為什麼倫理在投稿前處理
多數倫理問題（IRB、作者排序、利益衝突、過去發表）不能事後補救。一旦在投稿後被 reviewer / editor 發現，輕則 reject，重則撤稿 + lab 被列黑名單。
2026 必填的四項：(1) ICMJE 4 條件確認所有 author；(2) CRediT taxonomy 列每人貢獻；(3) 每位作者的 ORCID iD；(4) COI statement。
Most ethics issues (IRB, author order, conflicts of interest, past publications) can't be retroactively fixed. Discovered post-submission, the consequences range from reject to retraction + lab blacklist.
2026 required-fillables: (1) ICMJE 4-criteria check for every author; (2) CRediT taxonomy listing per-author contributions; (3) ORCID iD for each author; (4) COI statement.
一、ICMJE 4 條件
ICMJE (International Committee of Medical Journal Editors) 規定，必須同時符合 4 條件才能掛名 author（2026/01 更新版本）：
The ICMJE (International Committee of Medical Journal Editors) requires that authors must meet all 4 criteria (Jan 2026 revision):
實質學術貢獻
在 conception/design 或 data acquisition/analysis/interpretation 中有實質貢獻。
Substantial contribution to conception/design OR data acquisition/analysis/interpretation.
撰稿或關鍵修訂
參與撰寫 manuscript 或對重要 intellectual content 進行批判性修訂。
Drafted the work OR critically revised it for important intellectual content.
同意最終版本
同意 manuscript 最終要發表的版本 (sign-off on the final version)。
Approved the final version to be published.
負責 accountability
同意對研究的所有部分負責 (accountable for accuracy and integrity of all parts)。
Agree to be accountable for all aspects of the work, ensuring accuracy and integrity.
二、CRediT 14 角色
CRediT (Contributor Roles Taxonomy) 由 NISO 在 2014 年制定、2022 年成為 ANSI/NISO Z39.104 標準。Elsevier, Springer Nature, Cell Press 等主流期刊已強制要求每篇論文逐個作者列出 1-N 個角色。
CRediT (Contributor Roles Taxonomy) was developed by NISO in 2014 and became the ANSI/NISO Z39.104 standard in 2022. Elsevier, Springer Nature, Cell Press, and other major publishers now require per-author CRediT listings.
① Conceptualization
提出研究想法、formulation of overarching goals。通常是 PI / 主導學生。
Ideas, formulation of overarching goals. Usually PI / lead student.
② Methodology
設計方法、模型開發。技術上的 architect。
Method/model design. Technical architect.
③ Software
編寫程式、實作演算法。
Programming, algorithm implementation.
④ Validation
驗證 reproducibility、independent replication。
Verifying reproducibility, independent replication.
⑤ Formal analysis
執行統計、計算分析。
Statistical / computational analysis.
⑥ Investigation
實驗執行、資料收集 (wet-lab)。
Conducting experiments, data collection (wet-lab).
⑦ Resources
提供樣本、試劑、器材、計算資源。
Providing samples, reagents, equipment, compute.
⑧ Data Curation
清整、標註、archive 資料。
Cleaning, annotating, archiving data.
⑨ Writing — Original Draft
撰寫第一版 manuscript。
Drafting the initial manuscript.
⑩ Writing — Review & Editing
修訂、評論、潤稿。
Revisions, comments, polishing.
⑪ Visualization
製作圖表、視覺化呈現。
Producing figures & visualizations.
⑫ Supervision
研究指導、mentorship (典型 PI 角色)。
Research oversight, mentorship (typical PI role).
⑬ Project Administration
專案規劃、進度管理。
Project planning, schedule management.
⑭ Funding Acquisition
獲得 funding 支持本研究。
Acquired funding for the study.
