Humanize Academic Writing

Humanize Academic Writing for Social Sciences

Academic Integrity Statement

Purpose: This skill helps researchers improve the quality and naturalness of their own original ideas expressed through AI-assisted writing tools.

Ethical Use:

• ✅ Revising AI-drafted text based on your own research and ideas
• ✅ Improving writing quality for non-native English speakers
• ✅ Learning better academic writing patterns
• ❌ Using AI to generate ideas you don't understand
• ❌ Submitting work that doesn't represent your intellectual contribution

Principle: The goal is authentic scholarly communication, not deception.

---

Target Audience

Non-native English speakers in social sciences (sociology, anthropology, political science, education, psychology) who:

• Have original ideas and research
• Used AI tools to draft their text
• Need to humanize the writing style
• Want to reduce obvious AI patterns

---

When to Use This Skill

• User has AI-generated draft based on their own ideas
• Text feels "too perfect," mechanical, or repetitive
• Need to reduce AI detection markers
• Want authentic academic voice for social science writing
• Paragraph transitions feel robotic
• Language is overly abstract without concrete examples

---

Core Workflow

Step 1: Analyze the Text

First, run the AI detection analyzer to identify problematic patterns:

python scripts/ai_detector.py input.txt

The analyzer identifies:

• Repetitive sentence structures and lengths
• Overused AI transition phrases (Moreover, Furthermore, Additionally)
• Abstract/vague language patterns ("various aspects", "in terms of")
• Mechanical paragraph transitions
• Unnatural word choices for social sciences
• Low vocabulary diversity (Type-Token Ratio)
• Excessive passive voice
• Consecutive sentence similarity

Output: AI probability score + specific issues marked per paragraph

Step 2: Apply Targeted Rewriting Strategies

Based on detected issues, apply these fixes:

Strategy 1: Vary Sentence Rhythm (Fix Uniformity)

AI Pattern: All sentences are similar length (15-20 words)

Human Fix: Mix short (5-10), medium (15-20), and long (25-35) sentences

Example:

• AI: "This study examines social media impact. The research focuses on young adults. The analysis considers multiple factors."
• Human: "This study examines social media's impact on young adults, considering factors ranging from identity formation to civic engagement."

Strategy 2: Reduce Abstract Scaffolding

AI Pattern: Vague placeholder phrases that say little

Common culprits:

• "various aspects"
• "in terms of"
• "it is important to note that"
• "multiple factors"
• "different perspectives"

Human Fix: Replace with specific concepts, named theories, concrete examples

Example:

• AI: "In terms of the various aspects of social interaction, multiple factors play important roles."
• Human: "Social interaction depends on trust, reciprocity, and shared norms—factors that vary across cultural contexts."

Strategy 3: Eliminate Mechanical Transitions

AI Pattern: Overusing formal connectors at sentence starts

Overused words:

• Moreover,
• Furthermore,
• Additionally,
• In addition,
• It is important to note that

Human Fix: Use diverse transition strategies:

• Direct logical flow (no connector needed)
• "This pattern echoes..."
• "Building on this insight..."
• "Yet" / "Still" / "However" (sparingly)
• Implicit connections through content

Strategy 4: Add Scholarly Voice

AI Pattern: Generic academic tone without personality or critical engagement

Human Fix:

• Include appropriate hedging ("may suggest", "appears to", "potentially")
• Show critical engagement with sources
• Use disciplinary language naturally
• Demonstrate genuine intellectual grappling

Example:

• AI: "The data shows a correlation between X and Y."
• Human: "The data suggest a correlation between X and Y, though the causal mechanism remains unclear and warrants further investigation."

Strategy 5: Ground in Specificity

AI Pattern: Generic statements without grounding

Human Fix:

• Name specific theories/scholars
• Include concrete examples
• Reference particular contexts
• Cite actual studies with details

Example:

• AI: "Research has shown various effects of social media on society."
• Human: "Recent ethnographic work documents how Instagram reshapes young women's body image practices (Tiidenberg 2018), while experimental studies reveal minimal effects on political polarization (Guess et al. 2023)."

Step 3: Rewrite with Rationale

For each paragraph, follow this format:

Original (AI-generated): [Paste the original text]

Revised (Humanized): [Your rewritten version]

Rationale: Explain in 1-2 sentences what AI patterns you fixed. Examples:

• "Removed repetitive 'Moreover/Additionally' transitions and varied sentence rhythm (added one short sentence, one long); replaced 'various aspects' with specific concepts (trust, reciprocity, norms)."
• "Eliminated abstract scaffolding ('in terms of', 'multiple factors'); added concrete citation (Smith 2022) and specific research finding; included scholarly hedging ('suggests' rather than 'shows')."
• "Broke uniform 18-word sentences into varied lengths (8, 24, 15 words); removed mechanical 'Furthermore' openers; grounded claims in named theory (social capital) and specific context (urban China)."

---

Key Principles for Humanizing Text

1. Perplexity (Unpredictability)

• Problem: AI text is too predictable
• Fix: Add unexpected (but academically appropriate) word choices; vary syntactic structures

2. Burstiness (Rhythm Variation)

• Problem: AI uses uniform sentence lengths
• Fix: Mix short punchy sentences with longer complex ones; create natural reading rhythm

3. Specificity over Abstraction

• Problem: AI defaults to vague abstractions
• Fix: Use concrete examples, specific data, named theories; ground claims in particular contexts

4. Authentic Academic Voice

• Problem: Generic formal tone without personality
• Fix: Show genuine engagement with ideas; include appropriate hedging; demonstrate critical thinking

5. Natural Flow

• Problem: Mechanical transitions and paragraph connections
• Fix: Let content drive connections; use implicit logic; minimize formal connectors

---

Social Science Specifics

Disciplinary Language

Sociology:

• Key concepts: stratification, agency, habitus, capital, institutions, inequality
• Theoretical traditions: functionalist, conflict, symbolic interactionist, practice theory
• Common methods: ethnography, surveys, interviews, archival analysis

Anthropology:

• Key concepts: culture, ritual, kinship, liminality, positionality, thick description
• More reflexive voice acceptable
• Ethnographic detail valued

Political Science:

• Key concepts: institutions, power, legitimacy, governance, state capacity
• Causal inference language
• Hypothesis testing frameworks

Education:

• Key concepts: pedagogy, curriculum, equity, achievement gaps, learning outcomes
• Mixed methods common
• Policy relevance emphasized

Psychology (Social):

• Key concepts: cognition, behavior, attitudes, interventions, mechanisms
• Operational definitions critical
• Experimental designs prominent

Non-Native Speaker Considerations

Common AI Crutches: 1. Over-reliance on intensifiers ("very", "really", "quite") 2. Repetitive sentence starters 3. Overuse of formal connectors to signal logic

Strengths to Preserve:

• Clear logical structure (maintain this)
• Formal register (appropriate for academic writing)
• Careful grammar (don't over-casualize)

Areas to Humanize:

• Vary clause structures and sentence types
• Use field-specific terminology confidently
• Add appropriate scholarly hedging
• Include critical engagement with sources
• Ground abstractions in concrete examples

---

Additional Resources

For detailed guidance, see:

• [docs/rewriting-principles.md](docs/rewriting-principles.md): Comprehensive rewriting techniques with extended examples
• [docs/examples.md](docs/examples.md): Full before/after rewrites of different section types (intro, methods, findings, discussion)
• [docs/social-science-patterns.md](docs/social-science-patterns.md): Discipline-specific conventions and terminology

---

Scripts and Tools

ai_detector.py

Analyzes text for AI patterns and provides detailed scoring

# Basic analysis
python scripts/ai_detector.py input.txt

# Detailed output with paragraph-by-paragraph breakdown
python scripts/ai_detector.py input.txt --detailed

# JSON output for programmatic use
python scripts/ai_detector.py input.txt --json > analysis.json

text_analyzer.py

Provides quantitative metrics on text quality

# Analyze text metrics
python scripts/text_analyzer.py input.txt

# Compare before/after versions
python scripts/text_analyzer.py original.txt revised.txt --compare

Metrics provided:

• Sentence length distribution and variance
• Vocabulary diversity (Type-Token Ratio)
• Academic word usage frequency
• Transition word density
• Passive voice percentage
• Average sentence complexity

---

Example Workflow

1. User provides AI-generated text: "Can you help humanize this paragraph from my paper?"

2. Analyze first:

• Run ai_detector.py or manually identify patterns
• Note specific issues (e.g., "repetitive sentence structure, 3x 'Moreover', abstract language")

3. Rewrite strategically:

• Apply relevant strategies from above
• Maintain the user's core ideas and arguments
• Preserve accurate citations and data

4. Explain changes:

• Show original → revised
• Provide rationale explaining what AI patterns were fixed
• Help user learn for future writing

5. Verify improvements:

• Optionally run text_analyzer.py to confirm metrics improved
• Check that meaning and accuracy preserved

---

Tips for Effective Use

Do:

• ✅ Preserve the user's original ideas and arguments
• ✅ Maintain citation accuracy
• ✅ Keep the appropriate academic register
• ✅ Focus on patterns, not just individual words
• ✅ Explain your changes so users learn

Don't:

• ❌ Change the meaning or argument
• ❌ Add information not in the original
• ❌ Over-casualize academic language
• ❌ Remove all formal connectors (some are needed)
• ❌ Make text deliberately grammatically incorrect

Balance:

Academic writing should be:

• Clear but not simplistic
• Formal but not robotic
• Structured but not mechanical
• Precise but not pedantic

---

Common Pitfalls to Avoid

1. Over-correcting: Don't make every sentence wildly different in length. Natural variation exists within a range.

2. Removing all connectors: Some transitions are necessary for clarity, especially in complex arguments.

3. Adding colloquialisms: Academic writing should remain formal; avoid casual expressions.

4. Losing precision: Don't sacrifice technical accuracy for "naturalness."

5. Ignoring discipline: Social science subfields have different conventions—respect them.

---

Summary Checklist

After rewriting, verify:

• [ ] Sentence lengths vary (mix of short, medium, long)
• [ ] Mechanical transitions (Moreover, Furthermore, Additionally) removed or reduced
• [ ] Abstract placeholder phrases replaced with specific concepts
• [ ] At least one concrete example or named theory added
• [ ] Scholarly hedging included where appropriate
• [ ] Original meaning and arguments preserved
• [ ] Citations remain accurate
• [ ] Disciplinary language sounds natural
• [ ] Rationale provided explaining AI patterns fixed

---

This skill emphasizes authentic scholarly communication while respecting the intellectual work of non-native English speakers using AI tools responsibly.

🌏 双语文档更新完成！

✅ 修正完成

问题识别

之前的中文文档将示例翻译成了中文，但实际使用场景是：

• 用户：中文母语的社科研究者
• 目标：撰写英文学术论文
• 需求：改进AI生成的英文文本

修正方案

所有中文文档现在采用：

• ✅ 用中文讲解原则和策略
• ✅ 示例保持英文（改写前和改写后）
• ✅ 用中文解释问题和修改理由

---

📝 已更新的文件（4个）

1. docs/examples_CN.md

• ✅ 所有示例文本改为英文
• ✅ 问题分析用中文
• ✅ 修改理由用中文
• ✅ 添加了说明：示例是英文，解释是中文

2. docs/rewriting-principles_CN.md

• ✅ 所有策略示例改为英文
• ✅ 原则讲解用中文
• ✅ 添加了说明

3. docs/social-science-patterns_CN.md

• ✅ 所有学科示例改为英文
• ✅ 学科特点讲解用中文
• ✅ 添加了说明

4. README_CN.md & QUICKSTART_CN.md

• ✅ 添加了顶部说明，明确工具用途

---

📋 现在的文档结构

英文文档（面向英文用户）

• README.md - 英文说明
• QUICKSTART.md - 英文快速开始
• docs/rewriting-principles.md - 英文原则，英文示例
• docs/examples.md - 英文示例，英文解释
• docs/social-science-patterns.md - 英文指导，英文示例

中文文档（面向中文用户写英文论文）

• README_CN.md - 中文说明
• QUICKSTART_CN.md - 中文快速开始
• docs/rewriting-principles_CN.md - 中文讲解原则，英文示例
• docs/examples_CN.md - 英文示例，中文解释
• docs/social-science-patterns_CN.md - 中文讲解学科特点，英文示例

---

🎯 用户体验

中文用户（非英语母语学者）

1. 阅读 README_CN.md

• 看到明确说明：工具用于改进英文文本
• 用中文理解功能和用途

2. 查看 docs/examples_CN.md

• 看到英文的AI生成示例
• 用中文理解有哪些问题
• 看到英文的人性化版本
• 用中文理解如何修改的

3. 学习 docs/rewriting-principles_CN.md

• 用中文学习改写策略
• 看英文示例对比
• 用中文理解原理

4. 实际使用

• 把自己AI生成的英文文本交给Cursor
• Cursor按照技能改写成自然的英文
• 理解修改理由

---

🔄 更新到GitHub

cd "c:\Paper Project\AI DOC\First paper\humanize-academic-writing"

# 查看更改
git status

# 添加所有更改
git add .

# 提交
git commit -m "Fix Chinese documentation: Keep examples in English with Chinese explanations

- Update docs/examples_CN.md: English examples with Chinese rationales
- Update docs/rewriting-principles_CN.md: English examples with Chinese principles
- Update docs/social-science-patterns_CN.md: English examples with Chinese guidance
- Add clarifying notes to all CN files explaining bilingual approach
- Purpose: Help Chinese-speaking researchers improve English academic writing"

# 推送
git push

---

✨ 现在的优势

对中文用户

• ✅ 能用母语理解原则
• ✅ 能看到实际的英文示例（不是翻译）
• ✅ 能学习如何改进英文写作
• ✅ 不会因为中文翻译失去英文的细微差别

对英文用户

• ✅ 完整的英文文档
• ✅ 可以切换到中文版本（如果需要）

---

📊 文档统计（更新后）

---

🎉 完成！

现在中文文档正确地： 1. 用中文教学改写原则 2. 用英文展示实际示例 3. 用中文解释修改理由

这样中文用户可以：

• 用母语理解概念
• 看到真实的英文学术写作
• 学习如何改进自己的英文论文

立即更新到GitHub！

Contributing to Humanize Academic Writing

First off, thank you for considering contributing to this project! This skill aims to help researchers communicate their ideas more clearly and authentically.

