Research & Analysis
🔬 Structured thinking for academic research, due diligence, and investigative work.
Research benefits from ReasonKit’s multi-tool approach to ensure comprehensive, unbiased analysis.
Common Research Questions
Literature Review
rk think "What are the key debates in [your field]? What's established vs. contested?" --deep
Methodology Check
rk think "I'm planning to use [methodology] for my study on [topic]. What are the potential weaknesses?" --balanced
Claim Verification
rk think "Paper claims [specific claim]. How robust is this conclusion?" --paranoid
Research Direction
rk think "I want to study [topic]. What angles are underexplored? What might be impactful?" --deep
Example Analysis
Question: “I’m reviewing a paper that claims meditation reduces cortisol by 25%. Should I cite it?”
rk think "Paper claims meditation reduces cortisol by 25% (n=45, randomized). Should I cite this finding?" --paranoid
╔════════════════════════════════════════════════════════════╗
║ RESEARCH ANALYSIS: Citation Decision ║
╚════════════════════════════════════════════════════════════╝
┌────────────────────────────────────────────────────────────┐
│ 💡 GIGATHINK: Research Perspectives │
├────────────────────────────────────────────────────────────┤
│ 1. METHODOLOGY: Is RCT the right design for this? │
│ 2. SAMPLE SIZE: n=45 is small for physiological claims │
│ 3. EFFECT SIZE: 25% is large - extraordinary claims... │
│ 4. REPLICATION: Has this been replicated? │
│ 5. PUBLICATION BIAS: What about null results? │
│ 6. MEASUREMENT: How was cortisol measured? When? │
│ 7. BLINDING: Can you blind meditation studies? │
│ 8. CONTROL GROUP: What did control do? Placebo effect? │
│ 9. AUTHOR CONFLICTS: Financial interests in meditation? │
│ 10. META-ANALYSES: What do aggregated studies show? │
└────────────────────────────────────────────────────────────┘
┌────────────────────────────────────────────────────────────┐
│ ⚡ LASERLOGIC: Methodological Critique │
├────────────────────────────────────────────────────────────┤
│ ISSUE 1: Small sample size │
│ → n=45 has low statistical power │
│ → High risk of false positive (especially if p≈0.05) │
│ → Effect likely inflated (winner's curse) │
│ │
│ ISSUE 2: Large effect size suspicious │
│ → 25% cortisol reduction is unusually large │
│ → Larger effect sizes often shrink on replication │
│ → Compare to meta-analytic estimates │
│ │
│ ISSUE 3: Blinding challenges │
│ → Participants know if they're meditating │
│ → Expectation effects can be substantial │
│ → Hard to have true placebo in meditation studies │
└────────────────────────────────────────────────────────────┘
┌────────────────────────────────────────────────────────────┐
│ 🛡️ PROOFGUARD: Source Verification │
├────────────────────────────────────────────────────────────┤
│ TO VERIFY: │
│ • Journal impact factor and reputation │
│ • Author h-index and prior work │
│ • Funding sources │
│ • Replication attempts │
│ • Meta-analysis inclusion │
│ │
│ TRIANGULATION NEEDED: │
│ • Find 2+ independent replications │
│ • Check meta-analyses for pooled effect size │
│ • Look for pre-registered studies │
│ │
│ VERIFICATION QUESTIONS: │
│ • Is the data publicly available? │
│ • Was the study pre-registered? │
│ • What do systematic reviews conclude? │
└────────────────────────────────────────────────────────────┘
┌────────────────────────────────────────────────────────────┐
│ 🔥 BRUTALHONESTY: Academic Reality │
├────────────────────────────────────────────────────────────┤
│ UNCOMFORTABLE TRUTH: │
│ You want to cite this because it supports your thesis. │
│ Confirmation bias makes weak evidence look strong. │
│ │
│ UNCOMFORTABLE TRUTH: │
│ Single studies, especially with small n and large │
│ effects, rarely replicate. The replication crisis │
│ exists precisely because of papers like this. │
│ │
│ HONEST QUESTION: │
│ If this study showed 0% effect, would you still cite it? │
│ If no, you're cherry-picking. │
└────────────────────────────────────────────────────────────┘
SYNTHESIS:
Don't cite as standalone evidence. Instead:
1. Find and cite meta-analyses (stronger evidence)
2. If citing, note limitations (small n, needs replication)
3. Use language like "some evidence suggests" not "studies show"
4. Look for pre-registered replications
Research-Specific Profile
[profiles.research]
tools = ["gigathink", "laserlogic", "proofguard", "brutalhonesty"]
gigathink_perspectives = 15
laserlogic_depth = "exhaustive"
proofguard_sources = 5
proofguard_require_citation = true
timeout = 300
Research Quality Checklist
ReasonKit helps verify:
| Criterion | Question |
|---|---|
| Sample size | Is n sufficient for claimed effect? |
| Effect size | Is it realistic or suspiciously large? |
| Replication | Has it been independently replicated? |
| Pre-registration | Was hypothesis registered before data? |
| Conflicts | Are there financial/ideological conflicts? |
| Publication bias | Are null results published? |
| Methodology | Is design appropriate for question? |
Common Research Biases
| Bias | How ReasonKit Helps |
|---|---|
| Confirmation bias | BrutalHonesty challenges your preferences |
| Publication bias | ProofGuard asks about null results |
| Authority bias | LaserLogic evaluates arguments, not authors |
| Recency bias | GigaThink includes historical perspectives |
Academic Use Cases
Thesis Direction
rk think "My thesis proposal is [X]. Advisor likes it. What's wrong with it?" --deep
Peer Review Preparation
rk think "I'm submitting to [journal]. What will reviewers criticize?" --paranoid
Grant Writing
rk think "My grant proposal claims [X]. How would a skeptical reviewer attack this?" --deep
Debate Preparation
rk think "I'm presenting position [X]. What's the strongest counterargument?" --balanced
Tips for Research Analysis
-
Include methodology details — Design, sample size, statistical approach
-
Specify the claim precisely — Vague claims get vague analysis
-
Ask for counterarguments — “What’s wrong with this?” is valuable
-
Use paranoid for citations — Avoid citing weak evidence
-
Run before and after — Check assumptions before research, verify conclusions after