Becoming an AI Researcher
Topic 14 · Part 3 · Writing & publishing · ⏱ about 2 h

Integrity, Authorship & Using AI Tools

Plagiarism and how to cite properly, self-plagiarism and duplicate submission, authorship and contribution statements, honest reporting, and using LLMs responsibly within venue and university policies.

Show:
In this topic
Beginner

Cite properly, and never plagiarise

Plagiarism is presenting others' words, ideas, figures, code or results as your own. In research it can end careers. Rules of thumb:

  • Cite the source of every idea, method, dataset, result and figure you use or build on, including blog posts and code.
  • Paraphrase genuinely (your own structure and words) and cite; use quotation marks for exact wording, sparingly.
  • Reusing a figure needs permission or a licence that allows it, plus attribution.
  • Self-plagiarism: reusing large parts of your own published text without acknowledgement. Extending a workshop paper into a conference paper is usually fine if the venue allows it and you say so.
  • Dual submission: submitting the same work to two venues at once violates almost every venue's policy.
Beginner

Authorship

Authors are people who made a substantial intellectual contribution (ideas, design, experiments, analysis, writing) and can take responsibility for the paper. Discuss authorship and order early, and revisit it as contributions change. In AI the first author usually did most of the work; the last author is often the senior supervisor. Many venues use contribution statements (e.g. the CRediT taxonomy) to record who did what. Providing data or funding alone may merit acknowledgement rather than authorship, depending on norms; follow your institution's authorship policy.

Beginner

Honest reporting

Fabricating or falsifying data or results is serious misconduct. Less obvious practices also mislead: reporting only the best seed, tuning on the test set, dropping datasets where the method loses, changing the hypothesis after seeing results (HARKing) without saying so, or using an unfair baseline. The safeguard is the habits from Part 2: plan analyses, report all runs and comparisons, release code and data, and describe exploratory findings as exploratory.

Beginner

Using LLMs and AI tools responsibly

Policies differ by venue, publisher and university, and they change, so read the current policy for each submission. Common positions:

  • LLMs cannot be authors: they cannot take responsibility.
  • Using tools to improve grammar and wording is generally allowed; using them to generate substantial content, analysis or ideas often must be disclosed (some venues ask how they were used).
  • You are fully responsible for every claim, citation and line of code. LLMs invent plausible but nonexistent references: verify every citation against the actual paper.
  • Reviewers must not paste confidential submissions into public AI tools.
  • For theses and coursework, your university's academic integrity rules apply: ask your supervisor what is permitted.

Used well, AI tools speed up coding, searching and editing. Used badly, they produce confident errors with your name on them.

Check yourself

Quiz

1. You paraphrase a paper's explanation in your own words. You should…

2. Submitting the same paper to two conferences at once is…

3. An LLM suggests five references for your related work. You…

4. Changing your hypothesis after seeing results and presenting it as planned is called…

5. Can an LLM be listed as a co-author?

Go deeper

📚 Further reading

Trusted, mostly free sources: official documentation, well-known textbooks and courses. When you want to confirm or go deeper on anything here, start with these.

Ready