Latest Release: v1.1 — Structural Integrity Update
Formalizes enforcement rules and failure semantics. See release notes for details.

Phase-Aligned Inquiry

Most collaborative research fails not because the ideas are wrong, but because the structure allows confirmation bias and premature interpretation to corrupt results.

Phase-Aligned Inquiry (PACI) is a protocol designed to prevent that.

The Problem

Collaborative theoretical research—whether conducted by individual researchers, AI-assisted teams, or multi-agent systems—commonly fails through structural mechanisms rather than intellectual deficits:

• Confirmation bias at the execution layer: Measurement and interpretation are performed by the same component, allowing theory to contaminate data collection.
• Premature interpretation: Early signals trigger claims of success before systematic falsification is complete.
• Narrative momentum override: Compelling preliminary results cause contradictory evidence to be explained away rather than confronted honestly.

These failure modes persist because single-agent or role-conflated systems lack the structural separation necessary to prevent them. Adding more participants or computational resources does not solve the problem—it often amplifies it.

The Solution: Phase-Aligned Collaborative Inquiry (PACI)

PACI is a research methodology that prevents structural failures through role separation, sequence enforcement, and pre-commitment to failure criteria.

The PACI workflow separates inquiry into five strictly ordered phases.

Figure: Phase-Aligned Collaborative Inquiry protocol structure showing role separation and one-way phase enforcement.

1. Role separation: Generation, constraint, execution, falsification, and interpretation are performed by functionally distinct components. No single participant both produces and validates results.
2. Sequence enforcement: Each phase must complete before the next begins. Later phases cannot contaminate earlier ones. Violations invalidate downstream phases and require restart from the last uncontaminated phase.
3. Pre-commitment to failure criteria: Success and failure conditions are defined explicitly before data exists, preventing retroactive redefinition of observables.

PACI does not guarantee correct, fast, or interesting results. It guarantees that when hypotheses fail, they fail cleanly and informatively, with complete documentation of where and why failure occurred.

In practice, PACI prevents:

• Defining success after seeing results
• Interpreting data during execution
• Validating outcomes with the same component that generated them
• Quietly redefining thresholds after data exists

The Case Study

This repository demonstrates PACI through a physics case study where a falsifiable hypothesis about emergent spacetime properties was rigorously tested and cleanly falsified.

The protocol prevented premature interpretation, forced invariance testing, and enabled conceptual reclassification when the original claim failed.

The case study exists only to show that PACI works under real epistemic pressure.

The hypothesis failed. The data revealed patterns unrelated to the original claim. These are documented separately in case-study/07-what-the-wreckage-revealed.md — not as conclusions, but as preserved observations.

What This Repository Contains

The protocol is the primary contribution. Everything else in this repository exists to demonstrate, support, or contextualize it.

/methodology — The Core Contribution

• PACI Protocol ← Start here
• Role architecture details
• Sequence discipline framework
• Usage guidelines and preconditions
• Hard requirements and self-assessment checklist

/case-study — Demonstration

• Complete documentation of the HQG hypothesis lifecycle
• Phase-locked artifacts (Define → Constrain → Execute → Falsify → Interpret)
• Lessons extracted from clean falsification
• 07-what-the-wreckage-revealed.md — what the failure exposed that was unrelated to the original question

/technical — ⚠️ Planned

This section is not yet populated. It will contain simulation code, parameter sweep methodology, and data analysis procedures when available. The v1.0 tagged release does not include technical implementation files.

/appendix — ⚠️ Planned

This section is not yet populated. It will contain fuller conversation transcripts, decision point documentation, and multi-agent process notes when available. The v1.0 tagged release does not include appendix content.

Quick Start

If you want to use PACI for your own research:

• Read methodology/01-paci-protocol.md first. It contains the complete specification, including preconditions, role definitions, and failure patterns.

If you want to understand how PACI was derived:

• Start with case-study/00-context.md, then follow the phase documentation sequentially to see how falsification unfolded.

If you're evaluating whether PACI is credible:

• Read Section 5 (Guarantees / Does NOT Guarantee) and Section 4.6 (Observable Failure Patterns) in the protocol. These sections define what PACI can and cannot do.

If you're a skeptic:

• Good. Read methodology/01-paci-protocol.md Section 2 (Preconditions). If your research domain or team structure does not meet these hard requirements, PACI will not work for you. The protocol is explicit about its limitations.

If you're curious about what failed and what it revealed:

• Read case-study/07-what-the-wreckage-revealed.md. The original hypothesis was falsified. Something else was sitting in the data.

When To Use PACI

PACI is appropriate when:

• Your hypothesis is falsifiable (makes testable predictions)
• Confirmation bias is a significant risk
• Multiple participants or agents can specialize in different roles
• You can afford sequence discipline over speed
• You are prepared to document and learn from null results

PACI is not appropriate for:

• Pure ideation or brainstorming
• Unfalsifiable philosophical questions
• Consensus-building exercises
• Single-agent research without role separation
• Time-critical work requiring rapid iteration

Status

Component	Status
PACI Protocol (v1.1)	✅ Complete — structural integrity update
Case Study (HQG)	✅ Complete — all phases documented
Wreckage Note	✅ Complete
/technical implementation files	⚠️ Planned — not in current release
/appendix transcripts & meta-analysis	⚠️ Planned — not in current release

The core protocol and case study are stable. Planned sections will be added in future releases without affecting protocol integrity.

Version

PACI v1.1 — Structural Integrity Update (March 2026)

Changes from v1.0:

• Explicit enforcement rules for phase violations (Section 6)
• Defined INVALID vs INCONCLUSIVE outcome distinction
• Formalized allowed and prohibited rollback paths
• Introduced Temporal Innocence as a named core principle
• Clarified falsification scope constraints
• Annotated case study to align with updated enforcement vocabulary

Citation

If you use or adapt PACI in your research, please cite:

Phase-Aligned Collaborative Inquiry (PACI): A Protocol for Falsifiable Multi-Agent Research. (2026). Retrieved from https://github.com/leenathomas01/phase-aligned-inquiry

---

Related Work

For a complete catalog of related research:
📂 Research Index

Thematically related:

• PARP — Governance under opacity
• Embodied Agent Governance — External failure knowledge for agents
• The Continuity Problem — Why governance must precede autonomy

---

PACI is not a guarantee of success. It is a guarantee of honesty.