13 KiB
Multivariate Analysis of Coding.csv: Virtue Clustering and Associations
Date: 2026-03-28
Dataset: coding.csv
Texts Analyzed: 134
Unique Virtue Categories: 74
Average Virtues per Text: 2.78 (range: 1-5)
1. Executive Summary
This analysis examines 134 coded texts from two sources (AFP and PR) across 74 unique virtue categories. Using multiple multivariate techniques—clustering, network analysis, and association metrics—the study reveals:
- 4 distinct text clusters with one dominant cluster containing 86% of texts
- 3 major virtue communities representing different conceptual frameworks
- Strong ethical pairings (e.g., Care+Consent) that nearly always co-occur
- Source differences in conceptual complexity (AFP: more interconnected; PR: more focused)
2. Cluster Analysis of Texts
Using K-means clustering on binary virtue presence/absence vectors:
| Cluster | Size | Key Virtues | Sources | Interpretation |
|---|---|---|---|---|
| 1 | 5 texts | Memory, Imitation, Inheritance, Tradition | AFP, PR | Memory-focused texts - Historical and temporal continuity themes |
| 2 | 4 texts | Refusal, Embodiment, Resistance, Subversion | AFP only | Resistance discourse - Tactical opposition to systems |
| 3 | 115 texts | Adaptability, Tension Management, Accessibility, Design | AFP, PR | Core protocol cluster - Dominant protocol ethics discourse |
| 4 | 10 texts | Authenticity, Alignment, Inheritance | AFP, PR | Authenticity/Alignment cluster - Self-determination and tradition |
Key Finding: Cluster 3 represents the overwhelming majority (86%) of texts, suggesting a shared "protocol ethics" discourse across sources. Cluster 2 represents a distinct "resistance" discourse found only in AFP texts.
3. Strongest Virtue Associations
By Co-occurrence Count (raw frequency):
| Rank | Virtue Pair | Count | Notes |
|---|---|---|---|
| 1 | Accessibility + Situational Awareness | 4 | Practical context-sensitivity |
| 2 | Equity + Inclusivity | 3 | Justice framework |
| 3 | Balance + Tension Management | 3 | Managing contradictions |
By Jaccard Similarity (normalized association strength):
| Rank | Virtue Pair | Jaccard Index | Interpretation |
|---|---|---|---|
| 1 | Care + Consent | 0.750 | Nearly inseparable - Ethical foundation pair |
| 2 | Resistance + Subversion | 0.400 | Tactical cluster |
| 3 | Refusal + Subversion | 0.400 | Resistance tactics |
| 4 | Equity + Inclusivity | 0.375 | Justice-oriented |
| 5 | Refusal + Resistance | 0.333 | Activism tactics |
| 6 | Embodiment + Groundedness | 0.333 | Material presence |
| 7 | Agency + Freedom | 0.300 | Autonomy cluster |
Key Finding: The Care+Consent pairing (Jaccard = 0.750) is exceptionally strong, appearing together in 3 out of 4 possible texts where both concepts appear. This suggests an ethical foundation where care practices are inseparable from consent frameworks.
