/deep-research

Mode Selection Guide

Overview

deep-research provides 8 modes suited to different research stages and needs. This guide helps users select the most appropriate mode.

Decision Flowchart

User Input
    │
    ├── Have a clear research question?
    │   ├── Yes ──→ Have a text to review?
    │   │            ├── Yes ──→ review mode
    │   │            └── No ───→ Need PRISMA-compliant systematic review / meta-analysis?
    │   │                         ├── Yes ──→ systematic-review mode
    │   │                         └── No ───→ Need a complete report?
    │   │                                     ├── Yes ──→ full mode
    │   │                                     └── No ───→ Only need literature?
    │   │                                                 ├── Yes ──→ Need rapid paper comparison?
    │   │                                                 │            ├── Yes ──→ three-way-scan mode
    │   │                                                 │            └── No ───→ lit-review mode
    │   │                                                 └── No ───→ quick mode
    │   │
    │   └── No ──→ Want guided thinking?
    │              ├── Yes ──→ socratic mode
    │              └── No ───→ full mode
    │                          (Phase 1 interactive RQ clarification)
    │
    ├── Only need to verify specific facts?
    │   └── Yes ──→ fact-check mode
    │
    └── Not sure what you need?
        └── Describe your situation → System auto-recommends a mode

Detailed Mode Information

full mode (Complete Research)

Trigger Examples:

"Research the impact of AI on higher education quality assurance"
"Deep research on the impact of declining birth rates on Taiwan's higher education"
"Research the current state of SDGs implementation in Asian universities"

quick mode (Quick Research)

Trigger Examples:

"Quick research on blockchain in education"
"Quick research on the latest trends in educational technology"

review mode (Text Review)

Trigger Examples:

"Review this paper"
"Help me review this paper's methodology"
"Check this manuscript before submission"

lit-review mode (Literature Review)

Trigger Examples:

"Literature review on SDGs in higher education"
"Literature review: the evolution of quality assurance in Taiwan's higher education"

three-way-scan mode (WHY / HOW / WHAT paper comparison)

Trigger Examples:

"Compare these papers in WHY/HOW/WHAT format"
"Quick 3W scan of recent LLM-evaluation papers"
"Which of these should I read first?"

fact-check mode (Fact-Checking)

Trigger Examples:

"Fact-check these claims about Taiwan's university enrollment"
"Fact-check: Is the number of universities in Taiwan really declining?"
"Verify: 'OECD countries average 50% tertiary attainment rate'"

socratic mode (Guided Research)

Trigger Examples:

"Guide my research on higher education topics"
"Guide my research on educational technology"
"Help me think through my thesis direction"
"Help me think through my research topic"
「引導我的研究：高教品保」
「幫我釐清我的研究方向」
「幫我想想，我對少子化議題有興趣但不確定要研究什麼」
「我有個模糊的想法，想找研究題目」
「帶我做研究」

systematic-review mode (Systematic Review / Meta-Analysis)

Trigger Examples:

"Systematic review of AI-assisted assessment in higher education"
"Meta-analysis of the effect of active learning on STEM outcomes"
"PRISMA review of quality assurance frameworks in Asian universities"
"Evidence synthesis on the impact of accreditation on institutional improvement"

Common Misselection Scenarios

Mode Transitions

Common Transition Paths

socratic → full              Continue with complete research after Socratic completion
socratic → academic-paper    Write paper directly after Socratic completion
lit-review → full            Want complete analysis after literature review
lit-review → systematic-review  Author chooses a systematic review after initial lit survey
fact-check → full            Need deeper research after fact-checking
quick → full                 Worth going deeper after quick research
review → full                Need to re-research after review
systematic-review → academic-paper  Write up systematic review as a paper

deep-research to academic-paper Mode Mapping

deep-research vs academic-paper-reviewer Mode Mapping

Complete Academic Research Pipeline

Step 1: deep-research (socratic/full)
          ↓ Research Plan / Full Report
Step 2: academic-paper (plan/full)
          ↓ Paper draft
Step 3: academic-paper-reviewer (full/guided)
          ↓ Review comments
Step 4: academic-paper (revision)
          ↓ Revised paper
Step 5: [Repeat Steps 3-4 until passed]
          ↓ Final paper

Mode Transition Matrix

Rules for switching between modes mid-research. Not all transitions are safe.

Transition: quick → full

• When: Quick brief reveals the topic is more complex than expected
• Reusable: RQ Brief (as-is), initial keyword list
• Must Redo: Full literature search (quick only uses 5-8 sources), synthesis, verification
• Quality Delta: Full mode requires 15+ sources, 3+ databases, formal methodology design

Transition: lit-review → full

• When: Literature review reveals a gap worth investigating with original methodology
• Reusable: Complete bibliography, synthesis themes, evidence gap analysis
• Must Redo: Research design (methodology_patterns), data collection plan, ethics review (if primary research)
• Quality Delta: Full mode adds original research design; lit-review is secondary analysis only

Transition: socratic → full

• When: Socratic dialogue produces a well-formed RQ and user wants autonomous research
• Reusable: RQ Brief (with socratic_insights), accumulated INSIGHTs, scope definition
• Must Redo: Everything after RQ formulation (bibliography, synthesis, verification, report)
• Quality Delta: socratic mode only produces RQ Brief; full mode executes the complete pipeline

Transition: fact-check → full

• When: Fact-checking reveals a claim is part of a larger contested topic worth researching
• Reusable: Verified/debunked claims, source verification results
• Must Redo: RQ formulation (reframe from verification to inquiry), full bibliography, synthesis
• Quality Delta: Fact-check is binary (true/false/mixed); full mode produces nuanced analysis

Transition: lit-review → systematic-review

• When: The author chooses a systematic review, for example after the review-form note (deep-research/SKILL.md § Review-form note, #921) or because their target journal requires PRISMA compliance. ARS does not start this transition because the topic seems to warrant it
• Reusable: Initial keyword strategy, some identified sources (need re-screening)
• Must Redo: Protocol registration, formal inclusion/exclusion criteria, dual screening, risk of bias assessment, meta-analysis feasibility assessment
• Quality Delta: systematic-review requires protocol, RoB assessment, GRADE; lit-review has none of these

Transition: systematic-review → academic-paper

• When: Systematic review is complete and user wants to write it up as a journal article
• Reusable: Everything — PRISMA report is essentially the paper draft
• Must Redo: Formatting to target journal requirements, abstract restructuring
• Quality Delta: Minimal — systematic review output is already structured per PRISMA 2020

Prohibited Transitions

• full → quick: Cannot downgrade a full research to quick brief (loss of rigor)
• Any → socratic: Socratic mode is an entry point only; cannot transition into it mid-pipeline
• paper-review → full: Paper review evaluates existing work; full mode creates new research. These are fundamentally different tasks