科研 Skill
社会科学文献综述
lit-review
Runs a systematic literature review with search strategy, evidence table, thematic synthesis, and gap analysis. Use for systematic reviews or research syntheses.
SKILL.md
<skill_content>
Literature reviews in social science differ from biomedical reviews: gray literature (working papers, policy reports) carries significant weight, quasi-experimental designs are common, and the boundary between "published" and "working" research is fluid. This skill accounts for these disciplinary norms.
<mandatory_requirements>
</mandatory_requirements>
<research_question_framework>
For social science reviews, use the SPIDER framework (more flexible than PICO for non-clinical research):
- S (Sample): What population, group, or context?
- PI (Phenomenon of Interest): What intervention, policy, program, or phenomenon?
- D (Design): What study designs to include? (RCT, DID, IV, RDD, qualitative, mixed)
- E (Evaluation): What outcomes or impacts to evaluate?
- R (Research type): Quantitative, qualitative, or mixed methods?
Example: "What is the effect (E: learning outcomes, enrollment) of conditional cash transfer programs (PI) on educational attainment in low- and middle-income countries (S), as measured by experimental and quasi-experimental studies (D, R)?"
For policy evaluation reviews, also consider:
- Implementation context (what makes programs work or fail?)
- Heterogeneity (for whom, where, under what conditions?)
- Cost-effectiveness (what is the cost per unit of impact?)
- Scalability (do effects hold when programs scale?) </research_question_framework>
<database_guide>
Select at minimum 3 complementary databases appropriate for the domain:
Core Social Science Databases:
- Google Scholar: Comprehensive cross-disciplinary coverage, citation tracking
- SSRN: Working papers in economics, finance, law, political science
- NBER Working Papers: Leading economics research (often years before publication)
- JSTOR: Historical and current peer-reviewed articles across social sciences
- Web of Science / Scopus: Citation-indexed peer-reviewed literature
- EconLit: Economics-specific (AEA journals, books, working papers)
Policy and Development:
- J-PAL Evidence: RCTs in development economics
- 3ie Development Evidence Portal: Impact evaluations in international development
- World Bank Open Knowledge Repository: Policy research and working papers
- IMF Working Papers: Macroeconomics and public finance
- OECD iLibrary: Cross-country policy analysis
Specialized by Discipline:
- ERIC: Education research
- PubMed/PMC: Public health, health economics, epidemiology
- PolicyFile: US public policy research
- Campbell Collaboration: Systematic reviews in social sciences
- Cochrane Library: Health intervention reviews (when relevant)
- IZA Discussion Papers: Labor economics
- CEPR Discussion Papers: European economics research
- arXiv (econ, q-fin, stat): Quantitative methods, econometrics
Preprint and Open Access:
- RePEc/IDEAS: Economics working papers and articles
- OSF Preprints: Open science preprints across social sciences
- EdWorkingPapers (Annenberg): Education policy research
Database Selection Strategy:
- Primary database for breadth: Google Scholar or Web of Science
- Working paper repository: NBER, SSRN, or IZA (field-dependent)
- Specialized database: EconLit, ERIC, J-PAL, etc. (topic-dependent)
- Gray literature: World Bank, OECD, or government reports
- Citation chaining: Forward and backward from key papers </database_guide>
<quality_assessment>
Assess each study's methodological quality based on identification strategy and validity:
Study Design Hierarchy (for causal questions):
- RCTs (randomized controlled trials / field experiments)
- Regression discontinuity designs (RDD)
- Instrumental variables (IV)
- Difference-in-differences (DID) with pre-trends evidence
- Propensity score matching / synthetic control
- Observational with controls (OLS with covariates)
- Descriptive / correlational studies
- Qualitative / case studies
Quality Dimensions to Assess:
- Internal validity: How credible is the causal identification?
- Statistical power: Is the sample large enough to detect meaningful effects?
- External validity: How generalizable are the findings?
- Measurement quality: Are key variables well-measured?
- Pre-registration: Was the analysis pre-specified? (reduces p-hacking risk)
- Transparency: Are data and code available for replication?
- Robustness: Do results hold across specifications?
