Skip to content

The Research Paper

Every project on Askalot carries exactly one Research Paper: the single record of that research, from the first idea to the final verdict. It is opened during ideation, grown through design and fielding, and closed during analysis — so instead of a folder of disconnected artifacts, you finish a project holding one document that states what you set out to learn, how you measured it, who you reached, and what the data supports.

It is also the reason the AI agents behave as one team rather than four tools. Each agent reads the paper before it acts and writes its own findings back into it, so the Manager plans fieldwork against the goals the Designer recorded, and the Analyst grades the results against those same goals.

Opening it

The paper opens as its own page in a new tab, never as a panel inside another screen — it is the artifact, not a view.

Where you are How to open it
A project's page in Targetor Open the research paper
A Bundle's Overview in Balansor The Research Paper accordion
The code editor (Armiger) The Paper row in the Project panel

In the code editor you also get the paper mirrored into a read-only paper/ folder — one file per written chapter, refreshed each time you open the project. Editing those copies accomplishes nothing; they exist so you can read the paper beside the questionnaire you are writing.

Printing is how you get a PDF

The page carries a print stylesheet, so your browser's Print → Save as PDF produces the document — table of contents, chapters, and figures — without the surrounding application. There is no separate export step to run and no compile to wait for: the paper is assembled fresh every time you open it, so what you print is what the project holds right now.

The chapters

The paper has a fixed set of thirteen chapters, and all of them are listed from the start — whether or not anything has been written into them:

Abstract · Motivation · Related Work · Research Goals and KPIs · Instrument Design · Population and Sampling · Data Collection · Coding and Weighting · Studies · Data Quality Assessment · Discussion · Conclusion · Appendix

An unwritten chapter reads as "not yet written" rather than disappearing. That is deliberate: the structure is what tells you where the research stands, and a paper that only listed finished chapters would hide exactly the gaps you need to see.

Two chapters behave differently from the rest:

  • Studies is the only chapter whose contents you choose — see below.
  • Appendix draws the instrument rather than reproducing it. The figure shows the questionnaire's shape at block level — how many blocks, how many of them conditional — because a graph of every item in a real instrument is a hairball nobody reading a paper can use. When you need the instrument itself, download the QML from the questionnaire's page in Targetor; a download can name an exact version, so a paper can cite the instrument as it was rather than as it is now.

The Related Work chapter also carries your project's source material — a synthesis of the reference documents you uploaded, kept current as you add more. In the code editor it has its own Source Material row in the Project panel.

Who writes it, and how you stay in control

The paper is written by agents and read by people. You do not type into it directly, and that is the point: every change to it is attributable, versioned, and reviewable.

What reaches you instead is a proposal. When an agent wants to change a chapter, it files the proposed text; you approve or reject it. Approving lands that text as the chapter's new version, with your approval recorded alongside the agent that wrote it. Nothing is overwritten silently, and rejecting costs nothing.

  • A project's page in Targetor shows a badge with the number of proposals waiting for you, so a pending change is visible without opening the paper.
  • If a newer proposal arrives for the same chapter before you have reviewed the older one, the older one is closed automatically — you are never asked to choose between two versions of the same suggestion.
  • A rejection can carry a comment, so your reason survives with the record.

Agents are held to the same discipline between themselves. An agent must read a chapter immediately before changing it, and it changes a passage rather than replacing the whole chapter, so one agent's work never quietly erases another's. If the chapter moved since it read, its edit is refused and it has to look again.

Studies — choosing which analyses appear

Each time the AI Analyst evaluates a Bundle, it writes a Study: its verdict on that dataset. The Studies chapter is assembled from the Studies you have promoted, in an order you control.

Action What it does
Promote Puts that Bundle's Study into the paper. A successful Analyst run promotes automatically, so a completed analysis is never written but invisible
Demote Takes it back out. The text and its history survive — it is hidden, not deleted
Reorder Rearranges the promoted Studies into the sequence you want a reader to follow
Refresh Clears a staleness marker once you have brought the Study up to date

Writing and promoting stay independent, so an analysis you ran to check something never has to appear in the paper.

A stale Study keeps its words

Re-running the datasets behind a promoted Study marks it stale. The prose is left exactly as written and a marker tells your reader that the numbers underneath it have moved. Askalot will not rewrite your conclusions because the data changed — that judgement is yours.

Who can read it

A paper is reachable by the people who own the project, and by nobody else. The link carries only a project identifier, so a paper URL passed to someone outside the project is refused rather than read. Reading and writing the paper are open to every plan — no part of it is gated on a subscription tier.