AI Prompt for Ph.D.

這年頭不需要讀博班，但需要學習批判性思考+篩選參考資料+ 對AI下指令(Context and Prompt Engineering)

1.The Intake Protocol

請依據以下英文提示詞，使用繁體中文說明回覆

I’m going to share [NUMBER] papers on [TOPIC]. Before I ask any questions, please do the following:

1. List every paper in a table with columns: Author(s) | Year | Core Claim (one sentence, ≤20 words). If a paper has no explicit thesis, infer the central argument from its conclusions.

2. Group the papers into 2–5 clusters based on shared theoretical assumptions or frameworks. Name each cluster and briefly explain (1–2 sentences) what unites the papers within it.

3. Flag any direct contradictions between papers — where two or more authors make mutually exclusive claims about the same phenomenon. List each as: Paper A vs. Paper B — contested claim.

Do not summarize each paper individually. Focus only on the three tasks above.

2. The Contradiction Finder（矛盾發現器）

請依據以下英文提示詞，使用繁體中文說明回覆

Across all uploaded papers, identify the most significant points where two or more authors make claims that directly contradict each other.

• Only include genuine contradictions — mutually exclusive claims on the same issue.

• Exclude cases of mere difference in emphasis or scope.

Present your findings as a table with the following columns:

| Contested Claim | Position A (Paper, Year) | Position B (Paper, Year) | Root Cause of Disagreement |

For Root Cause, choose from: methodology, dataset, time period, definition of terms, or other (explain). Aim for 5–10 contradictions. If fewer exist, list all you find.

3. The Citation Chain（引用鏈）

請依據以下英文提示詞，使用繁體中文說明回覆

From the uploaded papers, identify the 3 concepts that appear most frequently across multiple papers (referenced by name, debated, or built upon).

For each concept, trace its intellectual history using only the evidence in the uploaded papers:

Concept Name:

• Origin: Who first introduced or defined it (within this set)?

• Challenge: Which paper(s) questioned or challenged it, and how?

• Refinement: Which paper(s) modified or extended it, and how?

• Current Status: Settled, contested, or still evolving — based on this literature?

Present each concept as a structured outline. If a concept lacks a clear challenge or refinement in these papers, state that explicitly rather than guessing.

4. The Gap Scanner（研究缺口掃描器）

請依據以下英文提示詞，使用繁體中文說明回覆

Based only on the uploaded papers, identify the 5 most significant research gaps that these papers collectively acknowledge, imply, or fail to address.

For each gap:

• Gap: [State the unanswered question clearly in 1–2 sentences]

• Why it exists: Choose from — methodological barrier, lack of data, topic too niche, assumed but untested, or ethical/logistical constraint. Explain briefly.

• Closest paper: Which uploaded paper came closest to addressing it, and where did it fall short?

• Path to resolution: What would be needed to close this gap (methodology, data, resources, etc.)?

Rank the 5 gaps from most to least significant, and briefly explain your ranking criterion (e.g., theoretical importance, practical impact, feasibility of resolution).

If fewer than 5 genuine gaps exist, list all you can identify and explain why the set is limited.

5. The Methodology Audit（方法論審計）

請依據以下英文提示詞，使用繁體中文說明回覆

Compare the research methodologies used across all uploaded papers.

Step 1 — Classification Table

Create a table: Paper (Author, Year) | Methodology Type | Data Source | Sample Size (if stated) | Key Limitation Noted by Authors.

Use the methodology type that best fits each paper. Don’t force papers into the categories below — add new categories as needed.

Suggested types: Survey, Experiment (RCT), Quasi-experiment, Simulation, Meta-analysis, Case study, Computational/ML, Literature review, Ethnography, Secondary data analysis.

Step 2 — Synthesis

• Which methodology type appears most frequently? Suggest why based on the papers’ stated rationale.

• Which methodology is absent or rare despite being relevant to the research questions?

Step 3 — Weakest Methodology

Identify the paper whose methodology is most vulnerable to criticism. Evaluate using these criteria: sample size adequacy, control for confounds, replicability, and transparency of reporting. State which criterion it fails most clearly.

6. The Master Synthesis（總體綜合）

請依據以下英文提示詞，使用繁體中文說明回覆

Using the uploaded papers as your only source, write a synthesis of this body of literature. Do NOT summarize individual papers. Instead, write across the entire literature:

1. Established consensus (~100 words): What does this field collectively agree on? Cite at least 2 papers that support each claim you make here.

2. Active debates (~100 words): What do researchers in this field meaningfully disagree about? Name the disagreeing positions without naming individual papers.

3. Strongest evidence (~100 words): What claims in this literature are supported by the most consistent, replicated, or methodologically robust evidence?

4. The key open question (~80 words): End with the single most important unanswered question in this field — the one whose resolution would most change the others.

Total: 400 words maximum. No hedging phrases like “it seems” or “some argue.” State clearly.

If the papers lack sufficient consensus to populate a section, say so explicitly.

7. The Assumption Killer（假設終結者）

請依據以下英文提示詞，使用繁體中文說明回覆

From the uploaded papers, identify the 5–8 most consequential assumptions that the majority of these papers share but never explicitly test, justify, or acknowledge as assumptions.

Focus on assumptions that are:

(a) foundational to the conclusions drawn, and

(b) plausibly false or context-dependent.

For each assumption:

• Assumption: [State it as a declarative claim, e.g., “X causes Y under all conditions”] Shared by: Name 2–3 papers that rely on it most heavily.

• Risk level: Rate as Low / Medium / High based on how much of the literature would be undermined if the assumption is false.

• Consequence: Explain what would change — would conclusions need revision (low impact), key findings be invalidated (medium), or the entire research paradigm collapse (high)?

Rank assumptions from most to least consequential.

8. The Knowledge Map Builder（知識圖譜建構器）

請依據以下英文提示詞，使用繁體中文說明回覆

Based only on the uploaded papers, create a structured knowledge map of this literature. Present it as a clean outline (no prose paragraphs).

KNOWLEDGE MAP

1. Central Claim: The single proposition that most of this field’s work tries to support, challenge, or refine. If no single claim unifies the field, name 2 competing centres instead.

2. Supporting Pillars (3–5): Well-established sub-claims with strong evidentiary support across multiple papers. For each: [Claim] — supported by: [Paper 1], [Paper 2]

3. Contested Zones (2–3): Areas of genuine, active disagreement. For each: [Issue] — [Position A] vs. [Position B]

4. Frontier Questions (1–2): Questions this literature raises but cannot yet answer. State as explicit questions.

5. Newcomer Reading List (3 papers): For each paper, state [Author, Year] — why a newcomer should read this first.

Selection criterion: foundational to understanding the field, not just most cited.

9. The ‘So What’ Test（「那又怎樣」測試）

請依據以下英文提示詞，使用繁體中文說明回覆

Summarize this entire body of research for a smart non-expert who has never read any of it. Respond in exactly three numbered points. Each point should be 2–3 sentences maximum.

Write as if speaking to an intelligent person with no domain knowledge.

1. What has been proven: The strongest, most reliable finding from this literature — stated as a direct claim with no hedging. No “suggests” or “may indicate.”

2. What is still unknown: The most significant thing this field has not yet figured out — stated honestly, without minimizing the uncertainty.

3. Why it matters: The single most important real-world implication. If no direct application exists, state the biggest theoretical consequence instead.

Rules: No jargon. No citations. No qualifications that weaken the core point.

If you cannot make a statement confidently based on the papers, say so — don’t fabricate certainty.