Domain research: Act, Input, Mission

AIM applied to research. Start from the law or mandate, collect four kinds of source, and say exactly what the research must deliver. AI helps you summarise and prepare, but only from sources you give it.

Mohd FikriMethods & frameworks6 min readBeginner
Baca dalam Bahasa Melayu
  • Draft: this article is under editorial review and may change.

You have just moved to a new kampung, and you want to hold a kenduri. You don’t start by ordering food. First you ask around. Is there a rule about using the surau hall? How was the last kenduri in this kampung run? Who is the mak cik everyone asks about cooking for 200 people? What does the Ketua Kampung expect to be told?

After a few days of asking, you know the rules, you have seen how others did it, and you know the people. Then you can plan a kenduri that fits the kampung, not one copied from somewhere else.

That is domain research: learning how a piece of work is done before you change it. A The area of work you are studying, with its own rules, documents, systems and people. For example, KRT grant claims or community mediation. is the area of work you are studying. This lesson shows how to research one with the AIM shape you learnt in the previous lesson, and how AI can help without making things up.

The idea in 30 seconds

Act: the law or mandate

Start from the rule that gives the work its purpose and its limits.

The idea in 30 seconds0:00 / 0:26
  1. Act: the law or mandate Start from the rule that gives the work its purpose and its limits.
  2. Input: four kinds of source Literature, standards, existing implementations and stakeholders.
  3. Mission: what to deliver Say what the research must hand over, how long, and by when.
  4. Check every claim AI works only from sources you supply. You check each point against the source.
  5. Research from sources, not from memory.

Key takeaways

If you remember nothing else, remember these.

  • Act: the law, policy or mandate that frames the domain, such as the Akta Rukun Tetangga 2012 (Akta 751) for KRT work.
  • Input: four kinds of source: literature, standards, existing implementations and stakeholders.
  • Mission: exactly what the research must deliver.
  • AI may summarise and compare only the sources you give it. Every claim gets checked against its source.

Ask around before the kenduri

Tap between the two research briefs below. Both ask for help with the same job. Only one tells AI where the answers must come from.

Act: the Akta Rukun Tetangga 2012 (Akta 751) and our department's KRT guide frame this work. The relevant extracts are attached.

Sources: (1) the grant claim section of our KRT guide, (2) the claim form, (3) notes from last month's meeting with three KRT treasurers, names removed. Use only these sources.

Summarise the current claim process in no more than 10 steps. For each step, name the source and page. List any step where the sources disagree.

What you typically get back

A short process map you can check line by line. Each step points to a page, so a wrong step is easy to spot and fix.

Same shape as AIM, one new letter

In the last lesson, AIM meant Actor, Input, Mission: who answers, what it needs to know, and what to produce. Domain research keeps the same shape. Only the first letter changes.

AIM prompt Domain research
A Actor: who the AI should be Act: the law or mandate that frames the work
I Input: what the AI can’t guess Input: the sources you collect
M Mission: what to produce Mission: what the research must deliver

Tap a stop, or press Play to walk through every source.

Domain research as AIM: the Act, four kinds of Input, and the MissionSix stops in a row. The Act, meaning the law or mandate, frames the domain. Four kinds of Input source follow: literature, standards, existing implementations and stakeholders. The Mission at the end says what the research must deliver.A · ActI · Input: four kinds of sourceM · MissionActlaw and mandateLiteraturereports, papersStandardsformats, rulesImplementationswhat others doStakeholderspeople to askMissionwhat to deliverDomain research as AIM: the Act, four kinds of Input, and the MissionA · ActI · Input: four kinds of sourceM · MissionActlaw and mandateLiteraturereports, papersStandardsformats, rulesImplementationswhat others doStakeholderspeople to askMissionwhat to deliver

The Act frames the domain, four kinds of source feed it, and the Mission says when the research is done. AI helps at each stop; you check and decide.

A for Act: start from the mandate

Every piece of government work rests on a mandate: a law, a policy, a directive or your department’s own rules. The Act tells you why the work exists, who may do what, and where the limits are.

For anything to do with KRT, the starting point is the Akta Rukun Tetangga 2012 (Akta 751), together with your department’s KRT guidelines. For other work it may be a policy, a circular or a standing instruction.

Three habits keep this part honest:

  • Use the official text. Get the Act or policy from an official source or your department’s legal or policy unit.
  • Check the version. Laws and guidelines get amended. Confirm which version is in force before you rely on it.
  • Never ask AI to quote a law from memory. It can produce section numbers and wording that look real and aren’t. Paste the relevant extract, and ask it to work only from that.