三、ORCID 與 ROR
👤 ORCID iD
16 位數字（如 0000-0002-1825-0097）唯一識別作者。2026 多數主流期刊（PLOS, Nature, Cell, Springer, Wiley）已強制要求 corresponding author 提供 ORCID，且鼓勵共同作者也都提供。
益處：避免姓名同名混淆（特別 Asian author）、自動 sync publication list、整合 funder reports。
A 16-digit unique author identifier (e.g., 0000-0002-1825-0097). By 2026, most mainstream publishers (PLOS, Nature, Cell, Springer, Wiley) require corresponding authors to provide ORCID, with co-authors strongly encouraged.
Benefits: disambiguates same names (especially Asian authors), auto-syncs publications, integrates with funder reports.
🏛️ ROR (Research Organization Registry)
類似 ORCID，但是給 institution 用的全球唯一 ID。例：「National Taiwan University」對應 ROR ID。許多期刊在 metadata 中已內含。
Like ORCID but for institutions — globally unique. E.g., "National Taiwan University" maps to a ROR ID. Many journals embed it in metadata.
四、學術誠信紅線
| 行為 | 定義與後果 |
|---|---|
| ① 偽造 / 篡改數據 | 最嚴重的學術不端，會撤稿 + 機構調查 + 終身學術除名。Retraction Watch 上多數案例都源於此。Most serious misconduct — retraction + institutional inquiry + permanent academic ban. Most Retraction Watch cases originate here. |
| ② 抄襲 | 含 self-plagiarism (重複使用自己過去文字)。期刊全部用 iThenticate 比對。中文寫作翻譯成英文也算。Includes self-plagiarism (reusing your own past text). Journals run iThenticate. Translating Chinese writing to English without citation also counts. |
| ③ 重複發表 | 同一研究在兩家期刊以同樣或近似形式發表。Conference abstract OK，full paper 不行。Same research in two journals in identical or near-identical form. Conference abstracts OK; full papers not. |
| ④ Salami Slicing | 把一個本應一篇的研究切成多篇 (least publishable units, LPU) 來增加發表數。Editor 會比對 cohort 重疊度。Slicing one study into multiple papers (least publishable units, LPU) to inflate publication count. Editors check cohort overlap. |
| ⑤ Ghost / Gift / Guest Authorship | Ghost = 真實貢獻者沒掛名 (常為產業 medical writer)；Gift = 沒貢獻但掛名 (常為老闆 / 系主任)；Guest = 知名學者掛名增加可信度。三者都違反 ICMJE。Ghost = real contributor uncredited (often industry medical writer); Gift = uninvolved person added (boss / chair); Guest = famous name added for credibility. All violate ICMJE. |
五、利益衝突 (COI)
2026 ICMJE 推薦的 Disclosure of Interest form 是業界標準。原則：寧可 over-disclose，不要 under-disclose。
需揭露的事項（每位作者）：
- 本研究相關 funding（含 grant number）
- 過去 36 個月任何企業的 employment、consultancy、honoraria、stock options、patent、travel support
- 家屬與本研究相關企業的關係
- 非金錢的 conflict (advocacy, religion, political)
The 2026 ICMJE Disclosure of Interest form is the industry standard. Principle: over-disclose rather than under-disclose.
To disclose (per author):
- Funding for this study (including grant numbers)
- Last 36 months of any corporate employment, consultancy, honoraria, stock options, patents, travel support
- Family-member relationships with companies relevant to this study
- Non-financial conflicts (advocacy, religion, political)
六、倫理檢查清單
🌳 送審前 8 項倫理檢查
📝 自我檢測
1. 一位 lab manager 純粹「執行 SOP 收集樣本」但沒參與設計與寫作。她應該？
1. A lab manager who only "ran SOPs and collected samples" without design/writing input. She should?
2. 你把同一個 cohort 切成 3 篇 paper 投稿，每篇分析不同 subset。這稱為？
2. Splitting one cohort into 3 papers analyzing different subsets is called?
3. 對 ORCID 的下列何者最正確？
3. Most accurate statement about ORCID?