O0-OBS-008

Differentiation Protocol: Unity vs. Absorption vs. Dissociation

**Version:** 1.0

**Research status:** PLANNED (PREREGISTERED PROTOCOL)

CLAIM STATUS: UNTESTED PREDICTION
EVIDENCE TYPE: COMPUTATIONAL/THEORETICAL (PLANNED)
PHYSICAL VALIDATION: NONE
INDEPENDENT REPLICATION: NOT YET ATTEMPTED
PHILOSOPHICAL PROVENANCE: O/0 ARCHIVE
ARCHIVE ENDORSEMENT: LIMITED TO REPORTED RESULT

Abstract

This document preregisters a differentiation protocol for distinguishing three computationally distinct states that all involve boundary change: Unity (boundary dissolution with maintained internal coherence), Absorption (boundary dissolution with internal coherence loss), and Dissociation (boundary fragmentation without dissolution). These three states may appear superficially similar—all involve altered boundary conditions—but have distinct computational signatures and distinct implications for agent identity and continuity. We define each state formally, specify measurable criteria, and preregister decision rules for classification. This protocol is designed to be applied to simulation data from SIM-001 and potentially to neural data from human studies (OBS-009).

Note: Conceptual provenance from the O/0 philosophical archive does not constitute empirical support.

Source proposition

From O0-META-011: "Boundary dissolution admits multiple forms: the observer may dissolve into a larger whole while maintaining coherence (unity), dissolve while losing coherence (absorption), or fragment while boundaries remain nominally intact (dissociation). These are computationally distinguishable." This protocol operationalizes the distinction.

Scientific audit

• **Testability:** HIGH. Each state is defined by measurable criteria (boundary integrity + internal coherence).
• **Falsifiability:** YES. If the three states cannot be reliably distinguished in practice (classification accuracy < 60%), the tripartite framework is empirically vacuous.
• **Novelty:** HIGH. Formal computational definitions of these phenomenological concepts are novel.
• **Potential confounds:** Intermediate or mixed states may exist that don't fit the tripartite classification. Addressed by including an "unclassified" category and testing classification completeness.

Research question

1. Can Unity, Absorption, and Dissociation be reliably distinguished using computational measures?

2. What preconditions predict which state will occur?

3. Are transitions between these states ordered (e.g., Unity → Absorption → Dissociation) or independent?

Operational definitions

**State 1: UNITY**

• Boundary condition: Dissolved. I(interior; exterior | boundary) > 0.5 bits (boundary no longer screens interior from exterior).
• Internal coherence: MAINTAINED. Φ(interior) remains within 80% of pre-dissolution value. The internal dynamics remain organized.
• Agent identity: Continuous. Information in the former interior is preserved in the larger system.
• Metaphor: A drop of dye dissolving into water while maintaining its color intensity.

Formal criteria:

• C_unity = (I_boundary > 0.5) ∧ (Φ_internal > 0.8 × Φ_baseline) ∧ (MI_preservation > 0.7)
• Where MI_preservation = I(state_internal(t); state_internal(t-Δ)) / I(state_internal(t_pre); state_internal(t_pre-Δ))

**State 2: ABSORPTION**

• Boundary condition: Dissolved. I(interior; exterior | boundary) > 0.5 bits.
• Internal coherence: LOST. Φ(interior) drops below 30% of pre-dissolution value. Internal dynamics become indistinguishable from surroundings.
• Agent identity: Discontinued. Information in the former interior is dissipated.
• Metaphor: A drop of dye dissolving into water and becoming uniformly diluted.

Formal criteria:

• C_absorption = (I_boundary > 0.5) ∧ (Φ_internal < 0.3 × Φ_baseline) ∧ (MI_preservation < 0.3)

**State 3: DISSOCIATION**

• Boundary condition: Intact but fragmented. Multiple sub-boundaries form: one original region becomes 3+ fragments each with I < 0.1 bits.
• Internal coherence: PARTITIONED. Φ(total_original) drops, but Φ(each_fragment) may be high.
• Agent identity: Multiplied. Multiple successors, none fully preserving the original.
• Metaphor: A crystal shattering into pieces that each retain internal structure.

Formal criteria:

• C_dissociation = (fragment_count ≥ 3) ∧ (Σ Φ_fragment > 0.5 × Φ_baseline) ∧ (max(Jaccard(fragment_i, original)) < 0.5)

**Auxiliary measures:**

• **Φ (integrated information):** Approximated as normalized mutual information between bipartitions of the region. Φ = min_partition MI(part_A; part_B) / H(region).
• **MI_preservation:** Temporal autocorrelation of internal state patterns.
• **Fragment count:** Number of distinct bounded regions arising from one predecessor within 50 timesteps.