I for Input: four kinds of source

Once you know the frame, collect what fills it. Domain research uses four kinds of source. Here is what each looks like if you are preparing to automate KRT grant claims.

Source What it is KRT grant claims example
Literature What has been written already Past internal reports on grant claims, briefing notes, your office’s own files
Standards The formats and rules the result must follow The claim form, the finance section of your KRT guide, your department’s finance rules
Existing implementations How the job is done now, here and elsewhere How your district handles claims today, how another district does it, how another agency handles grants to volunteer groups
Stakeholders The people who live the process A KRT chair, a KRT treasurer, a district officer, a finance officer

A good research pack has at least one item from each row. If a row is empty, that is your next task.

The same four rows work for other topics. If you are researching how to respond to community tension reports, your literature might be past incident summaries, your standards the reporting format your division uses, your implementations how other states handle early warnings, and your stakeholders a mediator and a district officer.

M for Mission: say what the research must deliver

Research without a Mission never ends. There is always one more document to read. A clear Mission says what you will hand over, how long it is and when it is due.

Act: our division's mandate on early warning of community tension, and its current reporting guideline (extracts attached).

Sources: five past incident summaries (anonymised), the reporting form, and notes from two interviews with mediators.

Produce a two-page brief for the division head: the current response steps, three gaps the sources show, and four questions to ask the state offices. Cite the source for every point. Due Friday.

What you typically get back

A two-page brief with a clear end point. Every gap and question traces back to a source you supplied, so the division head can check it.

Let AI summarise only what you give it

AI is a fast reader. Give it a stack of documents and it can summarise them, compare them and pull out the differences in minutes. The risk is that it fills gaps with things that sound right. This is called HallucinationWhen an AI states something false with confidence, for example inventing a figure, a circular or a study that does not exist.. The fix is to keep it tied to your sources and to check every claim.

Summarise from your sources only
You are a careful research assistant for a district unity office.

Below are four sources, labelled S1 to S4: the grant claim section of our KRT guide (S1), the claim form (S2), an internal note on common claim problems (S3), and interview notes from two KRT treasurers with names removed (S4). Use only these sources. Do not use outside knowledge.

Summarise what the sources say about why grant claims are returned. Give a table with the columns Reason, Source (S1 to S4, with page or paragraph), and Exact words (a short quote from the source).

If something is not in the sources, write "Not found in the sources". Do not guess.

When the table comes back, check it before you use it:

  1. Open each source at the page the AI points to.
  2. Find the quoted words. If they aren’t there, the row is wrong.
  3. Check that the reason actually follows from the quote.
  4. Delete or fix any row that fails, and note it.

Brief AI to prepare interview questions

Stakeholders tell you what documents don’t: where the process really gets stuck. AI can help you prepare good questions, so your time with a KRT chair or a district officer is well spent. Tap each coloured part of this brief.

Dissect a prompt

Interview questions for a KRT chair

Tap a coloured part of the prompt to see what it does.

You are an experienced community researcher who writes clear, neutral interview questions in Bahasa Melayu. KRT work is framed by the Akta Rukun Tetangga 2012 (Akta 751) and our KRT guide. The grant claim section is attached. Our source summary so far is attached. It shows that claims are often returned for missing receipts and late submission. The interviewee chairs a KRT, is a volunteer, and has 30 minutes in the evening. Draft 8 open questions about the grant claim process, in plain BM, in the order I should ask them. Mark any question that a district officer would answer better.

You ask the questions, not the AI. Listen, follow up, and record the answers with the person’s consent. A second version of the same brief, for a district officer, might focus on approval steps and common delays instead.

Common mistakes

  • Skipping the Act. Without the mandate, you can design something your department isn’t allowed to do.
  • Asking AI what the law says. Give it the text. Don’t ask it to remember.
  • Only reading documents. Documents show the official process. Stakeholders show where it really gets stuck.
  • No Mission. If you can’t say what you will hand over, and by when, the research isn’t ready to start.

Previous: the AIM method. Next in the path: Verb, Object, Classifier, where you learn to write findings anyone can trace.

Check your understanding

Check yourself

Domain research in five questions

Pick an answer, then tap Check answer. You can also use the arrow keys to move between answers.

Question 1 of 5: In domain research, what does the A stand for?

Nice work!

Lesson complete

You made it to the end. Take the quiz above if you have not yet, for bonus XP.

  • 0 XP
  • 0 day streak