📖 《把 AI 当助手——经济实证研究的诚信、方法与工具》双语网页版:每章中文部在前,English Edition 在后,内容与最新合订 PDF 一致。

附录 D · 诚信自查十问

问题 1：你的论文中是否有任何内容完全由 AI 生成且未声明？

对应：S0.0 / S11.1
自查方法：回忆写作过程，是否有整段/整节直接复制AI输出
合格标准：所有AI辅助内容已在声明中如实列出

问题 2：你引用的每一篇文献是否都经过 Zotero 或知网核实确实存在？

对应：S3.1 / S3.2
自查方法：随机抽取参考文献列表中的5条，逐一搜索确认标题/作者/年份/期刊正确
合格标准：0条虚构文献（AI幻觉引用是最常见的学术不端之一）

问题 3：你的每一个回归系数/统计数字是否都能从你的 do 文件中复现？

对应：S1.2 / S9.4
自查方法：清空 Stata → 从头运行 do 文件 → 对照论文表格逐格核对
合格标准：100%数字可复现，无手动修改

问题 4：你是否尝试过多种模型设定并只报告了显著的那个？

对应：S9.1
自查方法：回忆分析过程中“试了多少次”，是否每次都记录了
合格标准：所有合理设定均在稳健性检验或规格曲线中呈现

问题 5：你的问卷/访谈数据是否获得了被调查者的知情同意？

对应：S7.1
自查方法：检查是否有签署的知情同意书或录音开头的口头同意
合格标准：100%受访者知情同意，涉敏数据已脱敏

问题 6：你借用的量表是否注明了出处和原始信效度证据？

对应：S5.1
自查方法：检查问卷中每个量表是否有引用（如“借鉴Lusardi & Mitchell 2008”）
合格标准：所有非自编量表均有明确出处+原始α值报告

问题 7：你的抽样方式是否如实描述（非概率抽样没有说成随机抽样）？

对应：S5.2
自查方法：阅读论文“数据”一节，检查抽样描述是否与实际一致
合格标准：便利样本就写便利样本，不回避局限性

问题 8：你的理论框架图中每条箭头是否都有文献支撑？

对应：S2.1 / S4.1
自查方法：对照理论框架图，逐箭头找到参考文献列表中的对应条目
合格标准：0条无文献支撑的箭头

问题 9：你的 data/raw 文件夹是否保持只读（从未手动修改原始数据）？

对应：S8.1 / S9.3
自查方法：检查 data/raw/ 文件的修改时间是否与获取时间一致
合格标准：raw数据从获取后未被任何方式修改

问题 10：你的 AI 使用声明是否涵盖开题、投稿、答辩三个版本？

对应：S11.1 / S11.2
自查方法：检查三个版本文档是否存在且内容如实
合格标准：三版声明齐全、内容与实际操作一致、无遗漏

自查结果记录表

表 D.1 自查结果记录表。

L4|L4|L4|L4

问题 & 是/否/不确定 & 如“否”，整改措施 & 整改完成日期
1 & & &
2 & & &
3 & & &
4 & & &
5 & & &
6 & & &
7 & & &
8 & & &
9 & & &
10 & & &\

English Edition

Appendix D · Ten Integrity Self-Check Questions

Question 1: Is any part of your paper entirely AI-generated without disclosure?

Checkpoints: S0.0 / S11.1
How to self-check: recall the writing process—was any whole paragraph or section copied straight from AI output?
Pass standard: every AI-assisted passage is truthfully listed in the disclosure

Question 2: Has every reference you cite been verified in Zotero or CNKI as actually existing?

Checkpoints: S3.1 / S3.2
How to self-check: draw five entries at random from the reference list and search each one to confirm title, authors, year, and journal
Pass standard: zero fabricated references (AI-hallucinated citations are among the most common academic misconducts)

Question 3: Can every regression coefficient and statistic in your paper be reproduced from your do-file?

Checkpoints: S1.2 / S9.4
How to self-check: clear Stata → run the do-file from the start → compare cell by cell against the paper’s tables
Pass standard: 100% of numbers reproducible, with no manual edits

Question 4: Did you try several specifications and report only the significant one?