Quality Rating Scale:
- High: Strong identification strategy, adequate power, transparent methods, robust results
- Medium: Reasonable identification with some threats, adequate sample, partial robustness
- Low: Weak identification, small sample, or results sensitive to specification
Red Flags:
- P-values clustered just below 0.05 (possible p-hacking)
- No robustness checks or only one specification reported
- Effect sizes that are implausibly large
- Endogeneity concerns not addressed
- Missing data handled without sensitivity analysis
- Post-hoc subgroup analysis presented as primary finding </quality_assessment>
<prioritizing_papers>
Prioritize papers based on methodological rigor, venue quality, and influence:
Citation Count Thresholds (social science norms):
| Paper Age | Citations | Classification |
|---|---|---|
| 0-3 years | 10+ | Noteworthy |
| 0-3 years | 50+ | Highly Influential |
| 3-7 years | 50+ | Significant |
| 3-7 years | 200+ | Landmark Paper |
| 7+ years | 200+ | Seminal Work |
| 7+ years | 500+ | Foundational |
Note: Citation norms vary by subfield. Labor economics papers accumulate citations faster than political theory. Use these as rough guides.
Journal Tiers (Economics):
- Tier 1: American Economic Review, Quarterly Journal of Economics, Econometrica, Journal of Political Economy, Review of Economic Studies
- Tier 2: Review of Economics and Statistics, Journal of the European Economic Association, American Economic Journal (all), Journal of Finance, Journal of Labor Economics, Journal of Public Economics, Journal of Development Economics, Economic Journal
- Tier 3: Respected field journals (Journal of Human Resources, Journal of Health Economics, Journal of Urban Economics, etc.)
Journal Tiers (Political Science):
- Tier 1: American Political Science Review, American Journal of Political Science, Journal of Politics
- Tier 2: Comparative Political Studies, World Politics, International Organization, British Journal of Political Science
- Tier 3: Field journals (Political Analysis, Political Behavior, etc.)
Journal Tiers (Sociology):
- Tier 1: American Sociological Review, American Journal of Sociology
- Tier 2: Social Forces, Demography, Sociology of Education
- Tier 3: Field journals
Working Papers:
- NBER Working Papers carry significant weight (peer network vetting)
- SSRN/IZA papers should be assessed on methodology, not venue
- Check if working papers have been subsequently published
Identifying Seminal Papers:
- Cited by many of the other papers you find (appears across reference lists)
- Introduced a methodology now widely used (e.g., Angrist & Krueger 1991 for IV)
- Published in Tier-1 venue with high citation count
- Referenced in textbooks and survey articles
- Written by researchers recognized as field leaders </prioritizing_papers>
<search_techniques>
Boolean Search Construction:
- AND: narrows (both terms required)
- OR: broadens (either term)
- Quotes: exact phrase ("conditional cash transfer")
- Wildcards: education* matches educational, education, educating
Example for a CCT review: ("conditional cash transfer" OR "CCT" OR "cash transfer program") AND ("education" OR "school enrollment" OR "attendance" OR "learning outcomes") AND ("developing countries" OR "low-income" OR "Global South")
Citation Chaining:
- Forward citation search: Find papers citing a key paper (Google Scholar "Cited by")
- Backward citation search: Review references of key papers
- Snowball sampling: Start with 3-5 seminal papers, follow citation networks
- Prioritize papers appearing in multiple reference lists (likely foundational)
Search Refinement:
- Pilot search: Run broad terms, review first 50 results
- Note recurring keywords, author names, and journal names
- Refine search terms based on pilot results
- Run refined search across all selected databases
- Document each iteration for reproducibility
Gray Literature Search:
- NBER: Browse by program (Labor Studies, Public Economics, etc.)
