Evaluation Dimensions

Input: $ARGUMENTS

Overview

Universal dimensions for evaluating any claim, problem, or solution. Use external grounding to assess claims on these dimensions when ARAW alone cannot determine them.

External grounding helps:

1. Find problems/solutions ARAW wouldn’t find (blind spots)
2. Verify ARAW outputs on these dimensions
3. Calibrate confidence in ARAW findings

Steps

Step 1: Identify What’s Being Evaluated

1. Is it a CLAIM (something asserted to be true)?
2. Is it a PROBLEM (something needing solution)?
3. Is it a SOLUTION (something proposed to address a problem)?
4. Is it a PLAN (something proposed to achieve a goal)?
5. Is it an OUTPUT (something produced)?

Step 2: Apply Universal Dimensions

Truth / Accuracy:

• Is it factually correct?
• Are the claims verifiable?
• What evidence supports it? What contradicts?
• What’s the confidence level?
• → Check with external sources, data, domain experts

Completeness:

• Does it address everything it should?
• What’s missing?
• Are edge cases covered?
• Are limitations acknowledged?
• → Compare against known frameworks, checklists, standards

Consistency:

• Does it contradict itself internally?
• Does it contradict established knowledge?
• Does it contradict other claims by the same source?
• → Cross-reference across the output and against known facts

Relevance:

• Does it address the actual question/problem?
• Is everything included actually necessary?
• Is the scope appropriate (not too broad, not too narrow)?
• → Compare the output against the original objective

Feasibility:

• Can it actually be done?
• With available resources?
• In the available time?
• Given real-world constraints?
• → Check against physical, financial, organizational, and technical reality

Robustness:

• Does it work under normal conditions only, or also under stress?
• What happens when assumptions are violated?
• How sensitive is it to small changes in inputs?
• → Stress-test with edge cases, adversarial inputs, changed assumptions

Originality / Value-Add:

• Does it say anything new or just restate known things?
• Does it provide insight beyond the obvious?
• Is the framing itself valuable even if content is known?
• → Compare against what’s already known/available

Actionability:

• Can someone act on this?
• Are next steps clear?
• Is it specific enough to guide action?
• → Ask: “What would I do differently after reading this?”

Fairness / Balance:

• Are all perspectives represented?
• Is evidence selectively presented?
• Are counterarguments acknowledged?
• → Look for what’s conspicuously absent

Clarity:

• Can the target audience understand it?
• Are terms defined?
• Is the structure logical?
• → Show to a representative reader and check comprehension

Step 3: Score Each Dimension

Not all dimensions are equally important for every evaluation. Weight by context:

• For claims: Truth, Consistency, and Fairness matter most
• For solutions: Feasibility, Robustness, and Actionability matter most
• For plans: Completeness, Feasibility, and Clarity matter most
• For outputs: Clarity, Relevance, and Accuracy matter most

Step 4: Identify Critical Gaps

1. Which dimensions score below 3?
2. For each low score: is this fixable or fundamental?
3. Which low scores are most important given the evaluation type?
4. What would it take to raise each critical dimension?

Step 5: Report

EVALUATION DIMENSIONS:
Subject: [what was evaluated]
Type: [claim/problem/solution/plan/output]

| Dimension | Score | Weight | Weighted | Key Finding |
|-----------|-------|--------|----------|-------------|
| [dim] | [1-5] | [H/M/L] | [score] | [finding] |

Overall: [weighted average or qualitative summary]
Critical gaps: [dimensions below 3 that matter]
Improvement path: [what would most improve the evaluation]

External grounding used: [what sources/methods supplemented ARAW]
ARAW blind spots found: [what external grounding caught that ARAW missed]

When to Use

• Evaluating any claim, solution, or output
• When ARAW has been applied but you want dimensional verification
• When you need to communicate evaluation results systematically
• → INVOKE: /cv (convergent validation) for multi-method verification
• → INVOKE: /evaluate (category skill) for routing to appropriate evaluation

Verification

• Evaluation type correctly identified
• All 10 dimensions considered
• Dimensions weighted by relevance to evaluation type
• External grounding used (not just internal reasoning)
• Critical gaps identified with improvement paths