Scientific Critical Thinking

Scientific Critical Thinking

Overview

Critical thinking is a systematic process for evaluating scientific rigor. Assess methodology, experimental design, statistical validity, biases, confounding, and evidence quality using GRADE and Cochrane ROB frameworks. Apply this skill for critical analysis of scientific claims.

When to Use This Skill

This skill should be used when:

• Evaluating research methodology and experimental design
• Assessing statistical validity and evidence quality
• Identifying biases and confounding in studies
• Reviewing scientific claims and conclusions
• Conducting systematic reviews or meta-analyses
• Applying GRADE or Cochrane risk of bias assessments
• Providing critical analysis of research papers

Appraisal Workflow

1. Fix the question and unit. Record population, intervention/exposure, comparator, outcome, time point, effect measure, and target setting. Distinguish descriptive, predictive, and causal claims. Identify the independent experimental/sampling unit.
2. Extract evidence before judging. Locate the numerical result, uncertainty, denominators, protocol/registration, analysis plan, and relevant supplementary material. Keep missing reporting separate from evidence that a procedure was not performed.
3. Assess design and analysis. Check selection, confounding, measurement, attrition, multiplicity, dependence, and model assumptions. Good fit or optimizer convergence does not establish a uniquely identified parameter or a causal effect.
4. Choose the right appraisal framework. Reporting completeness, risk of bias, and certainty of a body of evidence answer different questions. Record the exact tool version and the result being assessed; do not turn checklist counts into a quality score.
5. Synthesize with scope intact. Examine independent replication, overlapping samples, missing evidence, and applicability. Apply GRADE per outcome/comparison when appropriate, with explicit domain reasons, rather than grading a whole paper by its design label.
6. Write a traceable critique. For each concern give the source location, observation, consequence for the claim, uncertainty, and a feasible remedy. Separate supported conclusions from assumptions and from clinical/policy recommendations.

Current framework versions, primary sources, and verification limits are in references/review_sources.md. The examples in the references are teaching examples, not empirical findings or validated patient-specific advice.

Visual Aids (Optional)

Only add figures when the user explicitly requests a diagram (for example, a GRADE flowchart, bias decision tree, or evidence-quality framework).

When figures help:

• Critical thinking framework diagrams
• Bias identification decision trees
• Evidence quality assessment flowcharts
• GRADE or risk-of-bias evaluation frameworks

How to create figures:

• Preferred: Use the scientific-schematics skill for AI-generated diagrams from a natural-language description
• Alternative: Build figures in your usual tools (draw.io, PowerPoint, matplotlib, etc.)

Run from the repository root, with OPENROUTER_API_KEY set:

python skills/scientific-schematics/scripts/generate_schematic.py "Illustrative GRADE appraisal: define outcome and comparison, assess certainty domains with reasons; keep recommendation decisions separate" -o figures/grade_flowchart.png --doc-type report

This optional command's CLI was checked with --help; paid generation was not exercised for this example. Follow that skill's current dependencies and review every generated label.

Disclosure: AI schematic generation sends your prompt to OpenRouter (a third-party API). Do not include unpublished sensitive details unless that transmission is appropriate for your project.

Core Capabilities

Seven capability areas, each with the questions to ask and what the answers imply, are in references/core_capabilities.md:

1. Methodology critique — design, controls, confounding, and whether the method can answer the question asked.
2. Bias detection — selection, measurement, publication, and cognitive biases.
3. Statistical analysis evaluation — power, multiplicity, p-value misuse, effect sizes.
4. Evidence quality assessment — study hierarchy, replication, and strength of inference.
5. Logical fallacy identification — the fallacies that recur in scientific argument.
6. Research design guidance — how to strengthen a design before data collection.
7. Claim evaluation — separating what was shown from what is being asserted.

Per-topic detail is in references/scientific_method.md, references/common_biases.md, references/statistical_pitfalls.md, references/evidence_hierarchy.md, references/logical_fallacies.md, and references/experimental_design.md.

Application Guidelines

General Approach

1. Be Constructive

   • Identify strengths as well as weaknesses
   • Suggest improvements rather than just criticizing
   • Distinguish between fatal flaws and minor limitations
   • Recognize that all research has limitations

2. Be Specific

   • Point to specific instances (e.g., "Table 2 shows..." or "In the Methods section...")
   • Quote problematic statements
   • Provide concrete examples of issues
   • Reference specific principles or standards violated

3. Be Proportionate

   • Match criticism severity to issue importance
   • Distinguish between major threats to validity and minor concerns
   • Consider whether issues affect primary conclusions
   • Acknowledge uncertainty in your own assessments

4. Apply Consistent Standards

   • Use same criteria across all studies
   • Don't apply stricter standards to findings you dislike
   • Acknowledge your own potential biases
   • Base judgments on methodology, not results

5. Consider Context

   • Acknowledge practical and ethical constraints
   • Consider field-specific norms for effect sizes and methods
   • Recognize exploratory vs. confirmatory contexts
   • Account for resource limitations in evaluating studies

Apply risk-of-bias tools to the right unit

For RoB 2, identify the specific result: outcome, time point, intervention comparison, numerical estimate, and effect of assignment versus adherence. Use the variant for individually randomized, cluster, or crossover trials; record signalling answers and justifications rather than assigning one blanket score to the whole paper. Different outcomes in the same trial can have different bias judgments. See the Cochrane RoB 2 guidance.

For non-randomized intervention effects, state whether using ROBINS-I 2016 or the ROBINS-I V2 November 2025 draft for follow-up/cohort studies. Do not mix their domains or algorithms. Diagnostic accuracy appraisal now uses QUADAS-3 (current tool v1.2), at the accuracy-estimate level. See the tool-specific sources before a formal assessment.