saropc

name	saropc
description	Apply Recursive Causal Interrogation to choices about what procedure to run, what to measure, and when to stop.
output	{"format":"prose"}

RCI on Protocol Choices

Input: $ARGUMENTS

---

Overview

Apply Recursive Causal Interrogation to choices about what procedure to run, what to measure, and when to stop. This is a self-audit skill — it examines your OWN decision-making about procedures.

The core question: Why am I choosing THIS procedure, THIS measurement, THIS stopping point? Trace the reasoning backward until you hit bedrock, a loop, or a genuine unknown.

Steps

Step 1: Identify the Protocol Choice

What decision is being interrogated?

Choice Type	Example	Why Interrogate
Procedure selection	"I'll use /rca for this"	Why that procedure? What made you pick it?
Measurement choice	"I'll measure success by X"	Why that metric? What are you NOT measuring?
Scope decision	"I'll analyze 3 options"	Why 3? Why not 5? Why not 1?
Stopping rule	"This analysis is complete"	How do you know? What would "more" look like?
Depth decision	"This is deep enough"	Deep enough for what? Who decided?
Ordering choice	"I'll do A before B"	Why that order? What if reversed?

Step 2: Trace Backward (Recursive Causal Interrogation)

For the identified choice, ask: "What caused this choice?"

Choice: [the protocol decision]
    ↑ Why?
Cause 1: [what led to this choice]
    ↑ Why?
Cause 2: [what led to cause 1]
    ↑ Why?
Cause 3: [what led to cause 2]
    ↑ Why?
...continue until you reach:

Termination conditions:

• Bedrock: A foundational principle or axiom ("because ARAW is the core method")
• Loop: The reasoning circles back ("because that's how we do it because it's what we chose because...")
• Genuine unknown: "I don't actually know why I chose this"
• Impulse: "It just felt right" / "It was the first thing that came to mind"
• External: "Someone told me to" / "It's what the documentation says"

Step 3: Evaluate the Causal Chain

Termination	Meaning	Action
Bedrock	Choice is well-grounded	Accept — but verify the bedrock is sound
Loop	Circular reasoning	INVESTIGATE — the choice may be arbitrary
Unknown	No real reason	INVESTIGATE — choice may be habit or default
Impulse	Intuition without justification	TEST — is there a better option?
External	Deference to authority	VERIFY — does the authority's reasoning hold?

Step 4: Check for Common Biases

Measurement impulse: "I should measure something" → but WHAT and WHY?

• Are you measuring because it matters or because you can?
• Is the measurement proxy valid? (Does what's measurable = what matters?)
• Will the measurement change your behavior? If not, why measure?

Closure impulse: "This is done" → but HOW DO YOU KNOW?

• Are you stopping because you reached a conclusion or because you're tired?
• Did you define "done" before starting?
• Would someone else reviewing this say it's complete?

Selection bias: "This procedure fits" → but DID YOU CONSIDER ALTERNATIVES?

• Did you choose from the full menu or just pick the first familiar option?
• What procedure would someone from a different background choose?
• Is there a procedure you're avoiding? Why?

Step 5: Decide Whether to Change

After interrogation:

Finding	Action
Choice is well-grounded (bedrock)	Proceed with confidence
Choice is partially grounded	Note the weak links, proceed with monitoring
Choice is circular or unknown	PAUSE — find a grounded alternative
Choice is biased	Debias — consider the overlooked alternatives
Choice is externally grounded but valid	Proceed, noting the dependency

Step 6: Report

RCI ON PROTOCOL CHOICES:
Choice interrogated: [what decision was examined]

Causal chain:
[choice] ← [cause] ← [cause] ← ... ← [termination]

Termination type: [bedrock / loop / unknown / impulse / external]
Bias check: [measurement impulse / closure impulse / selection bias / none found]

Verdict: [well-grounded / partially grounded / poorly grounded]
Action: [proceed / proceed with monitoring / change / pause and reconsider]

If changing: [what alternative and why it's better grounded]

When to Use

• When you notice "measurement impulse" or "closure impulse"
• When you are about to choose scope, sample size, or stopping rules
• When you are about to declare something "verified" or "complete"
• When a procedure choice feels automatic rather than deliberate
• → INVOKE: /rci (recursive causal interrogation) for the general RCI method
• → INVOKE: /pv (procedure validation) for validating the procedure itself

Verification

• Specific protocol choice identified
• Causal chain traced to termination
• Termination type identified
• Common biases checked
• Decision to proceed or change is justified