Problem Targeting

RFC-0000: Problem Targeting

Purpose

Ensure the system is addressing the correct causal layer before state collection or authorization begins.

This RFC implements Doctrine I: Mondai Ishiki — the discipline of problem consciousness.

The Law (from Doctrine I)

Action taken at the wrong causal layer will fail, regardless of local correctness or execution quality.

Therefore: No intervention may be authorized until the generating causal layer of the problem has been identified.

Core Concepts

Causal Layers Complex systems are composed of multiple abstraction layers. Failures observed at a given layer (e.g., UI, Output) are often generated upstream (e.g., Measurement, Incentive, Policy).

Problem Localization The act of identifying which causal layer generates the observed failure. Observed behavior is not evidence of causal origin.

Causal Ascent The disciplined process of moving upstream through abstraction layers until the failure's generating cause is found. Causal ascent terminates only when removing or correcting the identified layer would prevent the failure from arising.

Kadai Barashi (Problem Dissolution) A problem is dissolved when the conditions that generate it are removed, such that:

• The failure mode no longer arises
• Continuous monitoring is no longer required
• The original question becomes irrelevant

Status Codes

Interface

interface CausalValidation {
  problem_statement: string;
  causal_layer: "symptom" | "mechanism" | "root_cause" | "policy" | "incentive";
  upstream_layers_considered: string[];
  causal_ascent_complete: boolean;
  dissolution_possible: boolean;
  validation_source:
    | "user_confirmed"
    | "domain_expert"
    | "oracle_verified"
    | "unvalidated";
}

interface ProblemTargetingResult {
  status:
    | "VALIDATED"
    | "REQUIRES_CAUSAL_VALIDATION"
    | "CAUSAL_AMBIGUITY"
    | "WRONG_CAUSAL_LAYER";
  causal_validation?: CausalValidation;
  suggested_reframe?: string;
  upstream_questions?: string[];
}

Invariants

1. Causal validation precedes state collection. If the problem layer is ambiguous, required state cannot be meaningfully defined.
2. Symptom-layer requests in root-cause domains require reframing. The system MUST surface the causal gap before proceeding.
3. Dissolution dominates solution. If the problem can be dissolved (Kadai Barashi), this is preferable to repeated solution refinement.

Observed Signatures of Mislocalized Intervention

Systems should flag requests that exhibit:

• Repeated patching of downstream behavior
• Growing rule complexity and exception handling
• Apparent progress paired with persistent failure

These are diagnostic markers of symptom-layer intervention when the generating cause remains unaddressed.

Relationship to Other RFCs

Example

User Query: "My website is slow. What hosting should I use?"

Without RFC-0000:

• System collects state axes: current_host, traffic_volume, budget
• Produces hosting recommendation

With RFC-0000:

• System detects causal ambiguity: Is the problem the host, the application code, the database, the CDN, or the network?
• Returns CAUSAL_AMBIGUITY with upstream questions:
  • "Have you measured where the latency occurs?"
  • "Is the slowness consistent or load-dependent?"
  • "Have you profiled the application layer?"

The system refuses to optimize the wrong layer.

⸻