- SSRN: Search by keyword and sort by downloads or citations
- Government reports: Search agency websites directly
- Conference proceedings: ASSA/AEA meetings, APPAM, BREAD
- Dissertations: ProQuest Dissertations for emerging research </search_techniques>
Phase 1: Planning and Scoping (Cell 1)
- Define research question using SPIDER framework
- Develop search terms with synonyms and Boolean operators
- Select minimum 3 complementary databases
- Set date range, language, and geographic constraints
- Define inclusion/exclusion criteria (study design, population, outcomes)
- Specify review type: systematic, scoping, narrative, or meta-analysis
Phase 2: Systematic Search and Source Collection (Cell 2)
- Execute search across all selected databases
- Document search strings, dates, and result counts for each database
- Aggregate results and remove duplicates
- Build structured DataFrame of all identified sources
- Record how each source was found (which search, which database)
- Conduct citation chaining from key papers
Phase 3: Screening and Quality Assessment (Cell 3)
- Apply inclusion/exclusion criteria systematically
- Document exclusion reasons with counts
- Assess methodological quality of each included study
- Assign quality rating (high/medium/low)
- Create screening flow diagram (records found -> deduplicated -> screened -> included)
Phase 4: Thematic Synthesis and Gap Analysis (Cell 4)
- Identify 3-6 major themes across included studies
- Synthesize findings within each theme (NOT study-by-study)
- Compare effect sizes and directions across studies
- Weight synthesis by study quality
- Identify consensus findings, contested claims, and gaps
- Note methodological patterns and limitations across studies
Phase 5: Summary, Implications, and References (Cell 5)
- Synthesize key takeaways (what does the weight of evidence suggest?)
- Discuss implications for policy, practice, or future research
- List specific gaps that future research should address
- Acknowledge limitations of the review itself
- Provide properly formatted reference list
<output_format>
Each phase becomes a separate notebook cell. Pure Python code, no decorators or wrappers.
Cell 1 -- Search Strategy:
import pandas as pd
from datetime import date
# LITERATURE REVIEW: [Topic]
# ============================================================
search_strategy = {
"research_question": "What is the effect of [intervention] on [outcome] in [population/context]?",
"framework": "SPIDER",
"sample": "[population or context]",
"phenomenon": "[intervention, policy, or phenomenon]",
"design": "[RCT, DID, IV, RDD, mixed methods, etc.]",
"evaluation": "[outcomes to measure]",
"research_type": "[quantitative, qualitative, mixed]",
"search_terms": [
'("term1" OR "synonym1") AND ("term2" OR "synonym2")',
'"exact phrase" AND (outcome1 OR outcome2)',
],
"databases": [
"Google Scholar",
"NBER Working Papers",
"SSRN",
# Add domain-specific: EconLit, ERIC, J-PAL, 3ie, etc.
],
"date_range": "2010-2025",
"language": "English",
"inclusion_criteria": [
"Peer-reviewed articles or working papers from recognized institutions",
"Empirical studies with quantitative outcome measures",
"Study designs: RCT, DID, IV, RDD, or high-quality observational",
"Population: [specify]",
],
"exclusion_criteria": [
"Purely theoretical or opinion pieces without empirical evidence",
"Studies with sample size < [threshold]",
"Non-peer-reviewed sources without institutional affiliation",
"Studies outside geographic/temporal scope",
],
"review_type": "systematic", # systematic, scoping, narrative, meta-analysis
"date_executed": str(date.today()),
}
# Display search protocol
print("=" * 60)
print("SEARCH PROTOCOL")
print("=" * 60)
for key, value in search_strategy.items():
if isinstance(value, list):
print(f"\n{key}:")
for _item in value:
print(f" - {_item}")
else:
print(f"\n{key}: {value}")
Cell 2 -- Source Collection:
import pandas as pd
# Search Results by Database
# ============================================================
# Document search execution for reproducibility
search_log = [
{"database": "Google Scholar", "date_searched": "2025-01-15",
"search_string": '("conditional cash transfer" OR CCT) AND education',