Code of Conduct

Our Pledge

We are committed to providing a welcoming and inclusive environment for all contributors, regardless of:

• Experience level
• Native language
• Academic discipline
• Institution or career stage

Expected Behavior

• Be respectful and constructive in feedback
• Focus on improving the tool, not criticizing users
• Acknowledge that academic writing conventions vary across fields and cultures
• Maintain academic integrity principles

Unacceptable Behavior

• Encouraging academic dishonesty
• Discriminating against non-native English speakers
• Dismissing legitimate disciplinary differences in writing style

How to Contribute

Areas Where We Need Help

1. Discipline-Specific Guidance

• Economics, geography, communication studies, etc.
• Field-specific writing examples
• Conventions for different subfields

2. Language and Cultural Perspectives

• Non-English academic writing patterns
• Cultural differences in academic discourse
• Challenges specific to different language backgrounds

3. Technical Improvements

• Enhanced AI detection algorithms
• Better text analysis metrics
• Visualization tools
• Performance optimizations

4. Documentation

• More example transformations
• Clarifying existing documentation
• Translations (if expanding beyond English)

Getting Started

1. Fork the repository

   git clone https://github.com/yourusername/humanize-academic-writing.git
   cd humanize-academic-writing

2. Create a branch

   git checkout -b feature/your-feature-name

3. Make your changes

• Follow existing documentation structure
• Include examples where helpful
• Test scripts if modifying code

4. Commit with clear messages

   git commit -m "Add economics writing patterns to social-science-patterns.md"

5. Push and create Pull Request

   git push origin feature/your-feature-name

Contribution Guidelines

For Documentation Contributions

Adding Examples

• Provide both "Before (AI)" and "After (Human)" versions
• Include a "Rationale" explaining what was fixed
• Use realistic academic text (or create plausible examples)
• Cite real sources when possible

Adding Discipline-Specific Guidance

• Identify key concepts and vocabulary for the field
• Note writing style conventions (voice, citation style, structure)
• Provide 2-3 paragraph-length examples
• Cite classic works in the field

For Code Contributions

Python Scripts

• Follow PEP 8 style guidelines
• Include docstrings for functions and classes
• Add comments for complex logic
• Ensure code works with Python 3.7+
• Use only standard library (avoid external dependencies unless necessary)

Testing

• Test with various text lengths (100 words to 5,000 words)
• Include edge cases (very short sentences, no punctuation, etc.)
• Verify output format is user-friendly

Style Preferences

Documentation

• Use clear, accessible language
• Include concrete examples over abstract explanations
• Format code blocks with proper syntax highlighting
• Use headings to create scannable structure

Examples

• Academic but not jargon-heavy
• Realistic scenarios researchers face
• Show genuine improvement, not minor tweaks
• Explain why changes improve the text

Pull Request Process

1. Update documentation if you've changed functionality 2. Add your contribution to the relevant section of README.md 3. Ensure the PR description clearly explains:

• What you're adding/fixing
• Why it's needed
• How you've tested it (if code)

4. Request review from maintainers 5. Address feedback constructively

Academic Integrity Reminder

All contributions should support ethical use of AI in academic writing:

✅ Helping researchers express their own ideas clearly ✅ Improving writing quality and naturalness ✅ Teaching better academic writing practices

❌ Not helping people disguise plagiarism ❌ Not encouraging fabrication of research ❌ Not promoting academic dishonesty

If a contribution could be misused to support academic misconduct, it won't be accepted.

Recognition

Contributors will be acknowledged in:

• README.md (Contributors section to be added)
• Release notes for significant contributions
• Academic citation if project leads to publication

Questions?

• Open an issue for clarification
• Start a discussion in GitHub Discussions
• Email maintainers (see README)

License

By contributing, you agree that your contributions will be licensed under the MIT License.

---

Thank you for helping make academic writing clearer and more authentic!

改写示例：改写前后对比

🇬🇧 English | 🇨🇳 中文

本文档提供AI生成的英文学术文本转化为自然学术写作的完整示例，用中文详细解释修改理由。

📝 说明：所有示例文本都是英文（包括原文和改写版本），但问题分析和修改理由用中文解释，帮助非英语母语的中文研究者理解如何改进英文学术写作。

---

示例1：引言段落

原文（AI生成的英文）

Social media has become an important aspect of modern communication. Moreover, it has various effects on society. Additionally, researchers have studied this phenomenon extensively. Furthermore, the findings suggest multiple implications for understanding digital interaction. In addition, these platforms play a crucial role in political discourse. It is important to note that different perspectives exist on this topic.

发现的问题（中文解释）

• ❌ 句子长度均匀（9-13词）
• ❌ 机械过渡词（Moreover, Additionally, Furthermore, In addition）
• ❌ 抽象语言（"various effects", "multiple implications", "crucial role"）
• ❌ 重复句式结构（主-谓模式）
• ❌ 没有具体内容的通用主张

修订版（人性化的英文）

Social media platforms have fundamentally reshaped how individuals communicate, mobilize, and access information. While scholars extensively document these transformations (boyd 2014; van Dijck 2013), debates persist about their democratic implications—particularly regarding echo chambers and polarization (Sunstein 2017). Do platforms amplify pre-existing divisions, or do they create new forms of fragmentation? Drawing on panel data from three national surveys, this study examines...

修改理由（中文解释）

修复了：(1) 消除所有机械过渡词；(2) 变化句子节奏（12, 19, 15, 9词）；(3) 用具体概念替换抽象（"various effects" → echo chambers, polarization）；(4) 添加具体引用；(5) 包含修辞性问题以显示智力参与；(6) 混合陈述句和疑问句；(7) 使用破折号进行阐述。

---

示例2：文献综述部分

原文（AI生成的英文）

Previous research has examined social media effects on political engagement. Various studies have found different results. Smith (2020) discusses polarization. Moreover, Jones (2021) explores echo chambers. Additionally, Williams (2019) investigates filter bubbles. Furthermore, the literature suggests multiple factors are important. In terms of methodology, different approaches have been used. It is important to note that gaps exist in current research.

发现的问题（中文解释）

• ❌ 通用的"research shows"语言
• ❌ 全篇机械过渡
• ❌ 浅层引用整合（"discusses", "explores"）
• ❌ 模糊短语（"multiple factors", "different approaches"）
• ❌ 缺乏对来源的批判性参与
• ❌ 缺少综合或智力进展

修订版（人性化的英文）

Scholarship on social media's political effects divides into two camps. Optimists argue platforms democratize participation by lowering mobilization costs (Shirky 2008; Castells 2012), enabling movements from Occupy to Arab Spring. Skeptics counter that online activism rarely translates offline (Morozov 2011), creating "slacktivism" that substitutes symbolic gestures for sustained engagement. Recent empirical work complicates both narratives: Tufekci's (2017) ethnographic study reveals how Twitter simultaneously empowered and constrained Egyptian activists, while Margetts et al.'s (2016) behavioral data show online mobilization follows power-law distributions—most campaigns fail, but successful ones scale rapidly.

>

Yet this literature overlooks class dynamics. Existing studies disproportionately sample educated urban users, leaving rural and working-class experiences unexamined (notable exception: Schradie 2019). This gap matters theoretically...

修改理由（中文解释）

修复了：(1) 将文献结构化为竞争框架（乐观派 vs. 怀疑派）；(2) 学者作为有论证的行动者出现，而非通用的"discusses"；(3) 陈述具体主张（例如"lowering mobilization costs", "power-law distributions"）；(4) 添加批判性综合（"complicates both narratives"）；(5) 用理论理由识别研究空白；(6) 句子长度变化极大（9-28词）；(7) 使用冒号和破折号进行阐述；(8) 展示智力进展，引导研究贡献。

---

示例3：方法部分

原文（AI生成的英文）

This study uses qualitative methods. The research involves interviews with participants. Data was collected in several locations. Various themes emerged from the analysis. The findings were analyzed using thematic analysis. It is important to note that ethical considerations were addressed. Additionally, the study ensured participant confidentiality.

发现的问题（中文解释）

• ❌ 模糊方法论（"qualitative methods"）
• ❌ 过度被动语态
• ❌ 缺乏具体性（多少访谈？哪些地点？什么主题？）
• ❌ 通用短语（"various themes", "ethical considerations"）
• ❌ 机械的"Additionally"

修订版（人性化的英文）

Between January and July 2023, I conducted 42 semi-structured interviews with factory workers in Shenzhen, Dongguan, and Guangzhou—China's Pearl River Delta manufacturing hub. Participants, recruited through snowball sampling via labor NGO networks, ranged from 22 to 54 years old (median: 31), with 60% women reflecting the sector's gender composition.

>

Interviews lasted 45-90 minutes, conducted in Mandarin and recorded with consent. I employed Braun and Clarke's (2006) six-phase thematic analysis, coding inductively while remaining attentive to power dynamics in researcher-participant interactions (England 1994). Three dominant themes emerged: economic precarity, familial obligation, and contested notions of dignity. Each reveals tensions between workers' aspirations and structural constraints.

>

The university IRB approved all procedures. Participants received ¥100 compensation and chose pseudonyms appearing below.

修改理由（中文解释）

修复了：(1) 指定时间框架、样本量、地点及地理背景；(2) 详细说明招募方法和参与者人口统计；(3) 混合主动（"I conducted", "I employed"）和被动语态策略；(4) 命名具体分析框架（Braun & Clarke）；(5) 展示对权力动力学的反思性；(6) 具体命名三个主题；(7) 将伦理考虑自然整合到叙述中而非附加；(8) 句子结构变化极大；(9) 使用破折号进行地理说明。

---

示例4：研究发现部分

原文（AI生成的英文）

The data shows several important patterns. Moreover, participants expressed various concerns. Additionally, different themes emerged from the interviews. Furthermore, the findings suggest that multiple factors are relevant. In terms of experiences, participants described diverse situations. It is important to note that these results have significant implications. The analysis reveals that the phenomenon is complex.

发现的问题（中文解释）

• ❌ "data shows"（应该是"suggest"）
• ❌ 全部六个机械过渡
• ❌ 完全抽象（"various concerns", "different themes", "multiple factors"）
• ❌ 没有实际发现呈现
• ❌ 通用结论（"phenomenon is complex"）
• ❌ 缺少参与者声音或具体例子

修订版（人性化的英文）

Workers articulated a paradox: they needed WeChat to find jobs yet resented its intrusion into off-hours. Li, a 28-year-old migrant from Hunan, explained: "The boss adds you on WeChat. Then he sends work messages at 10 PM. You can't ignore him—he'll remember." This pattern appeared across 34 of 42 interviews, cutting across age and gender.

>

But responses varied by class position. Line workers passively complied, describing WeChat as "延长的工作" (extended work). Team leaders, however, leveraged platform affordances to assert autonomy—creating separate work and personal accounts, using status updates to signal unavailability. Wang, a 35-year-old team leader, strategically: "I post family photos at 9 PM. Shows I'm off-duty."

>

These micro-resistance tactics suggest agency within constraint. While scholars emphasize platforms' disciplinary functions (Fuchs 2014), workers actively negotiate boundaries—though success correlates with organizational power.

修改理由（中文解释）

修复了：(1) 消除所有抽象，以具体发现（悖论）开头；(2) 整合带直接引用的参与者声音；(3) 量化模式（34/42）；(4) 展示组内差异（line workers vs. team leaders）；(5) 包括工人实际使用的中文术语（"延长的工作"）；(6) 将实证发现连接到理论辩论（Fuchs）；(7) 提供细致结论（约束内的能动性，成功因权力而异）；(8) 句子节奏变化极大；(9) 使用冒号进行阐述；(10) 删除所有机械过渡。

---

示例5：讨论部分

原文（AI生成的英文）

These findings have important implications. Moreover, they contribute to existing literature. Additionally, the results suggest several theoretical points. Furthermore, this research demonstrates that the topic is significant. In terms of practical implications, various applications are possible. It is important to note that limitations exist. Future research should examine additional aspects.

发现的问题（中文解释）

• ❌ 完全抽象——没有陈述实际意义
• ❌ 所有机械过渡
• ❌ 全篇模糊短语（"several theoretical points", "various applications"）
• ❌ 没有具体内容的通用局限性提及
• ❌ 没有阐明智力综合或贡献

修订版（人性化的英文）

These findings challenge platform determinism. Rather than treating technology as imposing uniform effects, we see how organizational context mediates digital tools' meanings and uses. WeChat operates differently for line workers (disciplinary surveillance) versus team leaders (negotiation resource). This variation matters theoretically: scholars debating whether platforms empower or exploit workers (Scholz 2016; Fuchs 2014) overlook how positional power within existing hierarchies shapes platform politics.

>

The implications extend beyond workplace surveillance. If platform effects depend on users' structural positions, then interventions targeting "design" alone—algorithmic transparency, user controls—may prove insufficient. Power operates through, not just on, digital infrastructure.

>

Three limitations warrant mention. First, this study examines only manufacturing; service sector dynamics may differ. Second, focusing on WeChat excludes workers using alternative platforms (QQ, enterprise software). Finally, interviews capture reported behavior rather than observed practices—future research should employ workplace ethnography to triangulate findings.

修改理由（中文解释）

修复了：(1) 以清晰的理论贡献开头（"challenge platform determinism"）；(2) 具体说明发现如何推进辩论（命名学者，描述文献中的疏忽）；(3) 具体阐述实际意义（仅设计干预不够）；(4) 提供具体、实质性的局限性及理由；(5) 建议具体的未来方向（workplace ethnography）；(6) 删除所有机械过渡；(7) 使用破折号和斜体强调；(8) 变化句子结构；(9) 展示将实证发现连接到理论意义的真正智力综合。

---

示例6：摘要

原文（AI生成的英文）

This study examines social media use in workplace contexts. The research employs qualitative methods to investigate various aspects of digital communication. The findings reveal several important patterns. Moreover, the results contribute to existing literature. Additionally, practical implications are discussed. This research demonstrates that the topic is significant for understanding modern work environments.

发现的问题（中文解释）

• ❌ 全篇模糊（"various aspects", "several patterns"）
• ❌ 摘要中的机械过渡（不适当）
• ❌ 没有陈述具体发现或贡献
• ❌ 缺乏关键细节（哪里？谁？到底是什么？）
• ❌ 通用重要性主张

修订版（人性化的英文）

How do workplace surveillance technologies shape labor relations in China's manufacturing sector? Drawing on 42 interviews with factory workers in the Pearl River Delta, this study examines WeChat's dual role as job-search tool and managerial surveillance mechanism. Findings reveal a paradox: workers depend on platforms for employment yet resent their intrusion into personal time. Responses vary by organizational position—line workers passively comply while team leaders actively negotiate boundaries. These patterns challenge platform determinism, showing how existing power hierarchies mediate digital tools' meanings and effects. The findings suggest interventions targeting platform design alone cannot address surveillance absent structural workplace reform.

修改理由（中文解释）

修复了：(1) 以研究问题开头；(2) 指定方法、样本、地点；(3) 命名平台及其双重角色；(4) 陈述悖论（关键发现）；(5) 描述变化模式；(6) 阐述理论贡献；(7) 指出政策含义；(8) 全部8句，节奏变化；(9) 没有机械过渡；(10) 每句传达实质性信息。

---

示例7：引用密集段落

原文（AI生成的英文）

According to previous research, social media affects political behavior (Smith 2020). Studies have shown various effects (Jones 2021; Williams 2019). Moreover, scholars have noted important patterns (Brown 2022). Additionally, research demonstrates significant findings (Davis 2023). The literature suggests that multiple factors are involved (Miller 2021).

发现的问题（中文解释）

• ❌ 通用"According to research"框架
• ❌ 引用作为装饰而非实质
• ❌ 没有来自被引用作品的实际主张
• ❌ 机械过渡
• ❌ 全篇模糊（"various effects", "important patterns"）

修订版（人性化的英文）

Experimental studies consistently find that corrective information reduces misperceptions—by 8-15% in most samples (Nyhan & Reifler 2015; Porter & Wood 2019). Yet field effects prove elusive. When Facebook partnered with fact-checkers in 2017, flagged content circulated almost unchanged (Pennycook et al. 2018). Smith's (2020) panel study offers an explanation: corrections work for low-salience issues but backfire for identity-relevant topics, actually strengthening misperceptions among partisan respondents. This "backfire effect" remains contested—recent meta-analysis by Wood and Porter (2023) finds it rare—but the broader pattern holds: information alone cannot overcome motivated reasoning.

修改理由（中文解释）

修复了：(1) 学者成为提出具体主张的行动者；(2) 包含量化发现（8-15%，2017年）；(3) 命名具体干预（Facebook事实核查合作）；(4) 解释机制（"backfire effect"定义）；(5) 展示学术辩论（有争议的发现，元分析限定）；(6) 综合到更广泛的理论要点（motivated reasoning）；(7) 删除通用过渡词；(8) 实质性而非装饰性地整合7个引用；(9) 使用斜体强调；(10) 变化句子结构。

---

所有示例的共同模式

删除的内容

• 机械过渡（Moreover, Furthermore, Additionally）
• 抽象占位（"various aspects", "multiple factors"）
• 通用归属（"research shows", "studies have found"）
• 公式化短语（"It is important to note"）
• 均匀句子节奏

添加的内容

• 具体概念、理论、名称
• 相关的定量细节
• 参与者声音（在定性部分）
• 对来源的批判性参与
• 变化的句子结构
• 学术限定语和细致差别
• 具体例子和背景
• 真正的智力综合

一致的策略

1. 具体性：用具体细节替换每个抽象 2. 能动性：学者/参与者作为主动行动者，而非被动的"research" 3. 综合：显示想法之间的连接，而非仅仅列表 4. 声音：包括适当的限定语和批判性参与 5. 节奏：刻意变化句子长度和结构 6. 流畅：让内容创造连接，最小化机械过渡

---

学科特定示例（英文示例，中文解释）

社会学示例

AI生成的英文：

Social class plays an important role in educational outcomes. Moreover, various factors mediate this relationship.

人性化的英文：

Bourdieu's (1984) concept of cultural capital explains how middle-class families transmit educational advantages: not through income alone, but via embodied competencies, linguistic styles, and institutional know-how that schools reward. Lareau's (2003) ethnography specifies the mechanism—'concerted cultivation' parenting practices that align with teachers' expectations.

中文解释：用布迪厄的具体理论概念替换了"plays an important role"；命名了机制（文化资本的具体形式）；添加了Lareau的民族志来说明机制的运作方式。

人类学示例

AI生成的英文：

The ritual has various meanings for participants. Additionally, it serves multiple functions in the community.

人性化的英文：

The funeral ritual does not simply mourn the dead—it reconstitutes social order. When a Toraja elder dies, the village assembles for ceremonies lasting weeks, slaughtering dozens of buffalo in displays that affirm kinship networks and redistribute wealth (Volkman 1985). Participation is obligatory; absence signals social death.

中文解释：用具体的民族志细节（Toraja长者葬礼，宰杀水牛）替换了"various meanings"；说明了仪式的具体功能（确认亲属网络，重新分配财富）；包含了后果（缺席=社会死亡）。

政治学示例

AI生成的英文：

Institutions affect policy outcomes. Moreover, different types of institutions have various effects.

人性化的英文：

Electoral systems shape party fragmentation: proportional representation (PR) enables minor parties while first-past-the-post (FPTP) favors two-party systems (Duverger 1954). Recent panel data across 63 democracies confirms this pattern: PR countries average 4.2 parliamentary parties versus 2.7 in FPTP systems (Golder 2005).

中文解释：用具体的制度类型（PR vs. FPTP）替换了"institutions"；提供了量化证据（4.2 vs. 2.7个政党）；命名了经典理论（Duverger定律）和实证研究（Golder）。

---

自检问题（中文）

改写后，问自己：

1. 具体性：我是否用具体细节替换了每个"various/multiple/different"？ 2. 节奏：我的句子长度是否显著变化？ 3. 过渡：我是否消除了机械的"Moreover/Furthermore"？ 4. 声音：我是否包括适当的限定语和学术细致差别？ 5. 引用：学者/研究是否实质性整合，而非仅仅列出？ 6. 例子：我是否将抽象主张植根于具体实例？ 7. 综合：想法是否通过内容逻辑连接，而非仅仅正式连接词？ 8. 真实性：它听起来像有思想的人类学者思考，还是机器？

如果对任何问题回答"否"，进一步修订。

---

记住： 目标是真实的学术交流——展示对你主题的真正智力参与，而非愚弄检测软件。你的修订应该澄清和加强论证，而非掩盖它们。

更多指导请参见 rewriting-principles_CN.md。

Rewriting Examples: Before and After

This document provides complete examples of AI-generated text transformed into natural academic writing, with detailed rationales.

---

Example 1: Introduction Paragraph

Original (AI-generated)

Social media has become an important aspect of modern communication. Moreover, it has various effects on society. Additionally, researchers have studied this phenomenon extensively. Furthermore, the findings suggest multiple implications for understanding digital interaction. In addition, these platforms play a crucial role in political discourse. It is important to note that different perspectives exist on this topic.

Issues Identified

• ❌ Uniform sentence length (9-13 words each)
• ❌ Mechanical transitions (Moreover, Additionally, Furthermore, In addition)
• ❌ Abstract language ("various effects," "multiple implications," "crucial role")
• ❌ Repetitive sentence structure (Subject-Verb pattern)
• ❌ Generic claims without specifics

Revised (Humanized)

Social media platforms have fundamentally reshaped how individuals communicate, mobilize, and access information. While scholars extensively document these transformations (boyd 2014; van Dijck 2013), debates persist about their democratic implications—particularly regarding echo chambers and polarization (Sunstein 2017). Do platforms amplify pre-existing divisions, or do they create new forms of fragmentation? Drawing on panel data from three national surveys, this study examines...

Rationale

Fixed: (1) Eliminated all mechanical transitions; (2) Varied sentence rhythm (12, 19, 15, 9 words); (3) Replaced abstractions ("various effects") with specific concepts (echo chambers, polarization); (4) Added concrete citations; (5) Included rhetorical question to show intellectual engagement; (6) Mixed declarative and interrogative sentences; (7) Used dash for elaboration.

---

Example 2: Literature Review Section

Original (AI-generated)

Previous research has examined social media effects on political engagement. Various studies have found different results. Smith (2020) discusses polarization. Moreover, Jones (2021) explores echo chambers. Additionally, Williams (2019) investigates filter bubbles. Furthermore, the literature suggests multiple factors are important. In terms of methodology, different approaches have been used. It is important to note that gaps exist in current research.

Issues Identified

• ❌ Generic "research shows" language
• ❌ Mechanical transitions throughout
• ❌ Shallow citation integration ("discusses," "explores")
• ❌ Vague phrases ("multiple factors," "different approaches")
• ❌ No critical engagement with sources
• ❌ Lacks synthesis or intellectual progression

Revised (Humanized)

Scholarship on social media's political effects divides into two camps. Optimists argue platforms democratize participation by lowering mobilization costs (Shirky 2008; Castells 2012), enabling movements from Occupy to Arab Spring. Skeptics counter that online activism rarely translates offline (Morozov 2011), creating "slacktivism" that substitutes symbolic gestures for sustained engagement. Recent empirical work complicates both narratives: Tufekci's (2017) ethnographic study reveals how Twitter simultaneously empowered and constrained Egyptian activists, while Margetts et al.'s (2016) behavioral data show online mobilization follows power-law distributions—most campaigns fail, but successful ones scale rapidly.

>

Yet this literature overlooks class dynamics. Existing studies disproportionately sample educated urban users, leaving rural and working-class experiences unexamined (notable exception: Schradie 2019). This gap matters theoretically...

Rationale

Fixed: (1) Structured literature into competing frameworks (optimists vs. skeptics); (2) Scholars appear as agents with arguments, not generic "discusses"; (3) Specific claims stated (e.g., "lowering mobilization costs," "power-law distributions"); (4) Added critical synthesis ("complicates both narratives"); (5) Identified research gap with theoretical justification; (6) Varied sentence length dramatically (9 to 28 words); (7) Used colon for elaboration; (8) Showed intellectual progression leading to study's contribution.

---

Example 3: Methods Section

Original (AI-generated)

This study uses qualitative methods. The research involves interviews with participants. Data was collected in several locations. Various themes emerged from the analysis. The findings were analyzed using thematic analysis. It is important to note that ethical considerations were addressed. Additionally, the study ensured participant confidentiality.

Issues Identified

• ❌ Vague methodology ("qualitative methods")
• ❌ Excessive passive voice
• ❌ Lacks specifics (how many interviews? which locations? what themes?)
• ❌ Generic phrases ("various themes," "ethical considerations")
• ❌ Mechanical "Additionally"

Revised (Humanized)

Between January and July 2023, I conducted 42 semi-structured interviews with factory workers in Shenzhen, Dongguan, and Guangzhou—China's Pearl River Delta manufacturing hub. Participants, recruited through snowball sampling via labor NGO networks, ranged from 22 to 54 years old (median: 31), with 60% women reflecting the sector's gender composition.

>

Interviews lasted 45-90 minutes, conducted in Mandarin and recorded with consent. I employed Braun and Clarke's (2006) six-phase thematic analysis, coding inductively while remaining attentive to power dynamics in researcher-participant interactions (England 1994). Three dominant themes emerged: economic precarity, familial obligation, and contested notions of dignity. Each reveals tensions between workers' aspirations and structural constraints.

>

The university IRB approved all procedures. Participants received ¥100 compensation and chose pseudonyms appearing below.

Rationale

Fixed: (1) Specified timeframe, sample size, locations with geographic context; (2) Detailed recruitment method and participant demographics; (3) Mixed active ("I conducted," "I employed") and passive voice strategically; (4) Named specific analytical framework (Braun & Clarke); (5) Showed reflexivity about power dynamics; (6) Named the three themes concretely; (7) Integrated ethical considerations naturally into narrative rather than tacking on; (8) Varied sentence structure dramatically; (9) Used dash for geographic specification.

---

Example 4: Findings Section

Original (AI-generated)

The data shows several important patterns. Moreover, participants expressed various concerns. Additionally, different themes emerged from the interviews. Furthermore, the findings suggest that multiple factors are relevant. In terms of experiences, participants described diverse situations. It is important to note that these results have significant implications. The analysis reveals that the phenomenon is complex.

Issues Identified

• ❌ "Data shows" (should be "suggest")
• ❌ All six mechanical transitions
• ❌ Entirely abstract ("various concerns," "different themes," "multiple factors")
• ❌ No actual findings presented
• ❌ Generic conclusion ("phenomenon is complex")
• ❌ Lacks participant voices or concrete examples

Revised (Humanized)

Workers articulated a paradox: they needed WeChat to find jobs yet resented its intrusion into off-hours. Li, a 28-year-old migrant from Hunan, explained: "The boss adds you on WeChat. Then he sends work messages at 10 PM. You can't ignore him—he'll remember." This pattern appeared across 34 of 42 interviews, cutting across age and gender.

>

But responses varied by class position. Line workers passively complied, describing WeChat as "延长的工作" (extended work). Team leaders, however, leveraged platform affordances to assert autonomy—creating separate work and personal accounts, using status updates to signal unavailability. Wang, a 35-year-old team leader, strategically: "I post family photos at 9 PM. Shows I'm off-duty."

>

These micro-resistance tactics suggest agency within constraint. While scholars emphasize platforms' disciplinary functions (Fuchs 2014), workers actively negotiate boundaries—though success correlates with organizational power.

Rationale

Fixed: (1) Eliminated all abstractions, leading with concrete finding (the paradox); (2) Integrated participant voices with direct quotations; (3) Quantified pattern (34/42); (4) Showed within-group variation (line workers vs. team leaders); (5) Included Chinese term workers actually used; (6) Connected empirical findings to theoretical debate (Fuchs); (7) Offered nuanced conclusion (agency within constraint, success varies by power); (8) Varied sentence rhythm dramatically; (9) Used colon for elaboration; (10) Removed all mechanical transitions.

---

Example 5: Discussion Section

Original (AI-generated)

These findings have important implications. Moreover, they contribute to existing literature. Additionally, the results suggest several theoretical points. Furthermore, this research demonstrates that the topic is significant. In terms of practical implications, various applications are possible. It is important to note that limitations exist. Future research should examine additional aspects.

Issues Identified

• ❌ Entirely abstract—no actual implications stated
• ❌ All mechanical transitions
• ❌ Vague phrases throughout ("several theoretical points," "various applications")
• ❌ Generic limitations mention without specifics
• ❌ No intellectual synthesis or contribution clarified

Revised (Humanized)

These findings challenge platform determinism. Rather than treating technology as imposing uniform effects, we see how organizational context mediates digital tools' meanings and uses. WeChat operates differently for line workers (disciplinary surveillance) versus team leaders (negotiation resource). This variation matters theoretically: scholars debating whether platforms empower or exploit workers (Scholz 2016; Fuchs 2014) overlook how positional power within existing hierarchies shapes platform politics.

>

The implications extend beyond workplace surveillance. If platform effects depend on users' structural positions, then interventions targeting "design" alone—algorithmic transparency, user controls—may prove insufficient. Power operates through, not just on, digital infrastructure.

>

Three limitations warrant mention. First, this study examines only manufacturing; service sector dynamics may differ. Second, focusing on WeChat excludes workers using alternative platforms (QQ, enterprise software). Finally, interviews capture reported behavior rather than observed practices—future research should employ workplace ethnography to triangulate findings.

Rationale

Fixed: (1) Led with clear theoretical contribution ("challenge platform determinism"); (2) Specified how findings advance debates (named scholars, described oversight in literature); (3) Articulated practical implications concretely (design interventions insufficient); (4) Provided specific, substantive limitations with reasoning; (5) Suggested concrete future direction (workplace ethnography); (6) Removed all mechanical transitions; (7) Used dashes and italics for emphasis; (8) Varied sentence structure; (9) Showed genuine intellectual synthesis connecting empirical findings to theoretical implications.

---

Example 6: Conclusion

Original (AI-generated)

In conclusion, this study has examined social media and workplace dynamics. The research has made several contributions. Moreover, it has addressed important gaps. Additionally, the findings provide valuable insights. Furthermore, this work demonstrates the significance of the topic. Future research should continue to explore these issues. It is important to note that more investigation is needed.

Issues Identified

• ❌ Formulaic "In conclusion" opening
• ❌ Mechanical transitions
• ❌ Entirely generic ("several contributions," "valuable insights")
• ❌ Adds no new synthesis or provocative closing thought
• ❌ Weak, generic future research call

Revised (Humanized)

Platform capitalism's contradictions play out in the everyday: workers adopt the very tools that intensify exploitation, yet their adoption is neither passive acceptance nor pure resistance. Instead, we see ongoing negotiation—shaped by organizational hierarchies, constrained by labor market precarity, enabled by technological affordances.

>

This study advances scholarship in two ways. Empirically, it documents how class position within workplaces mediates platform experiences, complicating both celebratory and dystopian accounts. Theoretically, it suggests we cannot understand digital labor without attending to intersections of technological infrastructure and preexisting power structures.

>

The stakes are high. As platforms penetrate workplaces globally, understanding who benefits—and how subordinated groups carve out autonomy—becomes essential for imagining alternative futures. Technology is never just technology. It is power, reshaped.

Rationale

Fixed: (1) Eliminated "In conclusion" cliché; (2) Opened with provocative synthesis ("contradictions play out in the everyday"); (3) Clearly stated two contributions (empirical + theoretical); (4) Used italics for emphasis on key concept (intersections); (5) Ended with stakes/significance rather than generic "more research needed"; (6) Included sentence fragment for rhetorical emphasis ("It is power, reshaped"); (7) Removed all mechanical transitions; (8) Demonstrated genuine intellectual closure while opening to broader implications.

---

Example 7: Abstract

Original (AI-generated)

This study examines social media use in workplace contexts. The research employs qualitative methods to investigate various aspects of digital communication. The findings reveal several important patterns. Moreover, the results contribute to existing literature. Additionally, practical implications are discussed. This research demonstrates that the topic is significant for understanding modern work environments.

Issues Identified

• ❌ Vague throughout ("various aspects," "several patterns")
• ❌ Mechanical transitions in abstract (inappropriate)
• ❌ No specific findings or contribution stated
• ❌ Lacks key details (where? who? what exactly?)
• ❌ Generic significance claim

Revised (Humanized)

How do workplace surveillance technologies shape labor relations in China's manufacturing sector? Drawing on 42 interviews with factory workers in the Pearl River Delta, this study examines WeChat's dual role as job-search tool and managerial surveillance mechanism. Findings reveal a paradox: workers depend on platforms for employment yet resent their intrusion into personal time. Responses vary by organizational position—line workers passively comply while team leaders actively negotiate boundaries. These patterns challenge platform determinism, showing how existing power hierarchies mediate digital tools' meanings and effects. The findings suggest interventions targeting platform design alone cannot address surveillance absent structural workplace reform.

Rationale

Fixed: (1) Opened with research question; (2) Specified method, sample, location; (3) Named the platform and its dual role; (4) Stated the paradox (key finding); (5) Described variation pattern; (6) Articulated theoretical contribution; (7) Indicated policy implication; (8) All in 8 sentences with varied rhythm; (9) No mechanical transitions; (10) Every sentence conveys substantive information.

---

Example 8: Paragraph with Heavy Citations

Original (AI-generated)

According to previous research, social media affects political behavior (Smith 2020). Studies have shown various effects (Jones 2021; Williams 2019). Moreover, scholars have noted important patterns (Brown 2022). Additionally, research demonstrates significant findings (Davis 2023). The literature suggests that multiple factors are involved (Miller 2021).

Issues Identified

• ❌ Generic "According to research" framing
• ❌ Citations as decoration rather than substance
• ❌ No actual claims from cited works
• ❌ Mechanical transitions
• ❌ Vague throughout ("various effects," "important patterns")

Revised (Humanized)

Experimental studies consistently find that corrective information reduces misperceptions—by 8-15% in most samples (Nyhan & Reifler 2015; Porter & Wood 2019). Yet field effects prove elusive. When Facebook partnered with fact-checkers in 2017, flagged content circulated almost unchanged (Pennycook et al. 2018). Smith's (2020) panel study offers an explanation: corrections work for low-salience issues but backfire for identity-relevant topics, actually strengthening misperceptions among partisan respondents. This "backfire effect" remains contested—recent meta-analysis by Wood and Porter (2023) finds it rare—but the broader pattern holds: information alone cannot overcome motivated reasoning.

Rationale

Fixed: (1) Scholars become agents making specific claims; (2) Quantified findings included (8-15%, year 2017); (3) Named specific intervention (Facebook fact-checking partnership); (4) Explained mechanism ("backfire effect" defined); (5) Showed scholarly debate (contested finding, meta-analysis qualification); (6) Synthesized toward broader theoretical point (motivated reasoning); (7) Removed generic transition words; (8) Integrated 7 citations substantively rather than decoratively; (9) Used italics for emphasis; (10) Varied sentence structure.

---

Common Patterns Across All Examples

What Was Removed

• Mechanical transitions (Moreover, Furthermore, Additionally)
• Abstract scaffolding ("various aspects," "multiple factors")
• Generic attribution ("research shows," "studies have found")
• Formulaic phrases ("It is important to note")
• Uniform sentence rhythm

What Was Added

• Specific concepts, theories, names
• Quantitative details where relevant
• Participant voices (in qualitative sections)
• Critical engagement with sources
• Varied sentence structures
• Scholarly hedging and nuance
• Concrete examples and contexts
• Genuine intellectual synthesis

Consistent Strategies

1. Specificity: Replace every abstraction with concrete details 2. Agency: Scholars/participants as active agents, not passive "research" 3. Synthesis: Show connections between ideas, not just lists 4. Voice: Include appropriate hedging and critical engagement 5. Rhythm: Deliberately vary sentence length and structure 6. Flow: Let content create connections, minimize mechanical transitions

---

Discipline-Specific Examples

Sociology Example

AI: "Social class plays an important role in educational outcomes. Moreover, various factors mediate this relationship."

Human: "Bourdieu's (1984) concept of cultural capital explains how middle-class families transmit educational advantages: not through income alone, but via embodied competencies, linguistic styles, and institutional know-how that schools reward. Lareau's (2003) ethnography specifies the mechanism—'concerted cultivation' parenting practices that align with teachers' expectations."

Anthropology Example

AI: "The ritual has various meanings for participants. Additionally, it serves multiple functions in the community."

Human: "The funeral ritual does not simply mourn the dead—it reconstitutes social order. When a Toraja elder dies, the village assembles for ceremonies lasting weeks, slaughtering dozens of buffalo in displays that affirm kinship networks and redistribute wealth (Volkman 1985). Participation is obligatory; absence signals social death."

Political Science Example

AI: "Institutions affect policy outcomes. Moreover, different types of institutions have various effects."

Human: "Electoral systems shape party fragmentation: proportional representation (PR) enables minor parties while first-past-the-post (FPTP) favors two-party systems (Duverger 1954). Recent panel data across 63 democracies confirms this pattern: PR countries average 4.2 parliamentary parties versus 2.7 in FPTP systems (Golder 2005)."

---

Self-Check Questions

After rewriting, ask:

1. Specificity: Did I replace every "various/multiple/different" with concrete details? 2. Rhythm: Do my sentences vary significantly in length? 3. Transitions: Did I eliminate mechanical "Moreover/Furthermore"? 4. Voice: Did I include appropriate hedging and scholarly nuance? 5. Citations: Are scholars/studies integrated substantively, not just listed? 6. Examples: Did I ground abstract claims in concrete instances? 7. Synthesis: Do ideas connect through content logic, not just formal connectors? 8. Authenticity: Does it sound like a thoughtful human scholar thinking aloud?

If you answer "no" to any question, revise further.

---

Remember: The goal is authentic scholarly communication—demonstrating genuine intellectual engagement with your topic, not fooling detection software. Your revisions should clarify and strengthen arguments, not obscure them.

学术文本人性化改写原则

🇬🇧 English | 🇨🇳 中文

本文档用中文讲解将AI生成的英文学术文本转化为自然、人性化学术写作的详细策略。

📝 说明：所有示例都是英文（改写前和改写后），但用中文解释问题、修复方法和修改理由，帮助非英语母语者理解如何改进英文学术写作。

---

1. 变化句子节奏

问题

AI倾向于生成长度均匀的句子，创造单调的节奏，感觉机械化。

检测方法

• 所有句子在15-20词之间
• 句子长度标准差低
• 缺乏自然流畅和节奏

修复方法

策略性地混合句子长度：

• 短句 (5-10词)：强调关键点，创造冲击
• 中等句子 (12-20词)：标准学术discourse
• 长句 (25-35词)：复杂想法，多个从句

示例

AI模式（每句约18词，英文）：

This study examines the impact of social media on political engagement. The research focuses specifically on young adults in urban areas. The analysis considers multiple factors including education and income. The findings suggest several important implications for democratic participation.

人类模式（变化：8, 24, 11词，英文）：

This study examines social media's political effects among urban youth. Drawing on survey data from three cities, we find that platform usage correlates with civic engagement primarily among college-educated respondents, though effects vary by income. The implications are significant.

中文理由：混合短句(8)、长句(24)和中等句(11)；消除重复的"The [noun] [verb]"开头；将相关想法组合成复杂句子结构。

---

2. 消除机械过渡词

问题

AI过度使用句首的正式过渡词，创造清单式的感觉。

常见AI过渡词

• Moreover,
• Furthermore,
• Additionally,
• In addition,
• It is important to note that
• It should be noted that
• Significantly,

修复方法

使用多样化的过渡策略：

策略A：隐式连接（不需要连接词）

让逻辑流动自己说话。

AI（英文）: "Social media affects political views. Moreover, it influences voting behavior." 人类（英文）: "Social media affects political views and shapes voting behavior."

策略B：内容驱动的过渡

使用推进论证的实质性短语。

AI（英文）: "Moreover, previous research has limitations." 人类（英文）: "These studies, however, overlook rural contexts."

策略C：指示词 + 名词

引用前一句的具体内容。

AI（英文）: "Furthermore, data shows correlation." 人类（英文）: "This pattern echoes findings from earlier surveys."

策略D：变化的连词（谨慎使用）

• Yet / Still / However (对比)
• Thus / Hence (因果)
• Indeed / In fact (强调)

完整示例

AI模式（英文）：

Social media platforms have changed political communication. Moreover, they have altered how citizens engage with news. Additionally, these platforms affect political polarization. Furthermore, research shows concerning trends. In addition, younger users are particularly affected.

人类模式（英文）：

Social media platforms have fundamentally reshaped political communication—not just how citizens consume news, but the very structure of public discourse. These changes correlate with rising polarization, though causality remains contested. Youth appear particularly vulnerable.

中文理由：消除五个机械过渡；使用破折号进行阐述；包括学术限定语（"correlate"而非"cause"）；组合想法以实现流畅。

---

3. 替换抽象占位语

问题

AI用模糊短语填充空间，几乎不说任何具体内容。

常见犯规者（英文短语）

修复方法

用具体性替换抽象：

• 命名理论、学者、概念
• 提供具体例子
• 尽可能量化
• 引用特定背景

示例

AI模式（英文）：

In terms of the various aspects of social interaction, it is important to note that multiple factors play crucial roles. Different perspectives suggest that these elements contribute to outcomes in various ways. Research shows that these factors are significant across different contexts.

人类模式（英文）：

Social interaction depends on trust, reciprocity, and shared norms—factors that vary systematically across individualist versus collectivist cultures (Triandis 1995). Experimental evidence suggests trust matters most in economic exchanges, while shared norms govern political cooperation (Ostrom 2000).

中文理由：用三个具体概念（trust, reciprocity, shared norms）替换"various aspects"；命名理论框架和学者；指定背景（individualist vs. collectivist）；将主张植根于具体研究发现。

---

4. 添加学术声音和限定语

问题

AI经常过于肯定地陈述主张，或具有缺乏细微参与的通用学术形式。

修复方法

展示真正的学术思维：

适当的限定语（英文）

在适当时使用：

• "suggests" / "indicates" (而非"shows" / "proves")
• "appears to" / "seems to"
• "may" / "might" / "could"
• "potentially" / "possibly"
• "tends to"
• "in many cases" / "often"

AI（英文）: "The data shows that X causes Y." 人类（英文）: "The data suggest a correlation between X and Y, though the causal mechanism remains unclear."

批判性参与

展示智力角力：

• 承认局限性
• 注意相互竞争的解释
• 引用正在进行的辩论
• 指出什么仍不确定

示例（英文）：

AI: "Research demonstrates the effectiveness of the intervention."

人类: "While experimental studies report positive effects (Smith 2020; Jones 2021), critics question external validity given the controlled settings (Williams 2022). Field trials remain scarce."

中文理由：添加限定语（"report"而非"demonstrate"）；承认批评（external validity质疑）；指出证据缺口（field trials稀缺）。

---

5. 将抽象概念植根于具体性

问题

AI在没有具体锚定的情况下做出通用主张。

修复方法

始终提供：

• 具名学者/理论
• 具体例子
• 具体数据/统计
• 特定背景
• 实际研究（而非"research shows"）

示例

AI模式（英文）：

Research has shown that education affects various outcomes. Studies have found multiple effects across different populations. Scholars have noted important variations. Evidence suggests significant relationships exist. These findings have important implications.

人类模式（英文）：

Each additional year of schooling increases earnings by approximately 8-10% (Card 1999), though returns vary: Goldin and Katz (2008) find diminishing effects in saturated labor markets, while Duflo (2001) documents returns exceeding 15% in rural Indonesia. This heterogeneity suggests institutional context—credential recognition, labor market structure, skill complementarities—mediates educational impacts.

中文理由：用具名学者和具体发现（8-10%增长）替换"research shows"；提供定量估计；指定背景（饱和市场、印尼农村）；命名调节机制（credential recognition等）。

---

6. 变化段落开头

问题

AI经常以相似的方式开始段落，创造重复模式。

常见AI模式（英文）

• "This study examines..."
• "The research shows..."
• "Another important factor..."
• "In addition, ..."
• 每段都以"The [noun]..."开头

修复策略

策略A：以关键发现/主张开头

英文示例：

Contrary to conventional wisdom, social media use correlates negatively with polarization in our sample.

策略B：从背景/上下文开始

英文示例：

Since Facebook's algorithmic shift in 2015, scholars have debated...

策略C：用例子/说明开头

英文示例：

When Venezuelan activists mobilized via Twitter in 2019, traditional media initially ignored...

策略D：以问题开始（谨慎使用）

英文示例：

Why do some movements translate online activism into offline mobilization while others fail?

---

7. 主动与被动语态平衡

问题

AI学术写作经常过度使用被动语态（>50%的句子）。

何时使用主动语态：

• 强调研究者角色
• 描述你采取的行动
• 使主张更直接清晰

对比示例（英文）：

• 被动: "The data was analyzed using regression."
• 主动: "We analyzed data using regression."

何时使用被动语态：

• 行动者未知或不重要
• 保持学术距离
• 强调结果而非行动者

对比示例（英文）：

• 主动: "Researchers conducted the survey."
• 被动: "The survey was administered online." （行动者不重要）

---

总结：转换清单（中文）

改写AI文本时，系统地处理：

• [ ] 句子节奏：混合短(5-10)、中(12-20)、长(25-35)词句子
• [ ] 过渡：消除机械的"Moreover/Furthermore"；使用隐式流畅
• [ ] 抽象：用具体内容替换"various aspects/multiple factors"
• [ ] 声音：添加学术限定语（suggests而非shows）和批判性参与
• [ ] 植根：命名学者、理论、具体例子
• [ ] 段落开头：变化段落开始方式
• [ ] 引用：作为实质性内容整合，而非装饰
• [ ] 句法：混合简单、复合、复杂句子结构
• [ ] 语态平衡：策略性地混合主动和被动
• [ ] 战略强调：偶尔使用破折号或片段以实现节奏（罕见）

---

针对非英语母语者的特别提示（中文）

常见误区

1. 过度使用"in terms of" - 通常可以删除或直接陈述 2. 机械过渡依赖 - 用内容驱动的连接替换 3. 所有句子S-V-O结构 - 变化句式 4. 过于正式的词汇 - "utilize" → "use", "commence" → "begin"

你的优势（要保留）

• ✅ 清晰的逻辑结构
• ✅ 正式的学术语域
• ✅ 精确使用技术术语
• ✅ 仔细的语法

重点改进的地方

1. 句子节奏变化 2. 消除机械过渡词 3. 用具体内容替换抽象短语 4. 添加适当的学术限定语 5. 整合具体引用和例子

---

记住： 目标是真实的学术交流，而非机械应用规则。大声朗读你的修订——它听起来像一个有思想的人类学者在处理想法，还是一台生成文本的机器？

完整示例请参见 examples_CN.md。

Rewriting Principles for Humanizing Academic Text

This document provides detailed strategies for transforming AI-generated academic text into natural, human-like scholarly writing.

---

1. Sentence Rhythm Variation

The Problem

AI tends to generate sentences of uniform length, creating a monotonous rhythm that feels mechanical.

Detection

• All sentences between 15-20 words
• Low standard deviation in sentence length
• Lack of natural flow and pacing

The Fix

Mix sentence lengths strategically:

• Short sentences (5-10 words): Emphasize key points, create impact
• Medium sentences (12-20 words): Standard academic discourse
• Long sentences (25-35 words): Complex ideas, multiple clauses

Examples

AI Pattern (uniform ~18 words each):

This study examines the impact of social media on political engagement. The research focuses specifically on young adults in urban areas. The analysis considers multiple factors including education and income. The findings suggest several important implications for democratic participation.

Human Pattern (varied: 8, 24, 11 words):

This study examines social media's political effects among urban youth. Drawing on survey data from three cities, we find that platform usage correlates with civic engagement primarily among college-educated respondents, though effects vary by income. The implications are significant.

Rationale: Mixed short (8), long (24), and medium (11) sentences; eliminated repetitive "The [noun] [verb]" openings; combined related ideas into complex sentence structure.

---

2. Eliminate Mechanical Transitions

The Problem

AI overuses formal transition words at sentence beginnings, creating a checklist-like feel.

Common AI Offenders

• Moreover,
• Furthermore,
• Additionally,
• In addition,
• It is important to note that
• It should be noted that
• Significantly,

The Fix

Use diverse transition strategies:

Strategy A: Implicit Connection (No Connector Needed)

Let logical flow speak for itself.

AI: "Social media affects political views. Moreover, it influences voting behavior." Human: "Social media affects political views and shapes voting behavior."

Strategy B: Content-Driven Transitions

Use substantive phrases that advance the argument.

AI: "Moreover, previous research has limitations." Human: "These studies, however, overlook rural contexts."

Strategy C: Demonstrative + Noun

Reference specific content from prior sentence.

AI: "Furthermore, data shows correlation." Human: "This pattern echoes findings from earlier surveys."

Strategy D: Varied Conjunctions (Used Sparingly)

• Yet / Still / However (adversative)
• Thus / Hence (causal, less formal than "Therefore")
• Indeed / In fact (emphatic)

Examples

AI Pattern:

Social media platforms have changed political communication. Moreover, they have altered how citizens engage with news. Additionally, these platforms affect political polarization. Furthermore, research shows concerning trends. In addition, younger users are particularly affected.

Human Pattern:

Social media platforms have fundamentally reshaped political communication—not just how citizens consume news, but the very structure of public discourse. These changes correlate with rising polarization, though causality remains contested. Youth appear particularly vulnerable.

Rationale: Eliminated five mechanical transitions; used dash for elaboration; included scholarly hedging ("correlate" not "cause"); combined ideas for flow.

---

3. Replace Abstract Scaffolding

The Problem

AI fills space with vague phrases that say little concrete.

Common Culprits

The Fix

Replace abstractions with specificity:

• Name theories, scholars, concepts
• Provide concrete examples
• Quantify when possible
• Reference particular contexts

Examples

AI Pattern:

In terms of the various aspects of social interaction, it is important to note that multiple factors play crucial roles. Different perspectives suggest that these elements contribute to outcomes in various ways. Research shows that these factors are significant across different contexts.

Human Pattern:

Social interaction depends on trust, reciprocity, and shared norms—factors that vary systematically across individualist versus collectivist cultures (Triandis 1995). Experimental evidence suggests trust matters most in economic exchanges, while shared norms govern political cooperation (Ostrom 2000).

Rationale: Replaced "various aspects" with three specific concepts; named theoretical framework and scholars; specified contexts (individualist vs. collectivist); grounded claims in concrete research findings.

---

4. Add Scholarly Voice and Hedging

The Problem

AI often makes claims too definitively or with generic academic formality lacking nuanced engagement.

The Fix

Demonstrate genuine scholarly thinking:

Appropriate Hedging

Use when warranted:

• "suggests" / "indicates" (not "shows" / "proves")
• "appears to" / "seems to"
• "may" / "might" / "could"
• "potentially" / "possibly"
• "tends to"
• "in many cases" / "often"

AI: "The data shows that X causes Y." Human: "The data suggest a correlation between X and Y, though the causal mechanism remains unclear."

Critical Engagement

Show intellectual grappling:

• Acknowledge limitations
• Note competing explanations
• Reference ongoing debates
• Indicate what remains uncertain

AI: "Research demonstrates the effectiveness of the intervention." Human: "While experimental studies report positive effects (Smith 2020; Jones 2021), critics question external validity given the controlled settings (Williams 2022). Field trials remain scarce."

Disciplinary Positioning

Ground work in scholarly traditions:

AI: "This study uses qualitative methods to examine the phenomenon." Human: "Following interpretivist tradition, this ethnographic study examines how actors construct meaning within institutional constraints."

Examples

AI Pattern:

The research clearly demonstrates that social media causes political polarization. The data shows strong effects across all demographics. These findings prove the hypothesis is correct. The results indicate that intervention is necessary.

