3.2 Writing the Introduction and Background¶

What you will learn on this page

A practical workflow for drafting an Introduction (Japanese skeleton → English conversion → structure check → refinement)
Using AI tools to organize prior research: benefits and limits
Techniques for summarizing and synthesizing prior work
How to make your research gap more convincing
How to write research questions (RQs) in English
How to adjust the length of an Introduction
How to evaluate and revise AI-generated drafts

How this page fits in

The components of an Introduction (the hourglass flow) and the three moves of the CARS model are explained in
2.1 The IMRaD Structure and
2.2 Checking Conventional Expressions with Corpora.
This page assumes that knowledge and focuses on a hands-on workflow for writing an Introduction with AI.

A workflow for drafting an Introduction¶

If you align your workflow with the three moves of the CARS model, your structure becomes much clearer.

Workflow step	Related CARS move(s)	What you do
Step 1: Create a Japanese skeleton	Moves 1–3	Write 1–2 sentences for each core element in Japanese
Step 2: Ask AI to convert it into English	—	Convert the skeleton into an academic English paragraph
Step 3: Ask AI to check structure	Moves 1–3 presence check	Check whether the elements exist and appear in a reasonable order
Step 4: Strengthen the gap	Deepen Move 2	Make the gap specific and strengthen logical linkage
Step 5: Synthesize prior work	Refine Moves 1–2	Move from listing to integrated description

We explain each step below.

AI tools for organizing prior research¶

Before Step 1 (creating the Japanese skeleton), it helps to organize relevant prior studies efficiently. Below are tools that can support that work.

NotebookLM: reading and summarizing papers¶

Google's NotebookLM lets you upload multiple PDFs (and other files) as sources and then ask questions, generate summaries, and compare claims in a chat-style interface.

It can also generate slide decks and structured outputs such as comparison tables. Depending on settings and availability, it may also provide overview outputs such as narrated explanations. It is useful when you want to speed up literature organization.

Typical uses of NotebookLM

Upload multiple papers and ask cross-cutting questions such as: "What are the benefits of using ChatGPT for language learning and teaching?"
Extract recurring themes and repeated claims (pattern finding)
Identify areas that remain unclear in prior work (potential gaps)
Also useful beyond research: upload institutional documents and confirm key points of guidelines

NotebookLM assigns source references in its responses, which makes it easier to trace which paper a claim comes from.

NotebookLM

BiblioSeek: analyze, search, and summarize PDFs¶

BiblioSeek is a web application that lets you upload multiple PDF papers and use AI to analyze, search, and summarize them.

Supports switching between Japanese and English
Allows cross-document search and question-answering across uploaded PDFs
Uses the Gemini API (an API key is required)

PDF ThemeSort: theme-based classification of papers¶

PDF ThemeSort automatically classifies multiple PDF papers by theme.

You can either let AI propose themes automatically, or specify your own themes
You can switch between a faster mode and a higher-quality mode
It is helpful when you want to group papers quickly before writing a literature review

Other utilities¶

Attention: An AI-powered tool that extracts and analyzes themes from free-text data. It can also be used as an aid for organizing the background and motivation of your research.
PDF Auto-Rename Tool: Detects DOIs in PDF files and automatically renames them to "Author – Year – Title.pdf" format. Useful for managing large collections of papers.

Understand the limits of AI tools¶

These tools can greatly speed up literature organization, but keep the following limits in mind.

What they can do	What they cannot do / requires caution
Summarize and compare papers	Guarantee the accuracy of summaries (always verify against the original)
Classify papers by theme	Judge whether the classification criteria are appropriate (requires your domain expertise)
Suggest potential gaps	Evaluate the scholarly importance of a gap (requires knowledge of the field's context)
Extract patterns	Discover important papers you have not yet found (comprehensive searching must be done separately)

Use specialized tools for literature search

Searching for papers with general-purpose generative AI (such as ChatGPT) can prioritize open-access papers (including predatory journals), lack the depth needed for a literature review, or generate fabricated bibliographic information. For literature search, use the specialized tools listed below.