"results_found": 1240, "after_screening": 45},
{"database": "NBER", "date_searched": "2025-01-15",
"search_string": "conditional cash transfer education",
"results_found": 28, "after_screening": 12},
# ... more databases
]
search_df = pd.DataFrame(search_log)
print("SEARCH EXECUTION LOG")
print("=" * 60)
print(search_df.to_string(index=False))
print(f"\nTotal results: {search_df['results_found'].sum()}")
print(f"After screening: {search_df['after_screening'].sum()}")
# Evidence Table
# ============================================================
sources = [
{
"authors": "Author et al.",
"year": 2020,
"title": "Title of the study",
"journal": "Journal Name or Working Paper Series",
"methodology": "RCT", # RCT/DID/IV/RDD/PSM/observational/qualitative
"sample_size": "N=5,000",
"geographic_scope": "Mexico",
"key_finding": "CCT increased enrollment by 8 percentage points",
"effect_size": "8pp (95% CI: 5-11pp)",
"quality": "high", # high/medium/low
"relevance": "high", # high/medium/low
"source_db": "Google Scholar",
"doi_or_url": "https://doi.org/...",
},
# ... more sources
]
lit_table = pd.DataFrame(sources)
lit_table = lit_table.sort_values(["quality", "relevance", "year"],
ascending=[False, False, False])
# Summary statistics
print(f"\nTotal included studies: {len(lit_table)}")
print(f"\nBy methodology:")
print(lit_table["methodology"].value_counts().to_string())
print(f"\nBy quality rating:")
print(lit_table["quality"].value_counts().to_string())
lit_table
Cell 3 -- Screening Flow and Quality Assessment:
# SCREENING FLOW
# ============================================================
print("SCREENING FLOW DIAGRAM")
print("=" * 60)
flow = {
"Records identified through database searching": 1500,
"Additional records from citation chaining": 45,
"Records after deduplication": 1200,
"Records screened (title/abstract)": 1200,
"Records excluded at screening": 1050,
"Full-text articles assessed": 150,
"Full-text excluded (with reasons)": 108,
"Studies included in synthesis": 42,
}
for _stage, _count in flow.items():
print(f" {_stage}: n = {_count}")
# Exclusion reasons
print("\nExclusion reasons (full-text stage):")
_exclusion_reasons = {
"Wrong population/context": 35,
"Wrong outcome measures": 28,
"Insufficient methodology": 22,
"Duplicate/superseded version": 15,
"Not available in English": 8,
}
for _reason, _n in _exclusion_reasons.items():
print(f" - {_reason}: n = {_n}")
# QUALITY ASSESSMENT
# ============================================================
print("\n" + "=" * 60)
print("QUALITY ASSESSMENT SUMMARY")
print("=" * 60)
_quality_summary = lit_table.groupby(["methodology", "quality"]).size().unstack(fill_value=0)
print(_quality_summary)
# Flag any quality concerns
_low_quality = lit_table[lit_table["quality"] == "low"]
if len(_low_quality) > 0:
print(f"\nWARNING: {len(_low_quality)} low-quality studies included.")
print("These will be noted but down-weighted in synthesis.")
Cell 4 -- Thematic Synthesis:
# THEMATIC SYNTHESIS
# ============================================================
print("=" * 60)
print("EVIDENCE SYNTHESIS")
print("=" * 60)
# Theme 1
print("\n--- Theme 1: [Theme Name] ---")
print("Studies: [Author1 (Year), Author2 (Year), ...]")
print("Finding: [Synthesized finding across studies, not study-by-study]")
print("Strength of evidence: [Strong/Moderate/Weak]")
print("Consistency: [Consistent/Mixed/Contradictory]")
# Theme 2
print("\n--- Theme 2: [Theme Name] ---")
# ... same structure
# Consensus vs. contested
print("\n" + "=" * 60)
print("EVIDENCE MAP")
print("=" * 60)
print("\nConsensus findings (supported by multiple high-quality studies):")
print(" 1. ...")
print(" 2. ...")
print("\nContested or mixed findings:")
print(" 1. ... [Author1 finds X, but Author2 finds Y; difference may be due to ...]")
print("\nKnowledge gaps:")
print(" 1. ... [No studies examine ...]")
print(" 2. ... [Limited evidence on ... subpopulation]")
print(" 3. ... [Methodological gap: no RCTs on ...]")
print("\nMethodological patterns:")
print(f" - Most common design: {lit_table['methodology'].mode().iloc[0]}")
print(f" - Geographic concentration: {lit_table['geographic_scope'].value_counts().head(3).to_string()}")
print(" - Missing designs: [What study types are absent?]")