Hypothesis

H1: The three states (Unity, Absorption, Dissociation) are empirically distinguishable with classification accuracy > 80% using the defined criteria.

H2: Each state has distinct precursors: Unity is preceded by gradual boundary weakening with stable internals; Absorption is preceded by simultaneous internal degradation and boundary weakening; Dissociation is preceded by internal conflict (high variance in regional dynamics) with intact boundaries.

H3: The states form a partial order: Absorption can follow Unity (boundary dissolves, then coherence is lost), but Unity cannot follow Absorption (coherence cannot be recovered after dissolution).

Null hypothesis

H0_1: The three states cannot be reliably distinguished (classification accuracy < 60% using blinded raters or automated criteria).

H0_2: Precursor patterns are identical for all three states.

H0_3: Transitions are fully reversible (any state can follow any other state with equal probability).

Competing explanations

1. **Continuous spectrum:** The three states may be points on a continuum rather than distinct categories. Tested by examining the distribution of (boundary_integrity, internal_coherence) pairs—if bimodal/trimodal, categories are supported.

2. **Threshold artifact:** The specific thresholds (0.5, 0.8, 0.3) may create artificial categories that don't reflect natural clustering. Tested by applying unsupervised clustering (k-means, DBSCAN) to the measurement space and checking if k=3 is optimal.

3. **Timescale dependence:** At different timescales, the same event might be classified differently (e.g., Unity at t=10 becomes Absorption at t=1000). Addressed by specifying measurement window (100 timesteps post-event) and testing robustness across windows.

Formal model

State space representation:

• Each boundary-change event is characterized by a point in ℝ² : (boundary_integrity, internal_coherence)
• Boundary_integrity ∈ [0, 1]: 0 = fully dissolved, 1 = fully intact
• Internal_coherence ∈ [0, 1]: 0 = fully disordered, 1 = maximally organized

Expected clustering:

• Unity cluster: boundary_integrity < 0.3, internal_coherence > 0.7
• Absorption cluster: boundary_integrity < 0.3, internal_coherence < 0.3
• Dissociation cluster: boundary_integrity fragmented (multiple sub-boundaries), coherence partitioned

Classification algorithm:

1. Detect boundary-change event (Δ boundary_integrity > 0.3 over 50 timesteps)

2. Measure boundary_integrity and internal_coherence at t_event + 100

3. Measure fragment_count and MI_preservation

4. Apply decision tree:

• If boundary_integrity < 0.3 AND coherence > 0.7: → UNITY
• If boundary_integrity < 0.3 AND coherence < 0.3: → ABSORPTION
• If fragment_count ≥ 3 AND max_fragment_overlap < 0.5: → DISSOCIATION
• Otherwise: → UNCLASSIFIED

Methods

1. **Event generation:** Run SIM-001 for 50,000 timesteps on 100×100 grid. Record all boundary-change events (≥ 100 expected based on OBS-005 rates extrapolated).

2. **Automated classification:** Apply decision tree to each event. Record classification and confidence (distance from decision boundaries).

3. **Clustering validation:** Apply unsupervised clustering to (boundary_integrity, internal_coherence, fragment_count) space. Test whether k=3 provides optimal separation (silhouette score, BIC for Gaussian mixtures).

4. **Inter-rater reliability:** Have two independent analysis pipelines (different code, different authors) classify the same events. Compute Cohen's κ.

5. **Precursor analysis:** For each classified event, extract 500-timestep pre-event trajectories of all metrics. Train discriminative model to predict event type from precursors.

6. **Transition analysis:** For events that follow other events (within same region lineage), compute transition matrix. Test for forbidden transitions.

7. **Timescale robustness:** Repeat classification using measurement windows of {50, 100, 200, 500} timesteps post-event.

8. **Repetitions:** 20 independent runs.

Controls

• **Stable region control:** Apply classification criteria to randomly selected timesteps for stable regions (no boundary change). All should classify as "none of the above" (UNCLASSIFIED). If not, criteria are too permissive.
• **Simulated events control:** Artificially create each event type by manipulating grid states. Verify that the classification algorithm correctly identifies them.
• **Noise robustness control:** Add measurement noise to boundary_integrity and coherence values. Test classification stability.