Note that as of February 2026, Google Scholar added a feature called "Scholar Labs," which enables AI-powered paper search. This is a promising development worth watching.

Tool	Purpose	URL	Notes
Perplexity	AI search engine	perplexity.ai	Answers with source citations
Elicit	AI-powered paper search and summarization	elicit.org	Finds papers from research questions
Consensus	Evidence search across academic papers	consensus.app	Organizes findings in Yes/No format
SciSpace	Paper reading support and literature discovery (PDF chat)	scispace.com	Ask questions to PDFs and get cited key points
Connected Papers	Visualization of related papers	connectedpapers.com	Visual literature map
Research Rabbit	Paper recommendations	researchrabbit.ai	Integrates with Zotero

Choosing the right literature search tool (by purpose)

Broad keyword-based search: Google Scholar (start here to find foundational references)
Discovering related work via citation networks: Connected Papers / Research Rabbit — especially effective when you have a few "seed papers"
Quick summaries with evidence: Consensus (answers come with source references by design)
Structured comparison and extraction: Elicit (creates information extraction tables across multiple papers)

Note: All tools can miss relevant papers and produce inaccurate summaries. Always verify your final judgments against the original text (abstract, body, methods).

From literature organization to Introduction drafting

Use PDF ThemeSort to classify your collected papers by theme
Use BiblioSeek or NotebookLM to grasp the key points across each theme
Summarize the key points for each theme in Japanese notes
Use those notes to draft your Introduction following Steps 1–5 below

A good rule is to combine a search tool with a reading tool: search broadly, then read deeply.

Reference: how to progress a literature review¶

AI tools can make literature organization more efficient, but the literature review itself follows a systematic process. The Boston College Libraries guide describes the literature review as a six-stage process:

Stage	What you do (in brief)	Source (corresponding page)
1. Define the scope	Decide what to include and what to exclude	Scope of Review
2. Search	Gather papers from databases and other sources	Finding Information
3. Record	Save citation details and key points in a reusable format	Recording Information
4. Evaluate	Assess each study's strengths, weaknesses, evidence, and limitations	Evaluating Information
5. Organize	Rearrange by theme or other criteria to create a structure	Organizing the Review
6. Write	Draft the text so that the flow of research and the gap are clear to the reader	Writing the Literature Review

A literature review is not simply summarizing and listing collected studies. Think of it as the work of presenting the trajectory of research and the "gaps that remain" in a form that communicates clearly to readers.

This is where a critical difference from AI-based literature search emerges. AI search tends to cast a wide but shallow net, often centered on open-access literature. Its raw output is rarely at a level that can serve as a literature review. Whether you invest the effort to organize this material systematically is arguably what determines whether your paper achieves genuine depth.

The slides below present the literature review as a continuous process — from topic exploration → search and collection → storage and organization → analysis and synthesis → writing (the CLR: Comprehensive Literature Review Process). They explain the tools available at each stage (reference managers, CAQDAS, outliners, etc.) and how coding and categorization are key to integrating information. There is no single method that works for everyone, so it is important to try different organizational strategies and find the approach that works best for you.

← → keys to navigate slides 📎 Open in full screen

Step 1: Create an L1 (Japanese) skeleton¶

The most stable way to draft an Introduction with AI is to start with an L1 (in this case, Japanese) skeleton.

A practical skeleton uses four elements that map naturally onto CARS Moves 1–3.

Element	Related CARS move	What to write
What is important in this field	Move 1: Establishing a territory	The social or academic significance of the research area
What prior work has found	Move 1: Establishing a territory	An overview of key findings
What is still unknown (the gap)	Move 2: Establishing a niche	Issues or unanswered questions in prior work
What this study will do	Move 3: Occupying the niche	The purpose and approach of the study

Write 1–2 sentences for each in Japanese. Keep them simple and explicit.

Why start in L1?

When you start directly in English, it becomes easy to accept AI wording that you cannot fully evaluate.
If you stabilize the content in L1 first, you can keep content ownership while using AI mainly for language support.