Cell 5 -- Summary and References:
# SUMMARY AND IMPLICATIONS
# ============================================================
print("=" * 60)
print("SUMMARY")
print("=" * 60)
print("\nKey Takeaways:")
print(" 1. [What does the weight of evidence suggest?]")
print(" 2. [What is the range of effect sizes?]")
print(" 3. [What conditions moderate the effect?]")
print("\nImplications for Policy:")
print(" - ...")
print("\nImplications for Future Research:")
print(" - [Specific gaps to fill]")
print(" - [Methodological improvements needed]")
print(" - [Understudied populations or contexts]")
print("\nLimitations of This Review:")
print(" - [Search limitations: databases not searched, language restriction]")
print(" - [Potential publication bias]")
print(" - [Scope limitations]")
# REFERENCES (APA 7th Edition)
# ============================================================
print("\n" + "=" * 60)
print("REFERENCES")
print("=" * 60)
# Generate formatted references from the evidence table
for _idx, _row in lit_table.sort_values("authors").iterrows():
_ref = f"{_row['authors']} ({_row['year']}). {_row['title']}. {_row['journal']}."
if _row.get('doi_or_url'):
_ref += f" {_row['doi_or_url']}"
print(f"\n{_ref}")
</output_format>
<citation_format>
Default: APA 7th Edition (standard for social sciences)
In-text citations:
- Single author: Smith (2023) or (Smith, 2023)
- Two authors: Smith & Jones (2023) or (Smith & Jones, 2023)
- Three or more: Smith et al. (2023) or (Smith et al., 2023)
- Multiple citations: (Angrist, 2001; Duflo et al., 2011; Heckman, 2010)
- With page: (Smith, 2023, p. 42) or (Smith, 2023, pp. 42-45)
Reference list:
- Alphabetical by first author surname
- Hanging indent format
- Include DOI as URL when available
Journal article: Smith, J. D., Johnson, M. L., & Williams, K. R. (2023). Title of article in sentence case. Title of Periodical in Title Case, 22(4), 301-318. https://doi.org/10.xxx/yyy
Working paper: Smith, J. D. (2023). Title of paper (NBER Working Paper No. 31234). National Bureau of Economic Research. https://doi.org/10.xxx/yyy
Book: Author, A. A. (Year). Title of work: Subtitle. Publisher.
Chapter: Author, A. A. (Year). Title of chapter. In E. E. Editor (Ed.), Title of book (pp. xx-xx). Publisher.
Alternative: Chicago Author-Date (common in political science and sociology) follows similar in-text conventions but differs in reference formatting. Use whichever the user's field prefers.
Store all citation data in the DataFrame for programmatic formatting. </citation_format>
<common_mistakes>
</common_mistakes>
<interpretation_guide>
<interpreting_results>
- Weight evidence by study quality, not just quantity of studies
- Report effect size ranges across studies, not just average effects
- Note when effects are heterogeneous by context, population, or implementation
- Distinguish between absence of evidence and evidence of absence
- Consider publication bias: significant results are overrepresented in published literature </interpreting_results>
<red_flags>
- All included studies find effects in the same direction (possible publication bias)
- Effect sizes shrink as study quality increases (common pattern indicating bias)
- Geographic concentration (findings from one country may not generalize)
- Temporal clustering (field may have evolved since most studies were conducted)
- Funnel plot asymmetry when enough studies for meta-analysis </red_flags>
<next_steps>
- Strong consensus with high-quality evidence -> Report with confidence, note remaining gaps
- Mixed evidence -> Investigate sources of heterogeneity (context, methods, populations)
- Limited evidence -> Characterize what is known, emphasize need for more research
- Contradictory high-quality evidence -> Present both sides, analyze why results diverge </next_steps>
</interpretation_guide>
</skill_content>
示例提示词
帮我按系统性综述方法整理这批跨学科文献的检索策略与证据表。
所属插件
安装该插件会一并获得同领域的 58 个 Skill / MCP。
同插件的其他科研 Skill
- Mesa社会主体建模Builds Mesa agent-based models of opinion dynamics, segregation, social contagion, and network diffusion. Use when simulating emergent social behavior from agent rules.
- Mesa3主体建模引擎Builds agent-based models with the Mesa 3 AgentSet API, cell spaces, and batch_run sweeps for segregation, diffusion, and cooperation. Use for a Mesa 3.x ABM build or debug.
- 观察数据因果推断Estimates causal effects with DiD, IV, RDD, panel fixed effects, and propensity scores via statsmodels, linearmodels, DoWhy, and EconML. Use for observational or quasi-experimental causal estimation.
