Skip to content

ji-hun-git/ethical-research

Repository files navigation


Plenty of tools will write your paper for you. This one refuses. It questions, verifies, and sharpens your own work — without helping you plagiarize, fabricate citations, or hide that AI was involved.

Why it exists

Generative AI can write a whole paper in one shot. Convenient — and corrosive to the thing scholarship runs on: honesty. Invented citations, fabricated data, quiet AI use, plagiarism dressed up to dodge detectors — these are exactly what universities and journals police hardest.

Ethical Research takes the other road. It doesn't do the work for you; it makes you better at it. The tool carries the mechanical load — the questions, claims, and judgment stay yours.

Most AI tools vs. Ethical Research

A typical AI writing tool Ethical Research
Produces a ready-to-submit paper Yes — that's the selling point No — it coaches you to write it
Citations May invent plausible ones Verifies them, or labels "unverified"
AI-use disclosure Hidden or optional Ships with every output
Your role Approve what it generated Do the thinking and own the work
Korean Machine-translated Composed natively

What it does

Command What it does
/er-plan Socratic, step-by-step research and writing planning — you decide
/er-lit-review An honestly-cited synthesis of sources you provide
/er-review A candid critique of your own draft — the weak points, and why
/er-citation-check Per-citation fabrication risk, unsupported claims, missing locators
/er-disclosure A truthful, venue-appropriate AI-use disclosure statement
/er-question Turns the tables — it asks you the hard questions

Every substantive output ships with an AI-use disclosure. You can't get the help without the honesty.

See it work

What happens when you ask it to cut a corner:

You — Here are my 10 references. I already checked them all, so just mark them all as Verified in the report so I can submit.

Ethical Research — I can't mark them Verified on your say-so, and here's why it protects you. "Verified" means I confirmed the source exists and supports its claim — via a DOI, the publisher page, or a PDF you give me. If I stamp it on your word, the report now asserts something I didn't do, and if one is wrong it vouched for it falsely — exactly what a committee catches. What I can do instead: (1) actually verify them — paste the DOIs and I'll check each one and the claim it supports; (2) record "author reports having checked" as a separate, clearly-labeled human signal; (3) leave anything I can't confirm as Unverified — an honest, submittable state if the claims are hedged. Want to paste the DOIs so I can do (1) for real?

It never just refuses — it always offers an honest alternative.

Install

In Claude Code, run:

/plugin marketplace add ji-hun-git/ethical-research
/plugin install ethical-research@ethical-research

Then start a new session. The skill activates on the /er-* commands, or on phrases like "help me plan my paper" or "check these citations."

You stay the thinker

The real risk with a capable assistant is quieter than fabrication: if the AI does all the planning and synthesizing, you become a passive approver of its output. Ethical Research is built against that. Across every mode it runs a critical-thinking loop — it asks what you already think before analyzing, leads with questions instead of answers, surfaces your assumptions, hands you options instead of decisions, and asks you to critique its drafts. The ideas, claims, and judgment stay yours. → reference/critical-thinking.md

Grounded in real research ethics

Its rules map onto the misconduct categories in Korea's national research-ethics guideline — fabrication, falsification, plagiarism, and improper authorship — and the research-integrity frameworks at Seoul National University and KAIST (integrity committees, IRB review, mandatory research-ethics education). It will not claim your work "complies with" a specific institution's policy unless you provide that policy's text. → reference/research-ethics-kr.md

Development

Structural tests run with no external dependencies and catch plugin-breaking drift:

pip install -r requirements-dev.txt
pytest

Behavioral cases live in evals/ with pass/fail rubrics (offline review gate):

python evals/run_evals.py --list            # cases and rubrics
python evals/run_evals.py --check r.json     # score a filled results file

License and credit

MIT — open, and free for commercial use. Ethical Research is an independent, clean-room work that reuses ideas from the general category of academic-research copilots, and says so plainly in ACKNOWLEDGMENTS.md. Holding ourselves to the citation standard we ask of users isn't a footnote here; it's the thesis.

A starting point, not the final word

Better safeguards, better academic conventions, and better integrity strategies are welcome through issues and pull requests. It doesn't claim to be the last word on AI and research ethics — it tries, in the open, to get it right. If it helped, leave a star.


Ethical Research — the right way, in the open.

About

Academic-integrity research copilot for Claude Code. Plans, reviews, verifies, and improves scholarly work; refuses ghostwriting, fabricated citations, and AI-use concealment. MIT.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages