# Experiment Design Workbook

> ไฟล์นี้เป็น design record ไม่ใช่ผลวิจัย ช่องที่ยังไม่รู้ให้เขียน `TBD — owner/source/date` ห้ามให้ AI แต่งค่า citation approval หรือผลเพื่อเติมช่องว่าง

## 1. Intake

- Track: A own design / B published-study reverse engineering / C course dossier
- Problem:
- Objective/RQ:
- Candidate claim (one claim):
- Intended scope/population/environment:
- Evidence already verified:
- Target audience/venue family and verified scope:
- Unknowns requiring expert review:

## 2. Claim-to-hypothesis register

| ID | Type | Hypothesis / necessary condition | Why claim needs it | Observation required | Fail/inconclusive condition | Evidence/status |
|---|---|---|---|---|---|---|
| H-C1 | core | | | | | |
| H-M1 | mechanism | | | | | |
| H-V1 | measurement validity | | | | | |
| H-B1 | boundary/robustness | | | | | |
| H-S1 | safety/feasibility/compliance | | | | | |

Completeness attack from another group:

- Missing hypothesis/condition:
- Circular or untestable statement:
- Assumption incorrectly treated as evidence:

## 3. Experiment/study specification

| Element | Specification | Linked hypothesis ID | Evidence/rationale | Status/owner |
|---|---|---|---|---|
| Unit of analysis | | | | |
| Population/sample | | | | |
| Sampling/allocation | | | | |
| Intervention/exposure/phenomenon | | | | |
| Comparator/baseline/negative case | | | | |
| Parameters/settings | | | | |
| Primary outcome/metric | | | | |
| Secondary outcome/trade-off | | | | |
| Dataset/data source/version | | | | |
| Environment/context | | | | |
| Instrument/collection method | | | | |
| Procedure/timing | | | | |
| Confounder/bias/leakage | | | | |
| Analysis/decision rule | | | | |
| Failure/inconclusive rule | | | | |
| Target-venue evidence/reporting gap | | | | |

## 4. Instrument and measurement audit

| Construct | Instrument/proxy | Validity/calibration evidence | Reliability/repeatability | Pilot needed | Limitation/action |
|---|---|---|---|---|---|
| | | | | | |

## 5. Compliance and feasibility trigger screen

Use only: `no trigger identified`, `official determination required`, `approval/permission pending`, `blocked`.

| Trigger area | What in this design triggers review | Status | Responsible authority/person | Required document/action |
|---|---|---|---|---|
| Human participants/IRB | | | | |
| Personal/sensitive data/privacy | | | | |
| Animal/biological/biosafety | | | | |
| Site/data/device permission | | | | |
| License/IP/contract | | | | |
| Physical/psychological/environmental risk | | | | |

## 6. Result-response matrix

> ทุกแถวเป็น hypothetical scenario ไม่ใช่ research evidence

| Outcome state | Hypothesis IDs affected | Permitted conclusion | Forbidden conclusion | Claim status | Next action |
|---|---|---|---|---|---|
| Supports decision rule | | | | | |
| No difference / below threshold | | | | | |
| Opposite effect / harm / trade-off | | | | | |
| Inconclusive / invalid measurement or protocol | | | | | |

## 7. Red-team findings

| Finding | Location/element | Logic consequence | Response category | Revision/evidence/owner |
|---|---|---|---|---|
| | | | | |

Response category: `revise now`, `expert/statistician/IRB review required`, `collect pilot evidence`, `out of scope with justification`, `rebut with verified evidence`.

## 8. Design verdict

- Verdict:
  - [ ] design supports the candidate claim in principle
  - [ ] design supports only a narrower claim
  - [ ] design needs major revision
  - [ ] design is blocked pending evidence/approval/resource
- Narrowest claim currently supported by the design:
- Top three revisions:
  1.
  2.
  3.
- External review/approval required:
- What AI must not invent or decide: