Intent Analyzer

An advanced system for deeply understanding user intent by going beyond surface-level requests to discover underlying goals, unstated constraints, and true needs.

Overview

Intent Analyzer represents a sophisticated approach to understanding what users really want. Rather than taking requests at face value, it employs cognitive science principles to examine underlying intent, identify implicit assumptions, recognize unstated constraints, and help users articulate their true goals clearly.

This skill draws inspiration from coherent extrapolated volition in AI alignment theory—determining what someone would want if they "knew more, thought faster, and were more the person they wished they were." Applied practically, this means understanding not just what the user explicitly requested, but what they would have requested with complete knowledge of possibilities, perfect clarity about their goals, and full awareness of relevant constraints.

When to Use This Skill

Apply Intent Analyzer when:

• User requests are ambiguous or could be interpreted multiple ways
• Deeper understanding of goals would significantly improve response quality
• The stated request might be a proxy for an unstated underlying need
• Critical information appears to be missing or assumed
• Multiple reasonable interpretations exist and choosing wrong would waste effort
• Helping users clarify complex or poorly-defined problems
• Teaching or mentoring where understanding motivation improves guidance

This skill is particularly valuable for complex, open-ended, or high-stakes requests where misunderstanding intent could lead to significant wasted effort or poor outcomes.

Core Principles

Intent Analyzer operates on five fundamental principles:

First Principles Decomposition

Break down every request to its most fundamental goals. Question surface-level assumptions about what is being asked. Often, the stated request is a proxy for a deeper underlying need.

For example:

• "Summarize this document" might actually mean: seeking specific information within it, preparing for a meeting, evaluating whether to read it fully, or extracting key decisions
• "Help me write code" might actually mean: learning programming concepts, completing a specific project, debugging existing code, or understanding best practices

Identify these underlying intentions by decomposing the request to its fundamental purpose.

Probabilistic Intent Mapping

Every user message carries multiple possible interpretations with varying probabilities. Construct a probability distribution over potential intents considering:

• Context clues in the phrasing
• Domain patterns and common use cases
• Explicit and implicit information provided
• What's left unsaid or assumed

When multiple high-probability interpretations exist, explicitly acknowledge uncertainty and seek clarification rather than guessing. When one interpretation is clearly dominant (>80% confidence), proceed while remaining open to correction.

Evidence-Based Pattern Recognition

Recognize which category of request this represents based on established taxonomies:

• Creative task (writing, design, ideation)
• Analytical task (evaluation, comparison, assessment)
• Technical task (coding, configuration, troubleshooting)
• Learning query (explanation, teaching, understanding)
• Decision-making request (choosing between options, planning)
• Problem-solving (debugging, optimization, fixing issues)

Each category has characteristic patterns, common unstated assumptions, and typical underlying goals. Use pattern recognition to inform interpretation.

Constraint Detection

Identify both explicit and implicit constraints:

Explicit constraints: Directly stated requirements like word limits, specific formats, deadline pressures, technical requirements, or resource limitations

Implicit constraints: Emerge from context such as:

• Technical skill level (suggested by terminology used)
• Time pressure (urgent phrasing, "quick" requests)
• Resource limitations (mentions of budget, access, permissions)
• Domain-specific requirements (industry standards, company policies)
• Audience considerations (who will see/use the output)

Surface implicit constraints through strategic questioning when they significantly impact the response.

Socratic Clarification

When facing genuine uncertainty, engage in targeted questioning designed to reveal:

• Hidden assumptions and unstated premises
• True underlying goals beyond stated requests
• Critical constraints that weren't mentioned
• Context that would change the optimal approach

These questions are strategically chosen to disambiguate between competing interpretations, not to gather exhaustive details. Quality questions prevent wasted effort on wrong interpretations.

The Intent Analysis Process

Phase 1: Deep Analysis (Internal Processing)

Upon receiving a request, immediately engage in comprehensive internal analysis:

Intent Archaeology: Excavate the layers of intent:

• What is explicitly stated?
• What is implied?
• What domain knowledge or context is assumed?
• What expertise level does the phrasing suggest?
• Are there temporal constraints, quality requirements, or formatting preferences implied?

Goal Extrapolation: Construct a model of what the user is ultimately trying to achieve:

• Immediate goals (the task at hand)
• Higher-order goals (why they're doing this task)

For example, someone asking for code to scrape a website might be:

• Building a data pipeline (needs production-ready, maintainable code)
• Learning web scraping (needs educational code with explanations)
• Completing a school project (needs working code with documentation)
• Solving a one-time problem (needs quick, simple solution)

Each underlying goal suggests different optimal responses.

Constraint Detection: Identify constraints both explicit and implicit:

• Stated requirements (lengths, formats, deadlines)
• Contextual constraints (skill level, time pressure, resources)
• Domain requirements (standards, policies, compatibility)

Pattern Recognition: Map the request to established categories and identify which prompting patterns would be most beneficial. Is this analytical, creative, technical, learning-focused, or decision-oriented? Each benefits from different approaches.

Ambiguity Assessment: Quantify uncertainty in interpretation:

• High confidence (>80%): Proceed with dominant interpretation while noting alternatives
• Moderate confidence (50-80%): Proceed with interpretation but explicitly acknowledge assumption
• Low confidence (<50%): Seek clarification before proceeding

Phase 2: Decision Point

After internal analysis, choose between two paths:

Path A - High Confidence Interpretation: When analysis reveals clear dominant interpretation (confidence >80%), proceed directly while:

• Briefly noting interpretation if it involves meaningful assumptions
• Remaining open to correction if interpretation was wrong
• Framing response to make interpretation obvious

Path B - Clarification Required: When analysis identifies:

• Genuine ambiguity (multiple interpretations >30% probability each)
• Hidden assumptions that could lead to dramatically different responses
• Critical missing information that significantly impacts optimal approach

Engage in Socratic clarification before proceeding.

Phase 3: Socratic Clarification (When Needed)

When clarification is needed, ask strategic questions:

Disambiguation Questions: Distinguish between competing interpretations:

• "Are you looking to [interpretation A] or [interpretation B]?"
• "Is your goal [immediate goal] or [higher-order goal]?"
• "Would you prefer [approach X] or [approach Y]?"

Constraint Revelation Questions: Surface unstated constraints:

• "What will you use this for?" (reveals purpose)
• "Who is the audience?" (reveals formality/complexity needs)
• "What have you tried already?"