Be specific at the skeleton stage

If you write only generalities such as "Vocabulary learning is important" at this stage, the English output will remain equally abstract. Include information specific to your research, such as "the gap between receptive and productive vocabulary sizes among Japanese university students."

Example of a Japanese skeleton¶

Below is an example format. Replace the bracketed parts with your topic.

Background: 「近年、[研究領域] は [理由] により重要性が高まっている。」
Prior work: 「これまでの研究では、主に [A] と [B] が検討されてきた。」
Gap: 「しかし、[C] については十分に検討されていない。」
Purpose: 「本研究は [C] を対象に、[目的] を明らかにする。」

Here is a more concrete example from applied linguistics:

(1) Importance: 語彙知識は第二言語の読解力・聴解力と強く関連しており、
    語彙習得研究は応用言語学の中心的なテーマの一つである。
(2) Prior work: これまでの研究では、語彙サイズテスト（VST）を用いて
    受容語彙サイズを測定する研究が多く蓄積されている。
(3) Gap: しかし、受容語彙と産出語彙の乖離がどの習熟度段階で
    顕著になるかについては、十分に検討されていない。
(4) Purpose: 本研究は、日本人大学生200名を対象に、受容・産出語彙サイズの
    関係を習熟度別に分析する。

You will convert this into an English paragraph in Step 2.

Step 2: Ask AI to convert it into English¶

In Step 2, your goal is accuracy first. Ask AI to convert the Japanese skeleton into English while preserving meaning and claim strength.

Prompt: Japanese skeleton → English Introduction paragraph

Convert the Japanese text below into a natural academic English paragraph for an Introduction.

Requirements:
- Do not add new claims, examples, or citations
- Keep the strength of claims (do not make the gap sound stronger than in Japanese)
- Keep the paragraph order: background → prior work → gap → purpose
- Aim for clear sentences with short subjects and early verbs
- Include appropriate hedging expressions
- Do not insert specific citations (author name, year)

Japanese:
[Paste the 4-part Japanese skeleton here]

If you want the language level controlled, add a line such as "Write at around CEFR B2", "Use accessible academic English" or "Write in plain English."

Watch out for fabricated citations

AI sometimes inserts fictional citations such as "Smith (2020)" on its own. Either prohibit this explicitly in your prompt, or always check the output. Add real citations manually after verifying the original sources yourself.

Step 3: Ask AI to check structure¶

After conversion, check whether the paragraph still contains the intended elements and whether the order is reasonable.

Prompt: check the four elements

Check whether the English paragraph below includes these four elements:
(1) background/importance, (2) summary of prior work, (3) research gap, and (4) purpose of the present study.

For each element, quote the sentence(s) that correspond to it.
If any element is missing, point it out.
Do not rewrite the paragraph.

[Paste the English paragraph here]

If you want a check aligned with the CARS model, you can use a CARS-move check prompt and ask AI to label Move 1, Move 2, and Move 3 sentences. (When linking to the CARS section, use the English anchor convention described in this site.)

Step 4: Strengthen the gap¶

Many Introductions fail not because English is weak, but because the gap is vague. Strengthening the gap means making it specific, researchable, and logically necessary.

A strong gap typically has two features:

Specificity: what exactly is missing, in which context, for which population, with which method or outcome
Necessity: why that missing piece matters for theory, practice, or both

Common weak gap patterns¶

Weak gap phrasing	Why it is weak	How to improve
Few studies have examined X.	Simply saying "few" does not convey why the research is needed	Add what we would learn by examining X and why it matters
More research is needed.	A catch-all phrase that applies to any study	Specify what kind of research is needed and why
This area is under-researched.	Being under-researched does not automatically mean the topic is important	Show what problems arise as a consequence of the gap

Prompt: check the persuasiveness of your gap

Identify the sentence(s) in the Introduction below that describe the research gap.
Then evaluate them on the following criteria:

(1) Is the gap specific (not just "under-researched" but clear about what, why, and for whom)?
(2) Does the gap logically connect to the purpose of this study?
(3) Would a reader be convinced that this study is necessary?

