Build or revise literature reviews for economics and management empirical papers, especially when the review must be organized by theme, viewpoint, method, or mechanism rather than by article list; use when collecting text-identification methods, preparing research-gap and research-value transitions, or continuing the ai-innovation-lean project's theory-driven empirical writing under strict process boundaries.
Do not write novelty or gap claims beyond what these materials support.
2. Do not freewrite the whole review at once
Default staged order:
fix the chapter or section containers
write only the opening and closing paragraphs first
extract reusable paragraph套路 from reference papers
allocate paragraph套路 across the remaining middle paragraphs
plan sentence functions inside each paragraph
only then draft prose
For this project, the literature review should be built slowly and structurally, not generated as one full block.
3. Organize the review by unit, not by article sequence
Do not default to abstract grouping labels such as 第一类文献 or 研究线索. In Chinese economics and management journals, the review usually starts from the object, difficulty, or measurement issue itself, then moves to representative studies and only afterwards compresses them into the author's judgment.
Default review unit order:
theme
representative viewpoints or findings
method or measurement evolution
what remains unresolved
research gap and research value
Do not write A文献发现... B文献发现... C文献发现... unless the user explicitly wants an annotated bibliography.
4. Use a fixed paragraph logic
Preferred Chinese-journal paragraph order:
start from the concrete object, empirical difficulty, or mechanism at stake
show how existing studies handle that object
compress what is already relatively clear
identify the insufficiency that matters for this paper
bridge to the next block or to the paper
Avoid English-style category-first review prose such as "the first stream" or "the second line of research" unless the user explicitly requests that style.
Fixed review-paragraph identities should be planned separately:
本节首段: define the review object, the bottleneck, and the section order
本小节首段: define the block object and why this block matters
证据铺陈段: place representative studies and compress a local consensus
承上启下段: shift from one object to the next without changing the paper's main question
本小节末段: extract the insufficiency and hand off to the next block
本节末段: unify the gap and hand off to the next chapter
For each fixed paragraph identity, also set a sentence-role sequence, not only a paragraph label.
Use the same planning template throughout the review section:
段落身份
段落职责
段落套路
句子序列
当前义务
Preferred Chinese-journal paragraph order:
start from the concrete object, empirical difficulty, or mechanism at stake
show how existing studies handle that object
compress what is already relatively clear
identify the insufficiency that matters for this paper
bridge to the next block or to the paper
Avoid English-style category-first review prose such as "the first stream" or "the second line of research" unless the user explicitly requests that style.
For each literature block:
open with the problem this block addresses
summarize the main findings or viewpoints
explain the dominant methods or measurement approaches
point out the unresolved issue that matters for this paper
land on why the next block or the paper's question follows naturally
5. Learn paragraph套路 from high-quality references
When the user provides a Chinese reference article, decompose each paragraph into:
paragraph role
opening move
sentence-function sequence
landing move
reusable套路 label
Assume that academic paragraph organization uses a small number of recurring套路. Reuse those套路 deliberately instead of improvising every paragraph from scratch.
6. Treat the remaining paragraph plan as a distribution problem
After the chapter and section opening/closing paragraphs are fixed, decide how many remaining paragraphs are needed and assign套路 with variation.
Control for repetition:
avoid identical opening moves in adjacent paragraphs
avoid identical landing moves in adjacent paragraphs
avoid turning every paragraph into topic sentence + literature list + gap
keep the rhetorical rhythm varied but the logic stable
7. For text-identification methods, always build a method family map
Default method families are in references/text-identification-methods.md.
Use those families to compare:
what the method identifies
where it is commonly used
why it is useful here
what validation it requires
8. Keep project positioning stable
For ai-innovation-lean:
the paper is a 理论驱动的实证文章
Chapters 1 and 2 may be polished before later chapters are fixed
Chapters 3+ must not be written as formal manuscript text until literature, variables, data, and identification have been fixed by authority materials
Read references/ai-innovation-lean-boundaries.md when working on this project.
Output pattern
When asked to help with a literature review, usually produce:
a thematic map
a paragraph套路 map when a reference article is available
a sentence-function plan for key paragraphs
a gap statement that stays within the verified evidence
if useful, a draft paragraph or section that follows the project's journal style
References
For review structure and transitions: read references/review-architecture.md
For text-identification methods: read references/text-identification-methods.md
For this project's boundaries: read references/ai-innovation-lean-boundaries.md