Comparison - Impression-Informed Option Evaluation

Input: $ARGUMENTS

Core Principles

Impression before scoring. State your overall sense of which option is best BEFORE structured evaluation. This captures qualitative judgment that scoring can miss.
Tiers, not scores. Rate options as clearly better / slightly better / equivalent / slightly worse / clearly worse — not 2.3 vs 2.1. False precision obscures real differences.
Criteria serve the purpose. Don’t evaluate against generic criteria. Ask: “What does this need to accomplish?” The answer defines what matters.
Divergence is data. When your impression says “B is better” but your scoring says “A wins,” investigate. The divergence often reveals a missing criterion or a misjudged weight.
Kill clearly bad options early. Don’t waste effort scoring options that fail hard requirements. Eliminate first, then compare survivors.

The Process

1. State the Impression

Before any structured evaluation:

OPTIONS: [List all options]
IMPRESSION: [Which seems best, at first glance? Why?]
CONFIDENCE: [HIGH / MEDIUM / LOW]

2. Define Purpose and Criteria

What must this accomplish? Derive criteria from purpose.

REQUIRED — Must have. Failure = eliminated.

- [Criterion]: [What "pass" looks like]

IMPORTANT — Strongly preferred. Failure = significant penalty.

- [Criterion]: [What good looks like]

NICE-TO-HAVE — Bonus value. Absence acceptable.

- [Criterion]: [What it adds]

3. Eliminate on Required Criteria

For each option, check required criteria. PASS or FAIL.

| Option | [Req 1] | [Req 2] | [Req 3] | Status |
|--------|---------|---------|---------|--------|
| A      | PASS    | PASS    | PASS    | Survives |
| B      | PASS    | FAIL    | PASS    | Eliminated |
| C      | PASS    | PASS    | PASS    | Survives |

Eliminated options stop here. Record why. Note: under what conditions would they pass? (prevents premature rejection)

4. Compare Survivors on Important Criteria

For each important criterion, compare survivors in TIERS:

[Criterion name]:
- A: [Evidence/reasoning] → TIER
- C: [Evidence/reasoning] → TIER

Tiers: CLEARLY BETTER / SLIGHTLY BETTER / EQUIVALENT / SLIGHTLY WORSE / CLEARLY WORSE

5. Overall Assessment

CRITERION SUMMARY:
| Criterion | A | C | Edge |
|-----------|---|---|------|
| [Crit 1]  | Slightly better | — | A |
| [Crit 2]  | — | Clearly better | C |
| [Crit 3]  | Equivalent | Equivalent | — |

OVERALL: [Which option wins on the criteria that matter most?]

6. Divergence Check

IMPRESSION said: [option X seems best]
ANALYSIS says: [option Y scores best]

DIVERGENCE? [YES/NO]
If YES: What criterion is the impression weighting that the analysis isn't?
        → [Investigate. Often reveals a missing important criterion.]

7. Recommendation

RECOMMENDATION: [Option]
CONFIDENCE: [HIGH / MEDIUM / LOW]
REASONING: [2-3 sentences explaining why, incorporating both analysis and impression]
RISKS: [What could go wrong with this choice?]
REVERSIBILITY: [How easy to change course if wrong?]

Handling Difficult Cases

Situation	Approach
Two options very close	Focus on REVERSIBILITY — pick the easier one to undo
Many options (5+)	Eliminate on required criteria first, then compare top 3
No clear winner	Ask: “What additional information would make this clear?”
Impression and analysis diverge strongly	The divergence IS the finding — investigate what’s behind it
All options are bad	Say so. Suggest reframing the choice.

When Called by Other Skills

Comparison is a primitive. When called by decision_procedure, UAUA, or other skills:

Accept options and context from the caller
Return: impression, elimination results, tier comparison, divergence check, recommendation
Be direct about which option wins and why

Pre-Completion Check

Impression stated before analysis
Criteria derived from purpose (not generic)
Required criteria used for elimination
Survivors compared in tiers (not numeric scores)
Divergence between impression and analysis checked
Recommendation includes reasoning and risk