If improvements are needed, indicate the direction only (do not rewrite).

[Paste Introduction here]

Prompt: make the gap more specific (without exaggeration)

Improve the research gap sentence(s) below.

Requirements:
- Make the gap more specific (what, who, where, under what conditions)
- Do not exaggerate (avoid claims like 'no studies exist' unless supported)
- Keep the claim strength appropriate
- Provide 2 alternative gap sentences and a one-line reason for each

Gap sentence(s):
[Paste your gap sentences here]

Step 5: Synthesize prior work (Synthesis)¶

As noted in the Reference section on literature reviews, integration is the key to an effective literature review. Here, we look at how to express your organized prior research as actual English prose in an Introduction.
(See also: "avoiding the laundry list literature review")

Listing vs. synthesis¶

Style	Example	Assessment
Listing (avoid)	Smith (2020) found X. Jones (2021) found Y. Lee (2022) found Z.	✗: Readers must figure out the connections themselves
Synthesis (recommended)	Several studies have found that X is related to Y (Smith, 2020; Jones, 2021), although the effect may depend on Z (Lee, 2022).	✓: The author organizes findings around a theme

The listing style also tends to produce monotonous repetition of said / found, as noted in the discussion of reporting verbs in 2.2. Be conscious of avoiding this pattern.

Prompt: convert listing to synthesis

The following passage from an Introduction lists prior studies one by one.
Rewrite it as an integrated paragraph organized by themes, commonalities, and differences.

Requirements:
- Keep all citations (author, year) as they are
- Do not add new references or claims
- Use transition expressions to signal shifts between themes
- Let the gap emerge naturally from the synthesis

[Paste the listing-style passage here]

Frameworks for organizing prior research¶

As a preparatory step before synthesizing, it helps to create Japanese notes using the following framework. This makes it much easier to write an integrated Introduction.

Common grouping lenses include:

By method (experimental, survey, corpus, qualitative)
By population/context (age, proficiency, setting, country)
By construct (motivation, engagement, accuracy, etc.)
By finding pattern (consistent results vs mixed results)

Prompt: organize prior research using a framework

Read the summaries of the 5 papers below and organize them according to this framework:
(1) Common theme: the shared question addressed by this group of studies
(2) Consistent findings: conclusions that agree across studies
(3) Inconsistencies: points where results differ or are debated
(4) Methodological trends: methods commonly used across studies
(5) Remaining gaps: questions, populations, or methods not yet examined

Write each item in Japanese, using bullet points.
For the gaps, also suggest 1–2 candidate research questions that could follow from them.

[Paste summaries of each paper here]

Prompt: propose synthesis categories (you choose the final ones)

Based on the brief study notes below, propose 3–4 synthesis categories
and rewrite the literature summary as an integrated paragraph.

Requirements:
- Do not invent results not in the notes
- Keep citations only where provided
- Use cautious academic language (avoid overclaiming)

Study notes:
[Paste bullet notes: author-year + 1–2 lines each]

For conventional phrases and reporting verb choices, check:
2.2 Checking Conventional Expressions with Corpora

Writing research questions (RQs) in English¶

Clear RQs help readers understand your study design. They also help you write the purpose sentence in Move 3.

Common lead-in expressions¶

Pattern	Example
Direct presentation	The following research questions guided this study:
Derived from purpose	To address this gap, the present study investigated the following research questions:
Presented as hypotheses	Based on the literature, the following hypotheses were formulated:

Practical guidelines for writing RQs¶

Write RQs as interrogative sentences (To what extent does X affect Y?)
Write hypotheses as declarative sentences (It was hypothesized that X would significantly affect Y.)
Describe variables specifically (e.g., "vocabulary test scores" rather than "learning")
Make RQs answerable with your data and analysis
Use parallel structure if you have multiple RQs
Avoid vague verbs (for example, "explore") unless your design is truly exploratory
Match RQ wording with your Results section headings where possible

Prompt: check RQ phrasing in English

Check the following research questions for appropriateness as academic English.