Human Pattern:

The research suggests social media may contribute to polarization, particularly among politically engaged users (Boxell et al. 2017), though effect sizes remain modest and causality uncertain. Recent panel studies complicate this picture, finding minimal effects on attitude extremity (Guess et al. 2023). The mechanism—if one exists—likely involves selective exposure rather than direct persuasion.

Rationale: Added hedging ("may contribute," "likely involves"); cited specific studies; acknowledged contradictory evidence; specified conditional ("particularly among..."); maintained scholarly uncertainty where appropriate.

---

5. Ground Abstractions in Specificity

The Problem

AI makes generic claims without concrete anchoring.

The Fix

Always provide:

• Named scholars/theories
• Specific examples
• Concrete data/statistics
• Particular contexts
• Actual studies (not "research shows")

Examples

AI Pattern:

Research has shown that education affects various outcomes. Studies have found multiple effects across different populations. Scholars have noted important variations. Evidence suggests significant relationships exist. These findings have important implications.

Human Pattern:

Each additional year of schooling increases earnings by approximately 8-10% (Card 1999), though returns vary: Goldin and Katz (2008) find diminishing effects in saturated labor markets, while Duflo (2001) documents returns exceeding 15% in rural Indonesia. This heterogeneity suggests institutional context—credential recognition, labor market structure, skill complementarities—mediates educational impacts.

Rationale: Replaced "research shows" with named scholars and specific findings; provided quantitative estimates; specified contexts (saturated markets, rural Indonesia); named mediating mechanisms.

---

6. Vary Paragraph Openings

The Problem

AI often starts paragraphs similarly, creating repetitive patterns.

Common AI Patterns

• "This study examines..."
• "The research shows..."
• "Another important factor..."
• "In addition, ..."
• Every paragraph starting with "The [noun]..."

The Fix

Use diverse opening strategies:

Strategy A: Lead with Key Finding/Claim

Contrary to conventional wisdom, social media use correlates negatively with polarization in our sample.

Strategy B: Start with Context/Background

Since Facebook's algorithmic shift in 2015, scholars have debated...

Strategy C: Open with Example/Illustration

When Venezuelan activists mobilized via Twitter in 2019, traditional media initially ignored...

Strategy D: Begin with Question (Sparingly)

Why do some movements translate online activism into offline mobilization while others fail?

Strategy E: Use Transition Phrase (Occasionally)

Building on these insights, we now turn to...

Examples

AI Pattern (all start with "The [noun] [verb]"):

The study examines social media effects. The research focuses on political engagement. The analysis considers three platforms. The findings reveal important patterns. The data shows significant correlations.

Human Pattern (varied openings):

Social media platforms claim to foster democratic participation, yet evidence remains mixed. We examine this tension through panel data tracking 2,400 users across three election cycles. Surprisingly, heavy users show less political engagement offline—suggesting substitution rather than mobilization. This pattern echoes findings from earlier studies of slacktivism (Morozov 2011). But mechanism matters.

Rationale: Varied openings (claim, method statement, finding, reference, fragment); eliminated repetitive "The [noun]" pattern; used dash for elaboration; included sentence fragment for emphasis ("But mechanism matters").

---

7. Natural Citation Integration

The Problem

AI often cites generically: "Research shows..." or mechanically: "According to Smith (2020)..."

The Fix

Integrate citations as substantive content:

Pattern A: Scholars as Agents

AI: "According to research, trust matters." Human: "Putnam (2000) argues trust forms the bedrock of civic engagement."

Pattern B: Findings as Core Content

AI: "Studies have shown effects (Smith 2020; Jones 2021)." Human: "Experimental manipulations consistently reduce polarization by 15-20% (Smith 2020; Jones 2021), though field effects remain elusive."

Pattern C: Engage with Scholar's Argument

AI: "Smith (2020) discusses polarization." Human: "Smith's (2020) claim that algorithms drive polarization, while influential, rests on correlational evidence that admits multiple interpretations."

Examples

AI Pattern:

According to research, social media affects politics (Smith 2020). Studies have shown various effects (Jones 2021). Previous research has examined this topic (Williams 2019). Scholars have noted important patterns (Brown 2022).

Human Pattern:

Smith's (2020) provocative claim—that Facebook's algorithm caused Trump's election—sparked fierce debate. Critics note his correlational evidence cannot rule out selection effects (Jones 2021), while defenders argue the mechanism is theoretically sound (Williams 2022). Recent field experiments offer partial support: algorithmic changes reduce polarization by 10-15% (Brown 2023), though effects dissipate within weeks.

Rationale: Scholars appear as active agents; specific claims stated; critical engagement shown; quantitative findings included; competing interpretations presented; temporal dynamics noted.

---

8. Sentence Structure Variation

The Problem

AI defaults to simple declarative sentences: Subject-Verb-Object.

The Fix

Use varied syntactic structures:

Simple Sentence (S-V-O)

Trust matters.

Compound Sentence (Two Independent Clauses)

Trust matters for cooperation, but norms govern enforcement.

Complex Sentence (Dependent + Independent Clause)

While trust facilitates cooperation, norms govern enforcement.

Compound-Complex

Trust facilitates cooperation, but when norms are weak, even high-trust societies experience free-riding.

Sentence with Interrupting Clause

Trust—more than any other factor—predicts cooperative outcomes.

Sentence Beginning with Modifier

Despite high trust, cooperation failed when institutions were weak.

Inverted Structure

Not just trust but also reciprocity matters.

Examples

AI Pattern (all S-V-O):

Social media platforms affect political views. These platforms shape information consumption. Users encounter diverse content. Algorithms filter information. This filtering creates echo chambers.

Human Pattern (varied structures):

Social media platforms shape political views not through direct persuasion but by filtering information flows. When algorithms prioritize engagement—and outrage drives engagement—users encounter increasingly extreme content (Levy 2021). The result: echo chambers that reinforce existing beliefs.

Rationale: Used complex sentence with interrupting clause; varied structure; included causal connection; added citation; used colon for elaboration; mixed medium and long sentences.

---

9. Active vs. Passive Voice Balance

The Problem

AI academic writing often overuses passive voice (>50% of sentences).

The Fix

Strategic voice mixing:

Use Active When:

• Emphasizing agent/researcher role
• Describing actions you took
• Making claims more direct and clear

Passive: "The data was analyzed using regression." Active: "We analyzed data using regression."

Use Passive When:

• Agent is unknown or unimportant
• Maintaining academic distance
• Emphasizing object/result over actor
• Following disciplinary conventions (varies by field)

Active: "Researchers conducted the survey." Passive: "The survey was administered online." (agent unimportant)

Mix Both

AI Pattern (all passive):

The study was conducted in three cities. Participants were recruited through advertisements. Data were collected via surveys. Analysis was performed using SPSS. Results were interpreted using theory.

Human Pattern (mixed voice):

We conducted the study in Shanghai, Beijing, and Guangzhou, recruiting 1,200 participants through targeted social media advertisements. Surveys were administered online to maximize response rates. Regression analysis reveals three key patterns...

Rationale: Active voice for researcher agency ("We conducted," "recruiting"); passive where agent unimportant ("were administered"); active for findings ("reveals").

---

10. Strategic Use of Fragments and Dashes

The Problem

AI rigidly follows grammatical rules, never using fragments or dashes for emphasis.

The Fix (Use Sparingly)

Occasional fragments and dashes add natural rhythm:

Sentence Fragments (Rare, for Emphasis)

The data suggest a correlation. A strong one.

Trust matters. Not just for cooperation, but for societal cohesion itself.

Dashes for Elaboration/Interruption

Trust—more than any measured variable—predicts outcomes.

The mechanism remains unclear—though several hypotheses exist.

Warning: Use sparingly (1-2 times per page). Overuse becomes gimmicky.

Examples

AI Pattern (no fragments, no dashes):

The finding is significant. It suggests that trust matters. This is important for understanding outcomes. The mechanism requires further study.

Human Pattern (strategic use):

The finding is striking: trust predicts outcomes more powerfully than income, education, or political ideology—combined. Yet the mechanism remains opaque. More on this below.

Rationale: Used colon for emphasis; added dash for dramatic comparison; included fragment ("More on this below"); maintained scholarly tone despite informal elements.

---

Summary: Transformation Checklist

When rewriting AI text, systematically address:

• [ ] Sentence rhythm: Mix short (5-10), medium (12-20), long (25-35) word sentences
• [ ] Transitions: Eliminate mechanical "Moreover/Furthermore"; use implicit flow
• [ ] Abstractions: Replace "various aspects/multiple factors" with specifics
• [ ] Voice: Add scholarly hedging and critical engagement
• [ ] Grounding: Name scholars, theories, specific examples
• [ ] Paragraph openings: Vary how paragraphs begin
• [ ] Citations: Integrate as substantive content, not decoration
• [ ] Syntax: Mix simple, compound, complex sentence structures
• [ ] Voice balance: Mix active and passive strategically
• [ ] Strategic emphasis: Occasional dash or fragment for rhythm (rare)

---

Discipline-Specific Notes

Quantitative Social Science

• More passive voice acceptable (methodological sections)
• Precise terminology critical
• Hypothesis-testing language
• Causal inference vocabulary

Qualitative Social Science

• More active voice acceptable
• Reflexivity valued
• Thick description
• Interpretive language

Theory-Heavy Writing

• Abstract concepts necessary (but define them)
• Engage deeply with prior scholarship
• Show intellectual lineage
• Demonstrate critical synthesis

---

Remember: The goal is authentic scholarly communication, not mechanical application of rules. Read your revision aloud—does it sound like a thoughtful human scholar grappling with ideas, or a machine generating text?

社会科学写作模式与学科惯例

🇬🇧 English | 🇨🇳 中文

本文档用中文为社会科学各学科中人性化AI生成的英文学术写作提供特定领域指导。

📝 说明：所有示例文本都是英文，用中文解释各学科的写作特点、常见错误和修复方法，帮助非英语母语者写出符合学科规范的英文学术文本。

---

社会科学学科概览

每个社会科学领域都有独特的惯例、理论传统和写作风格。理解这些有助于创建真实的学术声音。

---

1. 社会学 (Sociology)

关键概念和词汇（英文术语）

• 理论框架：Functionalism, Conflict theory, Symbolic interactionism, Practice theory, Intersectionality
• 核心概念：Stratification, Inequality, Agency/Structure, Habitus, Capital (economic, cultural, social), Institutions, Norms, Socialization
• 当代辩论：Intersectionality, Precarity, Platform capitalism, Algorithmic governance

写作风格特点

• 抽象理论与实证基础之间的平衡
• 强调反思性传统（特别是定性工作）
• 混合方法日益普遍
• 与古典理论家（Marx, Weber, Durkheim, Bourdieu）和当代工作互动

自然的社会学表达示例

好的（英文）:

• "Drawing on Bourdieu's field theory..."
• "Social capital mediates..."
• "Intersecting axes of inequality..."
• "Institutions structure but do not determine..."
• "Reproducing class distinctions through..."

类似AI（避免，英文）:

• "Society has various structures..."
• "Social factors play important roles..."
• "Different groups experience things differently..."

自然社会学写作示例段落（英文）

Working-class respondents navigated labor markets through kinship networks—what Granovetter (1973) termed "strong ties." Yet these dense, homogeneous networks, while providing security, also constrained mobility by circulating information within narrow class boundaries. Middle-class professionals, conversely, leveraged "weak ties" across diverse social fields, accessing non-redundant information and opportunities. This pattern reproduces inequality not through individual deficits but via differential network structures that reflect and reinforce class positions.

中文解释关键特征：命名理论家和概念（Granovetter, "strong ties"）；具体说明机制（information circulation）；展示不平等再生产；平衡实证观察与理论解释；策略性使用被动语态（"This pattern reproduces..."）。

---

2. 人类学 (Anthropology)

关键概念和词汇（英文术语）

• 核心概念：Culture, Ritual, Kinship, Liminality, Positionality, Thick description, Symbolic systems, Material culture
• 理论传统：Structuralism, Interpretivism, Practice theory, Political economy, Feminist anthropology
• 方法：Ethnography, Participant observation, Multi-sited fieldwork

写作风格特点

• Thick description (Geertz)：丰富的、背景化的细节
• 第一人称可接受且常被偏好（"I observed..."）
• 关于研究者位置和权力的反思性
• 在特殊（民族志细节）和一般（理论洞见）之间平衡
• 重视文学质量——好的写作很重要

自然人类学写作示例段落（英文）

When I asked Liu why she burned spirit money at her grandmother's grave, she laughed: "You think I believe this?" Yet she returned monthly, meticulously folding golden paper into ingots. The ritual's meaning, I came to understand, lay not in metaphysical belief but in performative kinship—making visible her continued obligations to ancestors, demonstrating filial piety to living relatives who witnessed her devotion. Belief became beside the point. The act itself constituted family.