4. Virtue Communities (Network Analysis)
Using network thresholding on co-occurrence patterns, three major virtue communities were identified:
Community 1: "Protocol Mechanics" (~40 virtues)
Core operational virtues for protocol design and implementation
Central Members:
- Adaptability, Agency, Balance, Capture Resistance
- Care, Complex Systems Tolerance, Consent
- Constraint, Curiosity, Design, Emergent Properties
- Equity, Freedom, Institutional Critique, Iterative Development
- Networked Intelligence, Plurality, Replicability, Systems Thinking
Characteristics:
- Largest community spanning practical and ethical dimensions
- High connectivity to Adaptability and Systems Thinking (hub virtues)
- Brings together ethics (Care, Consent, Equity) with operational concepts (Design, Iterative Development)
Community 2: "Collective Intelligence" (3 virtues)
Focused on collaborative knowledge production
Members: Alignment, Collaboration, Networked Intelligence
Characteristics:
- Small but distinct community
- Emphasizes distributed, collaborative approach
- Connected to Community 1 through Networked Intelligence
Community 3: "Relational Ethics" (~9 virtues)
Focus on social and cultural connection
Members:
- Collectivity, Cultural Awareness, Empathy, Interdependence
- Plurality, Relationality, Respect, Spatial Awareness
- Plus contextual concepts
Characteristics:
- Strong ties to Community 1 through Relationality
- Emphasizes interpersonal and cultural dimensions
- Includes Plurality, suggesting diversity and multiplicity
5. Network Centrality Analysis
"Hub" Virtues (ranked by number of connections to other virtue types):
| Rank | Virtue | Connections | Key Neighbors |
|---|---|---|---|
| 1 | Adaptability | 25 | Agency, Resistance, Long-Term Vision, Design, Systems Thinking |
| 2 | Design | 23 | Agency, Equity, Emergent Properties, Inheritance, Constraint |
| 3 | Agency | 23 | Resistance, Inheritance, Refusal, Autonomy, Systems Thinking |
| 4 | Temporal Awareness | 19 | Emergent Properties, Long-Term Vision, Adaptability |
| 5 | Systems Thinking | 19 | Agency, Design, Long-Term Vision, Constraint |
| 6 | Collectivity | 17 | Interdependence, Agency, Shared Responsibility |
| 7 | Transgression | 17 | Refusal, Subversion, Care, Capture Resistance |
| 8 | Institutional Critique | 16 | Refusal, Design, Subversion, Agency |
| 9 | Plurality | 16 | Interdependence, Agency, Systems Thinking |
| 10 | Relationality | 16 | Interdependence, Accessibility, Care, Curiosity |
Key Finding: Adaptability is unequivocally the central hub of this virtue network, connecting to 25 other virtue concepts. This suggests it functions as a bridging concept across multiple ethical and practical domains.
6. Source Comparison (AFP vs. PR)
| Metric | AFP (62 texts) | PR (72 texts) | Interpretation |
|---|---|---|---|
| Unique virtue pairs | 221 | 143 | AFP texts show more conceptual diversity |
| Avg pairs per text | 4.06 | 2.22 | AFP texts are more conceptually dense |
| Network density | 8.2% | 5.3% | AFP has more interconnected virtue networks |
| Top virtues | Adaptability (8), Temporal Awareness (7), Collectivity (7), Institutional Critique (7) | Tension Management (10), Adaptability (9), Systems Thinking (9), Infrastructural Awareness (8) | AFP: critical/social; PR: technical/systemic |
AFP Code Profile (Academic/Critical)
- Dominant themes: Adaptability, Temporal Awareness, Collectivity, Institutional Critique
- Emphasis: Social processes, critical engagement, collective action
- Pattern: Higher virtue co-occurrence suggests more conceptually complex texts
PR Code Profile (Practical/Technical)
- Dominant themes: Tension Management, Systems Thinking, Infrastructural Awareness
- Emphasis: Technical complexity, managing contradictions, system design
- Pattern: More focused virtue profiles, strong emphasis on Adaptability
Key Finding: Both sources prioritize Adaptability, but AFP has more distributed emphasis across critical/social virtues, while PR emphasizes technical/systemic concepts. The 8.2% vs 5.3% network density difference suggests AFP texts engage with more complex conceptual interconnections.