- 效应量元分析Pools effect sizes across studies with fixed/random-effects models, heterogeneity stats, forest/funnel plots, and publication-bias tests. Use for a systematic review’s quantitative synthesis.
- 缺失数据插补处理Handles missing data by naming the MCAR/MAR/MNAR mechanism, then applying multiple imputation (MICE) or FIML instead of single imputation. Use when a dataset has missing values to treat.
- 多层线性混合模型Fits mixed-effects/multilevel models (random slopes, ICC, GLMMs) for nested or longitudinal data via statsmodels, bambi, or R lme4/brms. Use for grouped data with within- vs between-cluster questions.
- 定性比较分析QCARuns crisp-, multi-value, and fuzzy-set Qualitative Comparative Analysis via R’s QCA package. Use for configurational or necessary/sufficient-condition research.
- 质性编码与主题分析Builds codebooks and computes intercoder reliability (Krippendorff’s alpha, kappa) for thematic or content coding, including human-vs-LLM double coding. Use when coding text or reporting reliability.
- 社会网络结构分析Computes network centrality, community detection, and ERGM inferential models matched to the research question via networkx/igraph. Use for social-network analysis beyond basic visualization.
- 社科研究设计把关Routes a study to its research design (experiment, quasi-experiment, observational, qualitative) and pins the identifying assumption first. Use when choosing a study design.
- 社科推论结论审核Audits inferential claims against the design before publication: unwarranted causal language, p-value/CI misreadings, and multiple-comparison issues. Use when checking a results section.
- 量表测量质量把关Checks that a scale’s reliability, validity, and measurement invariance are established before trusting or comparing it across groups. Use when adopting or reporting a survey instrument.
- 社科研究流程总控Coordinates a study through design, measurement, sampling, analysis, and inference gates with a shared Design Passport. Use for an end-to-end methodology review of a whole study.
- 质性研究可信度把关Gates qualitative trustworthiness before write-up: positionality, credibility, transferability, dependability, confirmability per Lincoln and Guba. Use before writing ethnographic findings.
- 抽样方法与样本量把关Checks that the sampling frame, method, and size match the inference paradigm before data collection. Use when planning recruitment or judging sample generalization.
- 复杂抽样调查分析Analyzes complex-sample survey data with design-based inference: weights, strata, PSUs, and calibration via Python svy or R survey/srvyr. Use for weighted surveys like GSS, ANES, or DHS.
- 文本数据定量分析Analyzes text as social-science data with topic modeling, embeddings, and validated dictionary or classification methods. Use for computational text analysis beyond a keyword count.
- 政策议题框架分析Runs anthropological framing and metaphor analysis on contested-policy documents, producing actor-attributed CSVs for NVivo import. Use for policy-framing or metaphor-analysis coding.
- 跨国比较政治分析Merges QoG, Polity5, and Freedom House data and runs cross-sectional OLS with fixed effects and regime classification. Use for comparative political-institution research.
- 计算社会学方法Covers social-media data collection, bot detection, network homophily, echo-chamber analysis, and conjoint survey experiments. Use for large-scale digital-trace sociology research.
- 冲突事件数据分析Downloads ACLED/UCDP conflict data and runs spatio-temporal clustering, conflict-onset logit panels, and fatality estimation. Use for quantitative armed-conflict research.
- 联合分析调查实验Designs conjoint survey experiments and estimates AMCEs via OLS, marginal means, and respondent-level heterogeneity. Use when designing or analyzing a choice-based conjoint experiment.
- 统计方法教学规范Defines the baseline statistical teaching and reproducibility standards shared across this analysis skill family. Use as the methodology reference other skills in the family build on.
- 人口统计方法分析Computes life tables, Leslie-matrix population projections, Lee-Carter mortality forecasts, and Kitagawa decomposition. Use for formal demographic analysis of mortality or population structure.
常见问题
社会科学文献综述能做什么?
Runs a systematic literature review with search strategy, evidence table, thematic synthesis, and gap analysis. Use for systematic reviews or research syntheses。
如何安装社会科学文献综述?
在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「社会科学文献综述」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。
社会科学文献综述需要哪些权限?
该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。
社会科学文献综述经过官方审核吗?
是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。