Checkpoints: S9.1
How to self-check: recall how many attempts the analysis involved, and whether each one was recorded
Pass standard: every reasonable specification appears in the robustness checks or the specification curve

Checkpoints: S7.1
How to self-check: check for signed consent forms or verbal consent recorded at the start of the audio
Pass standard: 100% of respondents gave informed consent, and sensitive data are de-identified

Question 6: Do the scales you borrowed state their source and original reliability/validity evidence?

Checkpoints: S5.1
How to self-check: check that every scale in the questionnaire carries a citation (e.g. “adapted from Lusardi & Mitchell 2008”)
Pass standard: every non-self-developed scale has an explicit source plus the original α

Question 7: Is your sampling method described truthfully (no non-probability sample presented as random)?

Checkpoints: S5.2
How to self-check: read the “Data” section of the paper and check that the description matches what you actually did
Pass standard: a convenience sample is called a convenience sample, and limitations are not evaded

Question 8: Is every arrow in your theoretical framework diagram supported by the literature?

Checkpoints: S2.1 / S4.1
How to self-check: go arrow by arrow through the framework diagram and locate the corresponding reference-list entry
Pass standard: zero arrows without literature support

Question 9: Has your data/raw folder stayed read-only (never manually editing raw data)?

Checkpoints: S8.1 / S9.3
How to self-check: check that the modification time of the files in data/raw/ matches when they were obtained
Pass standard: raw data untouched in any way since acquisition

Question 10: Do your AI-use disclosures cover the proposal, submission, and defence versions?

Checkpoints: S11.1 / S11.2
How to self-check: check that all three documents exist and their content is truthful
Pass standard: all three declarations present, consistent with actual practice, with nothing omitted

Self-check record sheet

Table D.1 Self-check record sheet.

L4|L4|L4|L4

Question & Yes/No/Unsure & Remedial action if “No” & Date completed
1 & & &
2 & & &
3 & & &
4 & & &
5 & & &
6 & & &
7 & & &
8 & & &
9 & & &
10 & & &\

附录 D · 诚信自查十问 ​

问题 1：你的论文中是否有任何内容完全由 AI 生成且未声明？ ​

问题 2：你引用的每一篇文献是否都经过 Zotero 或知网核实确实存在？ ​

问题 3：你的每一个回归系数/统计数字是否都能从你的 do 文件中复现？ ​

问题 4：你是否尝试过多种模型设定并只报告了显著的那个？ ​

问题 5：你的问卷/访谈数据是否获得了被调查者的知情同意？ ​

问题 6：你借用的量表是否注明了出处和原始信效度证据？ ​

问题 7：你的抽样方式是否如实描述（非概率抽样没有说成随机抽样）？ ​

问题 8：你的理论框架图中每条箭头是否都有文献支撑？ ​

问题 9：你的 data/raw 文件夹是否保持只读（从未手动修改原始数据）？ ​

问题 10：你的 AI 使用声明是否涵盖开题、投稿、答辩三个版本？ ​

自查结果记录表 ​

English Edition ​

Appendix D · Ten Integrity Self-Check Questions ​

Question 1: Is any part of your paper entirely AI-generated without disclosure? ​

Question 2: Has every reference you cite been verified in Zotero or CNKI as actually existing? ​

Question 3: Can every regression coefficient and statistic in your paper be reproduced from your do-file? ​

Question 4: Did you try several specifications and report only the significant one? ​

Question 5: Did you obtain informed consent from respondents for your questionnaire or interview data? ​

Question 6: Do the scales you borrowed state their source and original reliability/validity evidence? ​

Question 7: Is your sampling method described truthfully (no non-probability sample presented as random)? ​

Question 8: Is every arrow in your theoretical framework diagram supported by the literature? ​

Question 9: Has your data/raw folder stayed read-only (never manually editing raw data)? ​

Question 10: Do your AI-use disclosures cover the proposal, submission, and defence versions? ​

Self-check record sheet ​