7. Frequency Distribution
Top 30 Virtues by Frequency:
| Rank | Virtue | Count | % of Texts |
|---|---|---|---|
| 1 | Adaptability | 17 | 12.7% |
| 2 | Tension Management | 13 | 9.7% |
| 3 | Accessibility | 13 | 9.7% |
| 4 | Temporal Awareness | 11 | 8.2% |
| 5 | Design | 11 | 8.2% |
| 6 | Institutional Critique | 10 | 7.5% |
| 7 | Agency | 10 | 7.5% |
| 8 | Relationality | 10 | 7.5% |
| 9 | Infrastructural Awareness | 10 | 7.5% |
| 10 | Systems Thinking | 10 | 7.5% |
| 11 | Plurality | 9 | 6.7% |
| 12 | Transgression | 9 | 6.7% |
| 13 | Collectivity | 8 | 6.0% |
| 14 | Inheritance | 8 | 6.0% |
| 15 | Authenticity | 7 | 5.2% |
| 16 | Long-Term Vision | 7 | 5.2% |
| 17 | Equity | 6 | 4.5% |
| 18 | Capture Resistance | 6 | 4.5% |
| 19 | Respect | 6 | 4.5% |
| 20 | Cultural Awareness | 6 | 4.5% |
| 21 | Spatial Awareness | 6 | 4.5% |
| 22 | Interdependence | 6 | 4.5% |
| 23 | Shared Responsibility | 6 | 4.5% |
| 24 | Situational Awareness | 6 | 4.5% |
| 25 | Memory | 5 | 3.7% |
| 26 | Embodiment | 5 | 3.7% |
| 27 | Inclusivity | 5 | 3.7% |
| 28 | Balance | 5 | 3.7% |
| 29 | Reciprocity | 5 | 3.7% |
| 30 | Emergent Properties | 5 | 3.7% |
8. Key Insights and Implications
8.1 The Three Pillars of Protocol Ethics
The analysis reveals three conceptual pillars that structure this discourse:
-
Adaptive Ethics (centered on Adaptability and Design): The capacity to adjust, learn, and evolve protocols in response to changing conditions
-
Relational Justice (centered on Care, Consent, Equity, Inclusivity): Ethical frameworks emphasizing relationship, respect, and justice
-
Systemic Resistance (centered on Refusal, Subversion, Institutional Critique): Tactical opposition and critique of existing systems
8.2 The Adaptability Paradigm
The overwhelming centrality of Adaptability (highest frequency, highest connectivity) suggests this is the core organizing concept. It bridges:
- Ethical dimensions: Equity, Care, Consent
- Operational dimensions: Design, Iterative Development, Systems Thinking
- Resistance dimensions: Capture Resistance, Resistance, Agency
8.3 Source Convergence and Divergence
- Convergence: Both sources treat Adaptability as central, suggesting a shared understanding that protocols must be capable of change
- Divergence: AFP emphasizes critical/social dimensions (Institutional Critique, Collectivity), while PR emphasizes technical/systemic dimensions (Tension Management, Systems Thinking)
- Integration: The most conceptually dense texts (highest network density) come from AFP, suggesting critical theory provides more complex conceptual interconnections
8.4 Unexpected Pairings
Several virtue pairs show unexpected strength:
- Care + Consent (0.750): Suggests an ethics of care cannot exist without consent frameworks
- Refusal + Subversion (0.400): Tactical language clusters together
- Equity + Inclusivity (0.375): Justice requires both fair distribution and openness
8.5 The Resistance Cluster
The small cluster of resistance-focused texts (4 texts in Cluster 2) represents a distinct discourse that:
- Appears only in AFP texts
- Coheres around Refusal, Resistance, Subversion, Embodiment
- Serves as a strategic counterpoint to the dominant protocol design discourse
- May represent the critical "edge cases" that test protocol boundaries
9. Methodological Notes
Analytic Techniques Used:
- K-Means Clustering (k=4): Identified text groups based on virtue profile similarity
- Network Analysis: mapped virtue co-occurrences and calculated centrality (degree = number of connections)
- Jaccard Similarity: normalized measure of virtue pair association (intersection/union)
- Community Detection: threshold-based clustering of highly connected virtue groups
Limitations:
- Small dataset (134 texts) limits statistical power
- K-means clustering is sensitive to initialization (used deterministic starting points)
- Binary coding (presence/absence) doesn't capture intensity or salience
- Limited to virtues 1-5; other dimensions not analyzed
Generated Files:
| File | Description |
|---|---|
cooccurrence_matrix.csv |
25×25 matrix of virtue co-occurrence counts |
jaccard_similarity_matrix.csv |
25×25 similarity matrix (Jaccard indices) |
strong_associations.csv |
Top 50 virtue pairs with association metrics |
virtue_profiles.json |
Individual virtue profiles for each text |
10. Recommendations for Further Analysis
-
Qualitative Deep Dive: Examine the 4 resistance-focused texts (Cluster 2) and the 10 authenticity-focused texts (Cluster 4) to understand the distinct discourses
-
Temporal Analysis: If dates are available, analyze how virtue frequencies change over time
-
Semantic Mapping: The Care+Consent pairing could be explored through close reading to understand the conceptual linkage
-
Source-Specific Models: Consider whether different theoretical frameworks might be needed for AFP vs. PR texts
-
Expand to Other Codes: Analysis currently limited to Virtue_1 through Virtue_5; expanding to other coding categories could reveal additional patterns
-
Visualization: Generate network graphs of virtue communities to make relationships visually explicit
Analysis generated using Python standard library (no external packages required). All calculations are fully reproducible.