Interpretation Space Search

Overview

Ambiguity means multiple interpretations are possible. Instead of guessing, systematically:

1. Generate all plausible interpretations
2. Score each against criteria
3. Select the best-supported interpretation

Goal

When input is ambiguous, generate all possible interpretations and search for the best one using explicit criteria.

Steps

Step 1: Capture the Input

Record exactly what was said/observed. Don’t interpret yet - just capture.

Output: Raw input

Step 2: Note the Context

Record all available context:

• Who/what is the source?
• What preceded this?
• What’s the broader situation?
• What do we know about the source?

Output: Context summary

Step 3: Generate Interpretations

Using generation methods, produce all plausible interpretations. Be comprehensive - include interpretations you doubt.

For each interpretation, state:

• The interpretation in clear terms
• What assumptions it requires

Output: List of interpretations

Step 4: Apply Obvious Filters

Remove interpretations that:

• Contradict known facts
• Are logically impossible
• Require absurd assumptions

Don’t remove just because unlikely - keep for scoring.

Output: Filtered interpretations

Step 5: Score Each Interpretation

For each interpretation, score on each criterion (1-10). Be consistent across interpretations.

Output: Scored interpretations

Step 6: Rank Interpretations

Sort by weighted total score. Note the gap between #1 and #2.

Output: Ranked interpretations

Step 7: Assess Confidence

Confidence based on:

• Score gap (large gap = high confidence)
• Absolute score (high score = good interpretation)
• Number of viable alternatives (fewer = higher confidence)

If confidence is low, consider:

• Seeking more information
• Testing interpretations
• Accepting ambiguity

Output: Confidence assessment

Step 8: Select and Verify

Select best interpretation. Verify it accounts for all key aspects of input. Note what would change the interpretation.

Output: Selected interpretation with caveats

When to Use

• Ambiguous communication
• Unclear requirements
• Multiple possible meanings
• Conflicting signals
• Need to understand intent

Verification

• Multiple interpretations were generated (not just the obvious one)
• Interpretations cover literal, intent, and context variations
• Each interpretation is internally coherent
• Scoring was consistent across interpretations
• Confidence assessment accounts for score gaps
• Selected interpretation accounts for key input features