1. Start with a question a study can answer
Most first papers fail long before the writing, because the idea is a topic, not a question. “AI in education” is a topic. “Does spaced quizzing improve two-week recall in school ICT classes compared with re-reading?” is a question: it names what you will measure and what you will compare it with.
Then check whether it has already been answered. Search Google Scholar, the ACM Digital Library, IEEE Xplore and Scopus for the key terms, and read the newest five or six relevant papers closely. Write one paragraph that says what is known, what is not, and where your study sits. If the answer already exists, narrow the question or plan a replication: some venues welcome replications, and they are an honest way for a first-time author to contribute.
2. Pick the venue before you write
The venue decides the length, the template, the style of evidence and the deadline, so choose it first. There are three main routes.
- A conference. Fixed deadlines, a page limit and a decision in a few months. In computing and human-computer interaction, a full paper at a good conference carries real weight. Usually at least one author must register and present.
- A journal. Rolling submission, longer papers, and slower review, often several months per round. The usual route in many other fields.
- A preprint server such as arXiv. Free and immediate, but not peer reviewed. It dates your work and gets it read; it is not a substitute for publication. Check the target venue’s preprint policy first, and note that arXiv can ask first-time submitters for an endorsement in some categories.
Before you commit, check the venue’s scope against your paper, how its peer review works, where it is indexed (Scopus, Web of Science, DBLP, or the Directory of Open Access Journals for open-access titles), who the publisher is, and what it charges. The Think. Check. Submit. checklist is a good, free way to test an unfamiliar journal.
Warning signs of a predatory venue
- An unsolicited email promising fast acceptance or a guaranteed decision date.
- A scope so broad it covers everything, or an editorial board you cannot verify.
- A fee requested before any review, or an invented “impact factor”.
- A name that closely imitates a well-known journal or conference.
A paper in such a venue can do more harm than none, because panels and employers increasingly check. If you are unsure, ask a supervisor before you submit.
3. Design the study and get ethics approval
Write the method down before you collect anything: who your participants are, how many, what you will measure, and how you will analyse it. Decide the analysis before you see the data, because choosing the test afterwards, to get the answer you hoped for, is how weak results get published and later retracted. Small samples are fine if the claims are modest and the limits are stated.
If your study involves people, whether they take surveys, use a prototype or are observed, you need ethics approval from your institution’s ethics committee or review board before you collect data. Reviewers and editors ask for the approval number, and approval given after the fact is rarely accepted. Participants must give informed consent, and their data must be stored and shared lawfully: under the GDPR where it applies, and under Sri Lanka’s Personal Data Protection Act for data collected there. Anonymise early.
4. Collect and analyse the data
Keep the raw data untouched and keep a log of every decision you make about it: what was excluded, why, and by what rule. Pick statistical tests that suit your data and your question, and report effect sizes with confidence intervals rather than a bare p-value. An honest “we found a small effect” is publishable; an unexplained exclusion of three awkward participants is not.
Where the venue and your participants allow it, share the data and the analysis code. Reviewers trust what they can check.
5. Write it the way reviewers expect
Most empirical papers follow the same shape: Introduction, Methods, Results and Discussion, often called IMRaD, with an abstract on top. Write in the order that is easiest: methods first, then results, then the introduction and discussion, and the abstract last. Use the venue’s own template from the first day; IEEE, ACM and Springer all publish them, and reformatting at the end costs days.
- Every claim in the abstract and the conclusion must be supported by a result you can point to.
- State the limitations yourself. Reviewers find them anyway, and a paper that names them reads as careful.
- Cite fairly, including work that disagrees with you, and never paste text you did not write.
- Check the venue’s rules on AI writing tools, and disclose any use where it asks for that.
7. Submit, and read the decision correctly
Submit through the venue’s system with a short cover letter that states the question, the contribution and why this venue. Anonymise the manuscript if review is blind, and give each author an ORCID iD, a free persistent identifier that follows you across papers.
The decision usually falls into one of five groups:
- Desk rejection. The editor declined it without sending it to reviewers, often for scope. It is not a verdict on quality.
- Reject. Reviewers saw it and found a serious problem. Read the comments; they are free advice for the next venue.
- Major revision. A good sign. The editor still wants the paper if you fix real problems.
- Minor revision. Close. Fix what is asked and resubmit.
- Accept. Rare on the first round.
Review takes weeks to many months and varies enormously between venues, so be wary of anyone who promises a date. Nobody who truly depends on reviewers can.
8. Answer the reviewers
A revision is two documents: the changed paper and a response letter. In the letter, thank the reviewers, then take each comment in turn: quote it, say what you changed and where (page and line), or explain politely, with evidence, why you did not. Never ignore a comment, and never argue from irritation.
The most common mistake is rewriting the letter and not the paper. Reviewers read the paper again, and a comment answered only in the letter will come back.
After acceptance: fees, licences and your record
Some journals charge an article processing charge (APC) to make your paper open access; others charge nothing, and many have waivers. Conferences charge a registration fee, and at least one author usually has to pay it and present. These costs fall on the authors and vary hugely, so ask about them at the start, not after acceptance.
You will also sign a copyright or licence agreement, often a Creative Commons licence for open access. Read it, and check what the publisher lets you share and when. Finally, put the paper on your ORCID record and your CV with its true status: accepted, in press or published.
Where help is worth having
A supervisor or co-author who has been through peer review saves months, mostly by stopping the avoidable mistakes above: a vague question, the wrong venue, missing ethics approval, a claim the data does not support. That is what Publish With Me is for: I supervise, co-author where I genuinely contribute, run the user study with you and walk you through peer review. It is free for students with a valid ID, and quoted for businesses. I cannot guarantee acceptance, because peer review decides, and article charges and conference fees are paid by the authors.