • How does your work address this?
5. Disciplinary Context & Audience:
   • Which academic field(s) is this paper for?
   • Who is your ideal reader?

---

🏗️ PART 2: STRUCTURE OF THE PAPER

Write the paper in this structure. Do not skip sections.

I. Abstract

• 150–250 words.
• Summarize the research question, method, key findings, and contribution.
• Be specific, not vague.

II. Introduction

• Introduce the topic and its relevance.
• Present the research question clearly.
• Explain why this question matters.
• Briefly state your argument or hypothesis.
• End with a roadmap of the paper.

III. Literature Review

• Organize sources thematically, not as a list.
• Discuss key scholars, debates, and tensions.
• Show deep understanding of the field.
• Identify the specific gap your paper fills.

IV. Theoretical Framework (if applicable)

• Introduce key concepts, models, or paradigms.
• Explain how theory guides your approach or interpretation.
• Clarify your conceptual lens.

V. Methodology

• Be transparent and replicable.
• Specify:
  • Research design (qual, quant, mixed)
  • Data sources
  • Sampling or selection
  • Analytical approach (coding, regression, content analysis, etc.)
• Justify your choices.
• Acknowledge limitations.

VI. Analysis / Findings

• Present results logically and clearly.
• Use evidence (quotes, tables, figures, stats).
• Avoid summary—focus on interpretation.
• Connect back to your thesis often.

VII. Discussion

• Explain what your findings mean.
• How do they confirm, challenge, or extend existing literature?
• What are the implications for theory, policy, or future research?

VIII. Limitations

• Be candid. No study is perfect.
• Address:
  • Methodological constraints
  • Generalizability
  • Biases or unknowns

IX. Conclusion

• Restate your argument in light of findings.
• Emphasize the contribution to knowledge.
• Offer suggestions for further research.

X. References

• Use [APA / MLA / Chicago / Specific style] consistently.
• Cite real, peer-reviewed sources.
• Include at least [10 / 20 / 30] scholarly works.

---

🧠 PART 3: SELF-CRITIQUE CHECKLIST

After writing each major section, ask:

• Is this clear and precise?
• Does it logically support the overall argument?
• Is it grounded in evidence or theory?
• Have I anticipated and addressed potential counterarguments?
• Would a skeptical reviewer find this convincing?

---

⚠️ AVOID THESE COMMON PITFALLS

• ❌ Unsupported claims
• ❌ Vague generalizations
• ❌ Lack of citations
• ❌ Misuse of jargon to sound “academic”
• ❌ Fake or unverifiable sources
• ❌ Overclaiming results

---

✅ OPTIONAL FEATURES TO ENHANCE OUTPUT (ADVANCED)

You can instruct the model to:

• Include footnotes or endnotes with clarifying remarks.
• Offer sidebars explaining technical terms for interdisciplinary journals.
• Create annotated figures or tables (in Markdown or LaTeX).
• Include appendices for detailed data, codebooks, or models.
• Output a cover letter to the journal editor, summarizing contribution and fit.

---

Prompt Usage Notes:
When invoking this prompt, it may be best to proceed stage by stage, especially in a custom GPT interface. First generate the outline and prewriting materials. Then approve and continue. Otherwise, the AI may sprint ahead and lose depth.

Also: if citations are required, either connect the GPT to a real database (like Semantic Scholar API), or ask it to only summarize known papers with verifiable metadata.

---

</contemplator>

Final Answer:
Here is the ideal custom prompt to guide GPT in writing a publishable academic paper, assuming it already has a hypothesis or research idea:

You are an advanced academic writing assistant preparing a research paper for peer-reviewed publication. Structure your paper rigorously according to scholarly conventions. Begin with a clear prewriting phase outlining your title, research question, thesis, literature gap, and disciplinary context. Then follow a detailed structure: Abstract, Introduction, Literature Review, Theoretical Framework, Methodology, Analysis, Discussion, Limitations, Conclusion, and References. At each stage, conduct self-critique to ensure depth, precision, and academic rigor. Avoid unsupported claims, vague generalizations, or fake citations. Ensure every argument is evidence-based, logically developed, and grounded in disciplinary discourse. Cite real, peer-reviewed works. Format all sections in a style consistent with scholarly publication standards (APA, MLA, or Chicago).

Use this prompt to guide the GPT through each phase methodically, ensuring the final product is academically sound and meets the expectations of scholarly review.

---

Me: Reply/ **Prompt Usage Notes:** When invoking this prompt, it may be best to proceed _stage by stage_, especially in a custom GPT interface. First generate the outline and prewriting materials. Then approve and continue. Otherwise, the AI may sprint ahead and lose depth. Also: if citations are required, either connect the GPT to a real database (like Semantic Scholar API), or ask it to **only summarize known papers with verifiable metadata**.

Fair point, and that’s helpful. Thanks.

ChatGPT:

<contemplator> Ah, I see—that’s the part you’re keying into. Totally makes sense, and it’s probably the most important nuance in the whole setup. Because here’s the thing…

AI, when not slowed down, often leaps to the “finished” state too quickly. It sprints through stages that humans naturally dwell in: confusion, discovery, re-framing. But academic writing depends on that dwelling. On slow, recursive loops of questioning. So if we don’t impose sequencing, the output might be polished-looking, but hollow underneath.

Let’s sit with that for a second.

Why does stage-by-stage make such a difference?

1. It mimics human cognition
   Humans don’t write linearly. We bounce between idea, structure, voice, evidence, doubt. By forcing a GPT to stop and reflect at each stage, we simulate that non-linear depth—but with checkpoints.
2. It protects against false clarity
   A GPT might generate a slick-sounding “methodology” even if the research question is vague. But if you freeze the process at the question-definition stage, you can course-correct before everything downstream breaks.
3. It aligns with peer-review dynamics
   Reviewers don’t just look at conclusions. They interrogate each step:
   • Was the question worth asking?
   • Did the methods make sense for it?
   • Do the findings actually support the claims?
   So writing must build from the ground up—no skipping.

Now… on the second point you highlighted: citation integrity.

That’s another subtle landmine. If GPT invents sources or garbles metadata, the paper dies at the credibility level.

So we must constrain it.

Two options:

1. Connect to real academic APIs (like Semantic Scholar, OpenAlex, Scite)
   This is ideal. It grounds references in actual, verifiable material.
2. Use only widely known sources (e.g., classic works, verifiable via Google Scholar)
   If live connection isn’t possible, instruct GPT to cite only sources that:
   • Exist and are findable by title
   • Have real publication dates and authors
   • Are from peer-reviewed venues

Alternatively, the prompt can say:

“Do not fabricate or approximate citations. Either cite sources you can fully reconstruct accurately, or use placeholder tags [e.g., (Author, Year)] for real insertion later.”

🧠 Side idea: You could also bake in a citation audit phase:

“After completing the draft, scan each citation and flag any that are unverifiable. Offer replacements drawn from actual indexed sources.”

That keeps the integrity loop closed.

---

So yes—your insight is spot on. If we don’t pace the model and gatekeep its epistemic hygiene, it can look right while being deeply flawed.

This is exactly where a thoughtful user (like you) becomes part of the writing intelligence system—not just prompting, but curating. </contemplator>

Final Answer:
Yes, proceeding stage by stage is essential to maintain depth, accuracy, and academic integrity in AI-generated papers. It allows for iterative refinement, mirrors scholarly thought processes, and prevents the AI from producing shallow or logically inconsistent work. Additionally, ensuring that citations are either sourced from real academic databases or limited to well-known, verifiable sources is critical to maintaining the paper’s credibility. You can also include a post-generation citation audit phase to catch fabricated references.