中文解释关键特征：第一人称研究者存在（"I asked", "I came to understand"）；民族志细节（folding golden paper into ingots）；参与者声音（直接引用）；从表面到更深层意义的解释转向；对自己假设的反思性；文学风格（短促有力的句子）。

---

3. 政治学 (Political Science)

关键概念和词汇（英文术语）

• 核心概念：Power, Institutions, Legitimacy, Governance, State capacity, Democracy, Authoritarianism, Political economy, Collective action
• 子领域：
• Comparative politics: Regime types, Democratization, Political institutions
• International relations: Anarchy, Balance of power, Norms
• American politics: Polarization, Elections, Representation
• Political theory: Justice, Rights, Legitimacy

写作风格特点

• 强调因果推断（特别是定量工作）
• 假设检验框架常见
• 更可能在方法中使用被动语态
• 辩论驱动：将论证与竞争解释对立
• 操作定义至关重要

自然政治学写作示例段落（英文）

We hypothesize that proportional representation (PR) increases minor party success by lowering electoral thresholds. To test this, we analyze 63 democracies from 1990-2020, operationalizing party system fragmentation via effective number of parliamentary parties (ENPP). OLS regressions, controlling for social cleavages and district magnitude, reveal PR systems average 1.8 additional parties (p<0.01) compared to first-past-the-post systems. This effect persists across specifications, suggesting electoral rules causally shape party systems.

中文解释关键特征：清晰假设；操作定义（ENPP）；具体方法（OLS regression）；带显著性水平的定量结果；用适当限定语缓和的因果语言（"suggesting"）；控制混杂因素。

---

4. 教育学 (Education)

关键概念和词汇（英文术语）

• 核心概念：Pedagogy, Curriculum, Equity, Achievement gaps, Learning outcomes, Accountability, Tracking, School effectiveness
• 理论框架：Constructivism, Critical pedagogy, Cultural capital, Reproduction theory
• 当代辩论：Standardized testing, Charter schools, Teacher quality, Digital learning

写作风格特点

• 混合方法常见（surveys + interviews + observations）
• 强调政策相关性
• 平衡理论框架与实践启示
• 常面向非学术受众（政策制定者、实践者）

自然教育写作示例段落（英文）

Teacher expectations shape student outcomes—particularly for marginalized groups. When Rosenthal and Jacobson (1968) randomly told teachers certain students would "bloom," those students indeed showed greater gains, suggesting expectations became self-fulfilling prophecies. Recent replications complicate this picture: stereotype threat effects appear strongest in high-stakes testing contexts (Steele & Aronson 1995), weaker in low-pressure classrooms. This conditionality suggests interventions should target evaluative structures, not just individual teacher beliefs.

中文解释关键特征：具体引用经典研究（Rosenthal & Jacobson）；最近的复杂化/限定证据；具体说明机制（self-fulfilling prophecies, stereotype threat）；将实践启示与发现联系；适合更广泛教育受众的可及语言。

---

5. 心理学 (Psychology - Social Psychology)

关键概念和词汇（英文术语）

• 核心概念：Cognition, Behavior, Attitudes, Interventions, Mechanisms, Priming, Cognitive dissonance, Attribution, Conformity
• 方法：Experiments, Surveys, Behavioral measures
• 强调：Mechanisms, Replication, Effect sizes

写作风格特点

• 操作定义至关重要
• Hypothesis → Method → Results → Interpretation 结构
• 强调效应量和统计功效
• 复制危机意识（最近工作）
• 方法部分更可接受被动语态

自然心理学写作示例段落（英文）

We hypothesized that mortality salience increases out-group prejudice (Terror Management Theory; Greenberg et al. 1990). Participants (N=240) were randomly assigned to write about either death (mortality salience) or dental pain (control). They then rated out-group members on warmth and competence scales (Fiske et al. 2002). Mortality salience significantly reduced warmth ratings (β=-0.34, p=.001, d=0.52) but not competence, suggesting threat activates affective rather than cognitive prejudice. However, this effect disappeared among participants high in openness (interaction: β=0.28, p=.02), indicating personality moderates existential threat responses.

中文解释关键特征：与理论联系的清晰假设（Terror Management Theory）；具体样本量（N=240）；操作定义；随机分配；引用验证量表（Fiske et al.）；带效应量的定量结果（d=0.52）；具体说明机制；包括调节分析；解释连接到理论。

---

非英语母语者特别指导（中文）

常见挑战和AI拐杖

挑战1：冠词使用 (a/an/the)

AI常帮助非母语者正确使用冠词，但过度使用正式模式。

类似AI（英文，过于正式）:

"The research examines the phenomenon of the social media."

自然（英文）:

"This research examines social media's effects."

中文提示：阅读你领域学者的文章；注意他们何时省略冠词（通常用于抽象概念："democracy", "trust"，而非"the democracy"）。

挑战2：介词短语

AI默认正式介词短语，听起来生硬。

类似AI（英文）:

• "in terms of"
• "with regard to"
• "with respect to"
• "in the context of"

更自然：

• "regarding" 或直接陈述
• 这些短语通常可以完全删除

挑战3：机械过渡词

非母语者常依赖AI实现句子变化，但AI使用机械过渡。

类似AI（英文）:

"Moreover, Furthermore, Additionally, In addition..."

自然替代（英文）：

• 直接连接（无连接词）
• "Yet," "Still," "However"（谨慎使用）
• "This pattern," "These findings," "Such dynamics"
• 内容驱动的过渡

挑战4：限定语使用

非母语者常在适当限定语方面挣扎。AI有时过度限定或不足限定。

限定不足（英文，过于肯定）:

"The data proves that X causes Y."

过度限定（英文，过于弱）:

"The data might possibly suggest that X could perhaps be related to Y."

适当限定（英文）:

"The data suggest that X may cause Y, though alternative explanations remain."

"The correlation between X and Y is consistent with a causal relationship."

---

学科特定的引用风格（中文说明）

社会学

• 常深入参与古典理论家和当代理论
• 引用实质性整合："Drawing on Bourdieu's (1984) concept of habitus..."
• 理论部分常先于方法

人类学

• 引用民族志经典（Geertz, Evans-Pritchard, Turner）
• 引用常作为对话伙伴："As Clifford (1986) notes..."
• 文献综述可能贯穿全文而非单独部分

政治学

• 定量工作大量使用括号引用：(Smith 2020; Jones 2021)
• 假设和先前发现部分广泛引用
• 复制研究明确引用所复制内容

教育学

• 与学术研究一起引用实践者相关工作
• 引用政策文件和报告（不仅仅是同行评审文章）
• 经常引用经典研究（Rosenthal & Jacobson, Coleman Report）

心理学

• 强调最近工作和复制
• 引用元分析以获得既定效应
• 按效应名称引用经典研究："the Milgram paradigm", "Asch conformity"

---

按学科的常见错误（中文说明）

社会学

❌ 忽视权力和不平等（社会学总是关注分层） ❌ 没有结构背景的个人层面解释 ✅ 展示微观互动如何反映/再生产宏观结构

人类学

❌ 没有厚描述跨文化泛化 ❌ 从叙述中移除研究者（民族志是反思性的） ✅ 平衡特殊细节与更广泛理论洞见

政治学

❌ 没有适当研究设计的因果语言 ❌ 忽视替代解释 ✅ 清楚具体说明因果主张和识别策略

教育学

❌ 没有实践启示的理论 ❌ 忽视实施挑战 ✅ 将发现连接到可行动建议（同时注意约束）

心理学

❌ 没有操作定义讨论构念 ❌ 忽视效应量（仅关注p值） ✅ 具体说明变量如何测量并报告效应大小

---

红旗：AI支架迹象（中文说明）

无论学科，这些英文模式表明AI生成：

• "various aspects/factors/perspectives"
• "Moreover, Furthermore, Additionally" 模式
• 重复"in terms of"
• "It is important to note that"
• "plays a crucial/important role"
• 所有句子相似长度
• 没有具体引用（只是"research shows"）
• 没有具体例子或数据
• 完全抽象语言

---

学科特定改写示例（英文示例，中文解释）

社会学示例

AI生成（英文）:

Social class plays an important role in educational outcomes. Moreover, various factors mediate this relationship.

人性化（英文）:

Bourdieu's (1984) concept of cultural capital explains how middle-class families transmit educational advantages: not through income alone, but via embodied competencies, linguistic styles, and institutional know-how that schools reward. Lareau's (2003) ethnography specifies the mechanism—'concerted cultivation' parenting practices that align with teachers' expectations.

中文解释：用布迪厄的具体理论概念（cultural capital）替换了"plays an important role"；说明了文化资本的具体形式（embodied competencies, linguistic styles, institutional know-how）；添加了Lareau的民族志来说明机制的运作方式（concerted cultivation）。

人类学示例

AI生成（英文）:

The ritual has various meanings for participants. Additionally, it serves multiple functions in the community.

人性化（英文）:

The funeral ritual does not simply mourn the dead—it reconstitutes social order. When a Toraja elder dies, the village assembles for ceremonies lasting weeks, slaughtering dozens of buffalo in displays that affirm kinship networks and redistribute wealth (Volkman 1985). Participation is obligatory; absence signals social death.

中文解释：用具体的民族志细节（Toraja长者葬礼，宰杀水牛）替换了"various meanings"；说明了仪式的具体功能（affirm kinship networks, redistribute wealth）；包含了后果（absence = social death）；使用破折号和分号增加节奏。

政治学示例

AI生成（英文）:

Institutions affect policy outcomes. Moreover, different types of institutions have various effects.

人性化（英文）:

Electoral systems shape party fragmentation: proportional representation (PR) enables minor parties while first-past-the-post (FPTP) favors two-party systems (Duverger 1954). Recent panel data across 63 democracies confirms this pattern: PR countries average 4.2 parliamentary parties versus 2.7 in FPTP systems (Golder 2005).

中文解释：用具体的制度类型（PR vs. FPTP）替换了"institutions"；提供了量化证据（4.2 vs. 2.7个政党）；命名了经典理论（Duverger's Law）和实证研究（Golder）；使用冒号连接主张和证据。

---

非英语母语者：要保留的优势（中文说明）

你的优势

1. 清晰的逻辑结构：保持这一点。明确的路标帮助读者。

2. 正式语域：学术写作应该是正式的。不要过度随意化。

3. 精确使用技术术语：如果你仔细学习了学科词汇，你可能比松散使用术语的母语者更精确地使用它。

4. 新鲜视角：非英语语言结构可以提供框架想法的替代方式。

要人性化的重点

1. 变化句子节奏（混合短和长句） 2. 消除机械过渡词（Moreover, Furthermore） 3. 替换抽象短语（"various aspects"）为具体内容 4. 添加具体例子和引用 5. 自然整合适当限定语（suggests而非shows）

---

清单：适合学科的人性化（中文）

改写后，验证：

• [ ] 自然使用特定领域的英文词汇
• [ ] 引用对你学科核心的学者/理论
• [ ] 遵循学科惯例（语态、结构）
• [ ] 平衡抽象/具体适合领域
• [ ] 参与你领域的关键辩论
• [ ] 匹配学科规范的语气（反思性 vs. 中性）
• [ ] 包括适当证据类型（民族志细节、统计数据等）
• [ ] 展示对学科对话的智力贡献

---

记住： 每个学科都因充分理由发展了写作惯例——它们反映认识论承诺和受众期望。人性化你的英文写作不仅意味着移除AI模式，还意味着在你的传统内体现真实的学术声音。

更多示例请参见 examples_CN.md。

Social Science Writing Patterns and Disciplinary Conventions

This document provides field-specific guidance for humanizing AI-generated academic writing in social science disciplines.

---

Overview of Social Science Disciplines

Each social science field has distinct conventions, theoretical traditions, and writing styles. Understanding these helps create authentic scholarly voice.

---

1. Sociology

Key Concepts and Vocabulary

• Theoretical frameworks: Functionalism, conflict theory, symbolic interactionism, practice theory, intersectionality
• Core concepts: Stratification, inequality, agency/structure, habitus, capital (economic, cultural, social), institutions, norms, socialization
• Contemporary debates: Intersectionality, precarity, platform capitalism, algorithmic governance

Writing Style

• Balance between abstract theory and empirical grounding
• Strong tradition of reflexivity (especially in qualitative work)
• Mixed methods increasingly common
• Engage with classical theorists (Marx, Weber, Durkheim, Bourdieu) and contemporary work

Example Phrases (Natural to Sociology)

Good:

• "Drawing on Bourdieu's field theory..."
• "Social capital mediates..."
• "Intersecting axes of inequality..."
• "Institutions structure but do not determine..."
• "Reproducing class distinctions through..."

AI-like (avoid):

• "Society has various structures..."
• "Social factors play important roles..."
• "Different groups experience things differently..."

Sample Paragraph (Natural Sociological Writing)

Working-class respondents navigated labor markets through kinship networks—what Granovetter (1973) termed "strong ties." Yet these dense, homogeneous networks, while providing security, also constrained mobility by circulating information within narrow class boundaries. Middle-class professionals, conversely, leveraged "weak ties" across diverse social fields, accessing non-redundant information and opportunities. This pattern reproduces inequality not through individual deficits but via differential network structures that reflect and reinforce class positions.

Key features: Named theorist and concept, specified mechanism, showed reproduction of inequality, balanced empirical observation with theoretical interpretation, used passive voice strategically ("This pattern reproduces...").

---

2. Anthropology

Key Concepts and Vocabulary

• Core concepts: Culture, ritual, kinship, liminality, habitus, positionality, thick description, symbolic systems, material culture
• Theoretical traditions: Structuralism, interpretivism, practice theory, political economy, feminist anthropology
• Methods: Ethnography, participant observation, multi-sited fieldwork

Writing Style

• Thick description (Geertz): Rich, contextual detail
• First-person acceptable and often preferred ("I observed...")
• Reflexivity about researcher position and power
• Balance between particular (ethnographic detail) and general (theoretical insight)
• Literary quality valued—good writing matters

Example Phrases (Natural to Anthropology)

Good:

• "During my 18 months in the village..."
• "Informants described..."
• "The ritual enacted symbolic boundaries between..."
• "Thick description reveals..."
• "Embodied practices reproduced..."