Criteria:
(1) Is the interrogative form grammatically correct (grammar, word order)?
(2) Are variables described specifically?
(3) Is the question empirically testable?
(4) Does it correspond to the gap presented in the Introduction?

If improvements are needed, provide one revised version.

Research questions:
[Paste your RQs here]

Gap presented in the Introduction (summary):
[Paste a summary of your gap]

Prompt: rewrite RQs for clarity and parallel structure

Rewrite the following research questions in clear academic English.

Requirements:
- Keep meaning unchanged
- Make wording parallel across RQs
- Use specific verbs (e.g., examine, test, compare) when appropriate
- Provide 2 versions: conservative (minimal change) and clearer (more revision)

RQs:
[Paste your RQs here]

Adjusting the length of an Introduction¶

As noted in 2.1, the Introduction typically makes up about 20–25% of a paper (roughly 1,200–1,500 words for a 6,000-word paper). AI-generated text may be too short or too long, so length adjustment is often necessary.

If it is too short¶

When an AI-generated Introduction is too short, the most common reason is that the overview of prior work is underdeveloped. Check for the following:

Does Move 1 (establishing a territory) include sufficient background?
Is prior work described synthetically (listing tends to be shorter)?
Is the gap explanation specific enough?

Prompt: identify where to expand the Introduction

The following Introduction is about 600 words, and the target is 1,200–1,500 words.
Identify areas that should be expanded while keeping the current content.

For each area:
- Which move does it belong to (Move 1/2/3)?
- What topic should be added (be specific)?
- Approximate additional word count

Do not add any English text. Provide instructions only.

[Paste Introduction here]

If it is too long¶

On the other hand, AI may generate verbose text or describe prior studies in too much detail.

Prompt: compress the Introduction

The following Introduction is about 2,000 words, and I want to compress it to 1,200–1,500 words.
Identify passages that can be deleted or compressed, in order of priority.

Criteria:
- Background that does not directly relate to the research questions is a compression candidate
- Passages that repeat the same claim in different wording are merge candidates
- Overly general opening sentences are deletion candidates

Provide deletion/compression instructions only. Do not rewrite.

[Paste Introduction here]

Evaluating and revising AI-generated drafts¶

AI can help you draft quickly, but you still need a robust evaluation step. Use the following criteria to review your AI-generated Introduction.

Criterion	What to check	Common problems
Content accuracy	Does the output match what you intended to convey?	AI adds "plausible-sounding claims" that you did not write
Logical flow	Is the hourglass structure maintained?	Move 2 (gap) appears abruptly with a weak connection to Move 1
Specificity	Is the text too abstract?	Sentences like "This has attracted growing attention" without concrete details
Clarity of the gap	Does the reader understand "why this study is needed"?	The logical connection between the gap and the study purpose is unclear
Citation validity	Are there any fabricated citations?	AI inserts nonexistent author names and years

When AI output is too abstract

AI often generates text that is "generally plausible but lacks specificity." Add concrete details about your data, participants, methods, and context to increase the uniqueness of your Introduction.

Prompt: increase the specificity of a generated Introduction

The following Introduction is too abstract and stays at the level of generalities.
Identify the places where information specific to my study should be added,
and indicate "what should be written here" concretely (e.g., participant characteristics,
data size, field-specific terminology).

Do not revise or add English text. Provide instructions only.

Study overview (in Japanese):
[Describe the specifics of your research in Japanese]

Introduction:
[Paste the AI-generated Introduction here]

Prompt: deviation check (Japanese vs English)

Compare the Japanese and English below.
Identify any deviations in meaning, claim strength, or missing information.

Output:
- List deviations only
- Do not rewrite

Japanese:
[Japanese skeleton]

English:
[English paragraph]

If you want a broader cross-section check (Introduction → Results → Discussion consistency), use the revision prompts in:
3.5 Revision Techniques