| ID | Claim | Prio | Pramāṇa | Cond | Tier | Ed | Attr | Ind | Conflict | Self | Tested | Ovr | PS | SR | CO | IC | Final | Verdict |
|---|
Edit any dropdown to score in bulk. Click an ID to open it in Add / Edit. Penalties and the fallacy checklist live in Add / Edit.
How a score is built — the whole model in one line
your inputs→
PS how good is the kind of knowledge?
SR how trustworthy is the source?
CO how many independent sources agree?
IC does it clash with the timeline?
→weighted sum→
− penalties→verdict
You never type a score. You answer questions about the claim, and the four measures are computed from your answers.
PS = Pramāṇa Strength · SR = Source Reliability · CO = Corroboration · IC = Internal Consistency.
While you type — the habits that keep the data clean
- One proposition per claim. If it has an “and”, split it.
- Match the date to the source. Don’t invent a day the source doesn’t give.
- Name the calendar. A bare year is ambiguous (Gregorian / Julian / Śaka / Vikrama).
- Separate the event from its dating — they rest on different evidence.
- Separate what the source says from what you infer.
- One canonical form per name, with aliases once — and disambiguate bare names (“Raja Ajit Singh of Khetri”).
- Cite the text and the edition or translation you actually used — never “the scriptures”.
- Mark literal or figurative; write “uncertain” rather than deciding silently.
- Never let a premise assume the conclusion — that is a प्रकरणसम.
- State the falsifier. If you can’t, it may not be a claim.
Full detail in the Guide tab. None of this is enforced by the engine yet — the entity layer and the interpretation rules are not built. These habits are what make those layers a mechanical clean-up later rather than a re-reading of everything.
New claim
Claim text
One proposition that could be true or false. If it contains an “and”, split it. Keep the source’s own wording for the assertion, and put your gloss in the notes. State dates only as precisely as the source does, and name the calendar.
ID
Kind
claim — an assertion to be scored. ground — a source or artefact that stops the regress (manuscript, inscription, photograph, or testimony accepted as reliable).
Status
Workflow state, not the score.
Source attribution (whose claim)
The person or work the claim comes from — written in its canonical form (“Swami Vivekananda (Narendranath Datta)”), with the edition or translation if it is a text. Kept separate from who typed it in.
Transcriber (who typed it)
The operator entering this record. Kept separate from the source, so provenance survives.
Reviewer inputs — these four groups produce PS, SR, CO and IC
Defects — the fallacy checklist. Each ticked defect is named and deducted.
Additional manual penalty
Total penalty
Tick only what genuinely applies; apply one per category. A ticked savyabhicāra here is a judgement — the system also detects a deviating reason automatically from weak inference links and warns you.
Necessary premises — and how strongly each one actually supports this claim
Add a premise
A claim is never stronger than its weakest link: each premise's contribution is its own score multiplied by its warrant — how strongly it actually supports this conclusion (व्याप्ति). A premise can be perfectly true and still support nothing here. If you name an upādhi (the condition that would break the inference) and do not exclude it, the link is treated as vitiated.
Established tenet and counterfactual
Conflicts with an established tenet (सिद्धान्त)
If this claim contradicts an accepted tenet, it is judged against that tenet, not only against its own premises — IC goes to 0 and a siddhānta penalty applies. Maintain the register in the Siddhānta tab.
What would falsify this claim? (तर्क)
State the test that, if it failed, would count against the claim. An empty answer suggests the claim may be unfalsifiable — which is itself worth knowing.
Reading of the passage
If the claim rests on a text, is the passage to be read literally or figuratively? This is what lets a dispute be recognised as an उपचारच्छल — a quibble that takes a figurative word literally (NS 1.2.14).
Objections — challenges made against this claim, and whether they are oppositions at all
New objection
Kind
A substantive objection is a real counter-claim and competes with this claim on the score. A jāti (24 kinds, NS 5.1), a nigrahasthāna (22 points of defeat, NS 5.2) or a chala (3 kinds, NS 1.2.11–1.2.14) is recorded and then set aside — Nyāya's point is that many objections are not oppositions at all, and scoring them as weak counter-claims would be a category error.
Priority (प्रयोजन) — what to score first. Never enters the score.
Priority — used only to order the work queue.
Live score
Focus on
Show
Edge labels show the warrant (व्याप्ति). A low warrant is a weak inference, whatever the premise's own score. Click any box to focus it.
Register of established tenets (सिद्धान्त) — the background knowledge claims are judged against
A siddhānta is a settled tenet that a valid cognition must not contradict. This register is the system's only background knowledge, and it is also where the system takes something on faith — so keep it short, explicit, and arguable. Gautama distinguishes four kinds (NS 1.1.26–1.1.31), and they carry different weight: a claim contradicting a common tenet — held by all schools, opposed by none — is in far more serious trouble than one contradicting a peculiar tenet, which is one school's position by definition.
Tenet
Kind of tenet
| ID | Tenet | Kind (सिद्धान्त प्रकार) | Penalty | Claims conflicting |
|---|
Model parameters — the calibration surface
Every number the scoring model uses is here. Change any one and every score re-derives. The ordering of pramāṇas is defensible from the shastra; the gaps between them are calibration parameters, placeholders until your data says otherwise. In a multi-user setting only an administrator should touch this tab.