AI-like (avoid):

• "Culture plays an important role..."
• "Different cultures have various customs..."
• "Ethnographic methods were used..."

Sample Paragraph (Natural Anthropological Writing)

When I asked Liu why she burned spirit money at her grandmother's grave, she laughed: "You think I believe this?" Yet she returned monthly, meticulously folding golden paper into ingots. The ritual's meaning, I came to understand, lay not in metaphysical belief but in performative kinship—making visible her continued obligations to ancestors, demonstrating filial piety to living relatives who witnessed her devotion. Belief became beside the point. The act itself constituted family.

Key features: First-person researcher presence, ethnographic detail (specific actions), participant voice (direct quote), interpretive move from surface to deeper meaning, reflexivity about own assumptions ("I came to understand"), literary style (short emphatic sentences).

---

3. Political Science

Key Concepts and Vocabulary

• Core concepts: Power, institutions, legitimacy, governance, state capacity, democracy, authoritarianism, political economy, collective action
• Subfield variations:
• Comparative politics: Regime types, democratization, political institutions
• International relations: Anarchy, balance of power, norms, international institutions
• American politics: Polarization, elections, representation
• Political theory: Justice, rights, legitimacy

Writing Style

• Causal inference emphasized (especially quantitative work)
• Hypothesis-testing frameworks common
• More likely to use passive voice in methods
• Debate-driven: Position argument against competing explanations
• Operational definitions critical

Example Phrases (Natural to Political Science)

Good:

• "To test this hypothesis, we..."
• "The dependent variable, measured as..."
• "Controlling for confounders..."
• "Causal identification relies on..."
• "This mechanism suggests..."
• "Electoral institutions shape..."

AI-like (avoid):

• "Politics involves various factors..."
• "Governments play important roles..."
• "Democracy has different effects..."

Sample Paragraph (Natural Political Science Writing)

We hypothesize that proportional representation (PR) increases minor party success by lowering electoral thresholds. To test this, we analyze 63 democracies from 1990-2020, operationalizing party system fragmentation via effective number of parliamentary parties (ENPP). OLS regressions, controlling for social cleavages and district magnitude, reveal PR systems average 1.8 additional parties (p<0.01) compared to first-past-the-post systems. This effect persists across specifications, suggesting electoral rules causally shape party systems.

Key features: Clear hypothesis, operational definitions, specific method (OLS regression), quantitative results with significance levels, causal language tempered with appropriate hedging ("suggest"), controlled for confounders.

---

4. Education

Key Concepts and Vocabulary

• Core concepts: Pedagogy, curriculum, equity, achievement gaps, learning outcomes, accountability, tracking, school effectiveness
• Theoretical frameworks: Constructivism, critical pedagogy, cultural capital, reproduction theory
• Contemporary debates: Standardized testing, charter schools, teacher quality, digital learning

Writing Style

• Mixed methods common (surveys + interviews + observations)
• Policy relevance emphasized
• Balances theoretical frameworks with practical implications
• Often addresses non-academic audiences (policymakers, practitioners)

Example Phrases (Natural to Education Research)

Good:

• "Student achievement, measured by..."
• "Pedagogical practices that..."
• "Equitable access requires..."
• "Classroom observations revealed..."
• "Learning outcomes varied by..."

AI-like (avoid):

• "Education has various effects..."
• "Teaching involves multiple factors..."
• "Schools play important roles..."

Sample Paragraph (Natural Education Writing)

Teacher expectations shape student outcomes—particularly for marginalized groups. When Rosenthal and Jacobson (1968) randomly told teachers certain students would "bloom," those students indeed showed greater gains, suggesting expectations became self-fulfilling prophecies. Recent replications complicate this picture: stereotype threat effects appear strongest in high-stakes testing contexts (Steele & Aronson 1995), weaker in low-pressure classrooms. This conditionality suggests interventions should target evaluative structures, not just individual teacher beliefs.

Key features: Classic study cited concretely, recent evidence that complicates/qualifies, mechanism specified (self-fulfilling prophecy, stereotype threat), practical implication linked to findings, accessible language appropriate for broader education audience.

---

5. Psychology (Social Psychology)

Key Concepts and Vocabulary

• Core concepts: Cognition, behavior, attitudes, interventions, mechanisms, priming, cognitive dissonance, attribution, conformity
• Methods: Experiments, surveys, behavioral measures
• Emphasis: Mechanisms, replication, effect sizes

Writing Style

• Operational definitions essential
• Hypothesis → Method → Results → Interpretation structure
• Effect sizes and statistical power emphasized
• Replication crisis awareness (recent work)
• More passive voice acceptable in methods sections

Example Phrases (Natural to Social Psychology)

Good:

• "Participants were randomly assigned to..."
• "The manipulation successfully induced..."
• "Mediation analysis reveals..."
• "Effect size (Cohen's d = 0.42) suggests..."
• "Replicating Smith et al. (2020), we find..."

AI-like (avoid):

• "Psychology studies various behaviors..."
• "Cognition involves multiple processes..."
• "Different factors affect attitudes..."

Sample Paragraph (Natural Social Psychology Writing)

We hypothesized that mortality salience increases out-group prejudice (Terror Management Theory; Greenberg et al. 1990). Participants (N=240) were randomly assigned to write about either death (mortality salience) or dental pain (control). They then rated out-group members on warmth and competence scales (Fiske et al. 2002). Mortality salience significantly reduced warmth ratings (β=-0.34, p=.001, d=0.52) but not competence, suggesting threat activates affective rather than cognitive prejudice. However, this effect disappeared among participants high in openness (interaction: β=0.28, p=.02), indicating personality moderates existential threat responses.

Key features: Clear hypothesis linked to theory, specific sample size, operational definitions, random assignment, validated scales cited, quantitative results with effect sizes, mechanism specified, moderation analysis included, interpretation connected to theory.

---

Non-Native English Speaker Considerations

Common Challenges and AI Crutches

Challenge 1: Article Usage (a/an/the)

AI often helps non-native speakers get articles right, but overuses formal patterns.

AI-like (overly formal):

"The research examines the phenomenon of the social media."

Natural:

"This research examines social media's effects."

Tip: Read articles by scholars in your field; notice when they omit articles (often with abstract concepts: "democracy," "trust," not "the democracy").

Challenge 2: Preposition Patterns

AI defaults to formal prepositional phrases that sound stiff.

AI-like:

• "in terms of"
• "with regard to"
• "with respect to"
• "in the context of"

More natural:

• "regarding" or just state directly
• Often these phrases can be deleted entirely

Challenge 3: Sentence Starters

Non-native speakers often rely on AI for sentence variety, but AI uses mechanical transitions.

AI-like:

"Moreover, Furthermore, Additionally, In addition..."

Natural alternatives (see Rewriting Principles document for full list):

• Direct connection (no connector)
• "Yet," "Still," "However" (sparingly)
• "This pattern," "These findings," "Such dynamics"
• Content-driven transitions

Challenge 4: Vocabulary Formality

AI may suggest overly formal vocabulary that sounds unnatural.

AI-like (too formal):

• "utilize" → use "use"
• "commence" → use "begin"
• "terminate" → use "end"
• "possess" → use "have"

Natural: Academic writing is formal but not archaic. Use straightforward verbs.

Challenge 5: Hedging and Modality

Non-native speakers often struggle with appropriate hedging. AI sometimes over-hedges or under-hedges.

Under-hedged (too definitive):

"The data proves that X causes Y."

Over-hedged (too weak):

"The data might possibly suggest that X could perhaps be related to Y."

Appropriate hedging:

"The data suggest that X may cause Y, though alternative explanations remain."

"The correlation between X and Y is consistent with a causal relationship."

Strengths to Preserve

Non-native speakers often bring valuable qualities to academic writing:

1. Clear logical structure: Maintain this. Your explicit signposting helps readers.

2. Formal register: Academic writing should be formal. Don't over-casualize.

3. Precise use of technical terms: If you've learned disciplinary vocabulary carefully, you may use it more precisely than native speakers who use terms loosely.

4. Fresh perspectives: Non-English language structures can offer alternative ways to frame ideas. Don't erase linguistic diversity entirely.

What to Humanize

Focus your humanizing efforts on:

1. Varying sentence rhythm (mix short and long) 2. Eliminating mechanical transitions (Moreover, Furthermore) 3. Replacing abstract phrases ("various aspects") with specifics 4. Adding concrete examples and citations 5. Integrating appropriate hedging naturally

Example Transformation: Non-Native Speaker

AI-generated (helping non-native speaker):

In terms of the various aspects of social media, it is important to note that multiple factors play crucial roles. Moreover, different perspectives exist on this topic. Additionally, research has shown various effects. Furthermore, these findings have significant implications. It should be noted that more investigation is needed.

Issues: This is heavily AI-scaffolded—probably helping with grammar/articles but creating mechanical, abstract writing.

Better revision (preserving non-native speaker's ideas, removing AI scaffolding):

Social media's political effects remain contested. Optimists emphasize democratization (Shirky 2008), while skeptics note slacktivism (Morozov 2011). My survey data from Shanghai (N=800) suggest both occur simultaneously: users engage politically online yet rarely mobilize offline. This paradox requires explanation. I argue institutional context—specifically, China's fragmented authoritarianism—enables some forms of digital activism while foreclosing others.

What changed: (1) Eliminated all "in terms of," "it is important to note," "various aspects"; (2) Removed mechanical transitions; (3) Added specific concepts and citations; (4) Included concrete data; (5) Stated argument directly; (6) Preserved formal register appropriate for academic writing; (7) Used first person naturally ("I argue"), which is increasingly acceptable.

---

Disciplinary Citation Styles

Sociology

• Often engages deeply with classical theorists (Marx, Weber, Durkheim) and contemporary theory
• Citations integrated substantively: "Drawing on Bourdieu's (1984) concept of habitus..."
• Theory section often precedes methods

Anthropology

• Ethnographic classics referenced (Geertz, Evans-Pritchard, Turner)
• Citations often as conversation partners: "As Clifford (1986) notes..."
• Literature review may be woven throughout rather than separate section

Political Science

• Heavy use of parenthetical citations in quantitative work: (Smith 2020; Jones 2021)
• Hypothesis and prior findings sections cite extensively
• Replication studies explicitly cite what they're replicating

Education

• Practitioner-relevant work cited alongside academic research
• Policy documents and reports cited (not just peer-reviewed articles)
• Classic studies (Rosenthal & Jacobson, Coleman Report) referenced frequently

Psychology

• Emphasizes recent work and replications
• Meta-analyses cited for established effects
• Classic studies referenced by effect name: "the Milgram paradigm," "Asch conformity"

---

Common Mistakes by Discipline

Sociology

❌ Ignoring power and inequality (sociology always attends to stratification) ❌ Individual-level explanations without structural context ✅ Show how micro-interactions reflect/reproduce macro structures

Anthropology

❌ Generalizing across cultures without thick description ❌ Removing the researcher from the account (ethnography is reflexive) ✅ Balance particular detail with broader theoretical insights

Political Science

❌ Causal language without appropriate research design ❌ Ignoring alternative explanations ✅ Clearly specify causal claims and identification strategy

Education

❌ Theory without practical implications ❌ Ignoring implementation challenges ✅ Connect findings to actionable recommendations (while noting constraints)

Psychology

❌ Discussing constructs without operational definitions ❌ Ignoring effect sizes (focusing only on p-values) ✅ Specify how variables are measured and report effect magnitudes

---

Interdisciplinary Work

If your work spans disciplines:

1. Explicitly bridge traditions: "Drawing on political science theories of institutions and anthropological attention to meaning-making..."

2. Use vocabulary accessible across fields: Define technical terms from each discipline.

3. Address multiple audiences: Show how work contributes to each field's debates.

4. Respect disciplinary conventions: Match citation style, voice, structure to target journal/field.

---

Red Flags: Signs of AI Scaffolding

Regardless of discipline, these patterns indicate AI generation:

• "Various aspects/factors/perspectives"
• "Moreover, Furthermore, Additionally" pattern
• "In terms of" repeatedly
• "It is important to note that"
• "Plays a crucial/important role"
• All sentences similar length
• No specific citations (just "research shows")
• No concrete examples or data
• Entirely abstract language

---

Checklist: Discipline-Appropriate Humanizing

After rewriting, verify:

• [ ] Used field-specific vocabulary naturally
• [ ] Cited scholars/theories central to your discipline
• [ ] Followed disciplinary conventions (voice, structure)
• [ ] Balanced abstract/concrete appropriate to field
• [ ] Engaged with key debates in your area
• [ ] Matched tone to disciplinary norms (reflexive vs. neutral)
• [ ] Included appropriate evidence type (ethnographic detail, statistics, etc.)
• [ ] Showed intellectual contribution to disciplinary conversations

---

Remember: Each discipline has evolved writing conventions for good reasons—they reflect epistemological commitments and audience expectations. Humanizing your writing means not just removing AI patterns but embodying authentic scholarly voice within your tradition.