Research · Four-arm workflow · Diagram B · Full rebuild

Arm B: ask the one question that splits the most diseases

Every arm answers the same three questions. What do I ask next? When do I stop? What may I say? Arm B ranks one question by how many diseases its answer would split. It refuses by coverage.

Arm B: ask the one question that splits the most diseases Arm B does a fixed-rule walk of the approved catalogue with word bands. It asks the one question that would split the most diseases still on the list. It refuses by coverage when the catalogue does not cover the Health Event. while an answer could still change the advice and the question limit allows another question. 1 HEAR Model only turns Thai into facts. It decides nothing. 2 BUILD THE LIST Sort signs 3 ways. Then a plain word. Never a number. Reportables stay. Likely Possible Cannot exclude 3 CHOOSE WHAT TO ASK Rank questions by how many diseases an answer would split. Pick one. 4 ASK One question per message. Unasked ones are written down with why. 5 DECIDE WHETHER TO STOP Stop when no answer could change advice. 4 + 1, about five. Hard stop at 10. 6 DECIDE WHAT MAY BE SAID Name one disease, state an honest tie, or send to the veterinarian. Never a percentage. 7 SAY Model writes Thai from approved wording. Second and last time. 8 RECORD Evidence Trace (the full written record): the facts, each disease's tally and band, the question chosen, every question skipped and why, and the reason for stopping. 9 REFUSAL / VET DOOR BY COVERAGE METAPHOR ONE QUESTION SPLITS THE PILE two groups the answer splits most RMC GRID 48x3=144 · B grid baseline acc 74/144 beside C1 77/144 · RMC GRID 48x3=144 · B wrong reportable flag 46/144 beside C1 30/144 · answer key follows Arm B LEGEND Stage Accent / door Record Main flow Loop / write

Full sentences · matched to the numbered chips

  1. 1

    HEAR

    The model is used only to turn the Thai into facts. It decides nothing.

  2. 2

    BUILD THE LIST

    The code looks at each disease in the approved catalogue. It sorts the signs the catalogue lists for that disease into supporting / contradicting / not yet asked. Then the code gives it a plain word from a fixed 8-row table: Likely, Possible, or Cannot exclude. Never a number. The code orders the diseases still on the list. Reportables stay in the protected set. The code always assesses them.

  3. 3

    CHOOSE WHAT TO ASK

    The code ranks every possible question by how many diseases still on the list its answer would split. It picks one.

  4. 4

    ASK

    The code asks one question per message. It writes down every question it does not ask, with the reason.

  5. 5

    DECIDE WHETHER TO STOP

    The code stops when no remaining answer could change what we advise the farm to do. It also stops when the question limit is reached: 4 answers, plus 1 extra safety question. That is about five questions at most in an ordinary case. There is a hard stop at 10 asking turns.

  6. 6

    DECIDE WHAT MAY BE SAID

    The code names one disease, states an honest tie, or sends the case to the veterinarian. It never shows a percentage.

  7. 7

    SAY

    The model is used a second and last time. It only writes the Thai from approved wording.

  8. 8

    RECORD

    The Evidence Trace records the facts and each disease's tally and band. It also records the question chosen, every question skipped and why, and the reason for stopping.

  9. 9

    REFUSAL / VET DOOR

    REFUSE BY COVERAGE. Does the approved catalogue cover this Health Event at all? If not, the door sends the case to the veterinarian. No confidence number is needed.

Caption

Fixed-rule walk of the catalogue

Code owns every clinical decision. The model appears only at the two edges. Nothing is deployed.

B is the same starting machine (this arm) on RMC GRID 48x3=144. It is the shared reference the C arms are compared against. The grid is hard. B grid baseline acc 74/144 beside C1 77/144. B wrong reportable flag 46/144 beside C1 30/144. B mean questions 3.58 beside C1 3.84. The answer key follows Arm B, so the test tilts toward B-like answers.

Changed from

Changed from A

Changed from A: it takes only the shared nine-stage frame. It adds a full rebuild. The full rebuild uses an approved catalogue. It asks the one question that splits the most diseases. It stops only when no remaining answer could change the advice. It refuses by coverage, with reportables in the protected set. It records an Evidence Trace.

How to read

Marks on this drawing

  • Cyan: CHOOSE WHAT TO ASK, and the open coverage door
  • Bands stay grey: Likely, Possible, Cannot exclude are words, not probabilities
  • Loop: keep asking while an answer could still change the advice and the question limit allows another question
  • Score strip: RMC GRID 48x3=144, B is the grid baseline beside C1; the answer key follows Arm B

Research only - nothing is deployed - no percentage ever reaches the Farm Operator - the 14 Disease Profiles are vet-reviewed - reviewed, not signed.