Predictions

**Expected Outcome Matrix:**

| Metric | Expected | Decision threshold |

|--------|----------|-------------------|

| Classification accuracy (automated) | > 85% | > 70% to support framework |

| Inter-rater κ | > 0.75 | > 0.6 to support reliability |

| Optimal k (unsupervised) | 3 | k=3 must be within top-2 by BIC |

| Event rate (Unity) | 15-25% of boundary events | - |

| Event rate (Absorption) | 30-45% of boundary events | - |

| Event rate (Dissociation) | 10-20% of boundary events | - |

| Unclassified | 15-30% | < 50% for framework utility |

| Precursor discriminability (AUC) | > 0.75 | > 0.65 to support H2 |

| Forbidden transitions found | ≥ 1 | Any asymmetry supports H3 |

**Decision rules:**

• If automated accuracy > 70% AND κ > 0.6 AND k=3 is optimal: **SUPPORT tripartite framework**
• If accuracy < 60% OR κ < 0.4 OR k=3 is not in top-3: **REJECT framework, revise definitions**
• If unclassified > 50%: **Framework incomplete, additional categories needed**
• If precursor AUC > 0.65: **SUPPORT H2 (distinct precursors)**
• If transition asymmetry p < 0.05 (χ² test): **SUPPORT H3 (partial ordering)**

Falsification criteria

The tripartite framework is falsified if:

1. Unsupervised clustering does not support k=3 (k=2 or k=1 is strongly preferred by all metrics), OR

2. Automated classification accuracy < 60% (states are not reliably distinguishable), OR

3. > 60% of events are UNCLASSIFIED (categories cover minority of phenomena).

Results / Expected Outcomes

Protocol not yet applied to data. Expected outcomes based on theoretical analysis:

• Three states should be distinguishable because they occupy different regions of the (boundary, coherence) plane
• Absorption expected to be most common (dissolution typically disrupts coherence)
• Unity expected to be least common (maintaining coherence during dissolution requires specific conditions)
• Dissociation expected to be intermediate
• Some transition asymmetry expected (dissolution is easier than reconstitution)

Uncertainty

• **Threshold calibration:** The specific numerical thresholds (0.5, 0.8, 0.3) are informed estimates. Optimal thresholds may differ from preregistered values. Sensitivity analysis planned.
• **Φ approximation:** True integrated information is computationally intractable. Our approximation (minimum bipartition MI) may not capture all relevant coherence structure.
• **Event frequency:** If events are very rare (< 20 total in 50,000 steps), statistical power for transition analysis will be low.
• **Category completeness:** There may be a fourth (or fifth) boundary-change type not anticipated by this framework.

Limitations

1. The tripartite classification is motivated by phenomenological reports (meditation traditions) but tested only computationally—generalization to human experience is not established.

2. Φ approximation may classify some high-coherence states as low-coherence due to approximation error.

3. Binary classification (boundary intact vs. dissolved) may not capture gradual dissolution dynamics.

4. "Identity continuity" is defined information-theoretically but philosophical identity criteria may differ.

5. The protocol assumes clear event boundaries; gradual state changes may not fit the event-based analysis framework.

Replication status

Protocol preregistered. Not yet applied to data.

Data and code

• Protocol specification: This document
• Classification algorithm: decision_tree_classifier.py (to be implemented)
• Validation suite: clustering_validation.py (to be implemented)
• Pre-registration: This document. Timestamp: 2026-07-26.
• Planned data source: SIM-001 extended runs (50,000 steps)

Relationship to philosophical archive

The archive describes Unity, Absorption, and Dissociation as phenomenologically distinct states reported in contemplative traditions. This protocol translates those phenomenological descriptions into computational criteria, enabling empirical test. Success of the classification scheme would support the claim that these are genuinely distinct states (not just different descriptions of the same thing). Failure would suggest the distinctions may be phenomenological artifacts without computational correlates.

References

• Tononi, G. (2008). Consciousness as integrated information. *Biological Bulletin*, 215(3).
• Josipovic, Z. (2019). Nondual awareness: Consciousness-as-such as non-representational reflexivity. *Progress in Brain Research*, 244.
• Millière, R., et al. (2018). Psychedelics, meditation, and self-consciousness. *Frontiers in Psychology*, 9.
• Bayne, T., & Chalmers, D. J. (2003). What is the unity of consciousness? In *The Unity of Consciousness* (Cleeremans, A., ed.).
• Metzinger, T. (2020). Minimal phenomenal experience. *Philosophy and the Mind Sciences*, 1(I).

Revision history

| Date | Version | Changes |

|------|---------|---------|

| 2026-07-26 | 1.0 | Initial protocol pre-registration |