A Non-Reductive Relational Analysis System
KindSense is a computational framework for understanding complex systems without collapsing their inherent complexity. It implements a multi-hypothesis, omni-lateral approach to sense-making that preserves all signals, trajectories, and explanations.
Traditional prediction systems reduce complexity to make it manageable. KindEarth takes the opposite approach: preserve complexity while making sense of it.
- No Data Loss - Every metric, trajectory, and hypothesis is preserved
- Omni-Lateral Exploration - Scan inward, outward, forward, and inverse from every focal point
- Multi-Logic Analysis - Apply 10 different logic systems to the same data in parallel
- Relational Thinking - Focus on connections, not just individual metrics
- Non-Extractive Epistemology - Understanding emerges without destroying the source
MetricField: Multi-dimensional metric field with zero discardingRelationalMatrix: Pairwise metric interactions with forward/inverse transformationsCurvatureComputer: Stability classification and regime shift detection
FocalScanner: Creates and manages focal points in metric space- Omni-lateral scanning: Inward, outward, forward, and inverse directional analysis
- Relational tension tracking: Promotes high-tension regions for deeper exploration
SeedBank: Preserves all trajectory seeds without deletion- Pareto frontiers: Avoids single-optimum collapse
- Sensitivity analysis: Tracks how small changes create large effects
- 10 Logic Systems:
- Causal ↔ Reciprocal
- Linear ↔ Circular
- Economic ↔ Ecological
- Scarcity ↔ Abundance
- Individual ↔ Relational
- Parallel execution: Same data, multiple interpretations
- Invariant detection: What stays true across all logics?
git clone https://github.com/KindPath-Collective/KindSense.git
cd KindSense
pip install -e .
pip install flask numpy pytestfrom kindearth.core.metric_field import MetricField
from kindearth.hypothesis.logic_transforms import HypothesisEngine
# 1. Create a metric field
field = MetricField()
# 2. Ingest data (nothing is discarded)
data = {"metric_a": 1.5, "metric_b": 2.3, "metric_c": None}
field.ingest(data)
# 3. Get metric matrix
matrix = field.get_metric_matrix()
# 4. Run multiple hypotheses in parallel
engine = HypothesisEngine()
metric_names = ["metric_a", "metric_b", "metric_c"]
outcomes = engine.run_parallel(matrix, metric_names)
for outcome in outcomes:
print(f"{outcome.logic_type.value}: stability={outcome.relational_stability:.3f}")Your first R&D project analyzing the Australian Dollar rally is ready!
python webapp/app.py
# Visit http://localhost:5000- Live AUD/USD Rate - Real-time updates
- 10 Logic Perspectives - Causal, Ecological, Abundance, etc.
- Mean Explanation - "The rally is driven by..."
- Inverse Explanation - "The rally would reverse if..."
- Sensitivity Zones - Where small changes matter most
- Stability State - Stable, transitional, or unstable
- Multi-dimensional metric field
- Pairwise metric interactions
- Weak/missing metrics as valid signals
- Persistent field snapshots
- Time-aligned ingestion
- Polarity inversion
- Forward and inverse outcomes
- Divergence calculation
- Curvature as gradient magnitude
- Stability state classification
- Focal set sampling
- Sliding focal windows
- Inward/outward/forward/inverse scans
- Directional delta recording
- Focal point promotion
- Dormant region preservation
- Gap metrics and evolution
- Convergence/divergence detection
- Non-destructive reduction
- Trajectory seed bank
- Attention weighting without deletion
- Pareto frontier computation
- Premature convergence detection
- Inverse trajectory tracking
- Sensitivity analysis
- 10 logic systems
- Parallel hypothesis execution
- Relational stability comparison
- Invariant detection
- Hypothesis ecology management
KindEarth/
├── src/
│ └── kindearth/
│ ├── core/
│ │ ├── metric_field.py # Multi-dimensional field (zero discard)
│ │ ├── relational.py # Pairwise metric interactions
│ │ └── curvature.py # Stability & regime shift detection
│ ├── reduction/
│ │ └── seed_bank.py # Non-destructive trajectory preservation
│ ├── focal/
│ │ └── focal_point.py # Omni-lateral scanning
│ ├── hypothesis/
│ │ └── logic_transforms.py # 10 logic systems + parallel engine
│ ├── output/
│ │ └── explainer.py # Multi-perspective explanation generator
│ └── case_studies/
│ └── aud_rally.py # AUD rally end-to-end pipeline
├── webapp/
│ ├── app.py
│ └── templates/
│ ├── base.html # Shared layout with nav
│ ├── index.html # Dashboard home
│ ├── analysis.html # Multi-logic hypothesis table
│ ├── focal.html # Omni-lateral scan results
│ ├── curvature.html # Stability & regime shifts
│ ├── explanations.html # Mean/inverse narratives
│ └── sensitivity.html # Sensitivity zones & invariants
├── tests/
│ └── test_metric_field.py # 66 tests across all modules
├── requirements.txt
├── setup.py
└── README.md
- Reduce complexity → single prediction
- Discard weak signals → miss early warnings
- Single logic → narrow perspective
- Extract insights → destroy context
- Preserve complexity → multiple valid explanations
- Honor weak signals → early warning system
- 10 parallel logics → holistic understanding
- Non-extractive → context intact
"You're not just building a prediction system — you're building a new epistemology."
KindEarth is an experiment in making sense without destroying. It asks: What if we could understand complex systems without reducing them to single stories? What if weak signals matter? What if dormant possibilities are worth preserving?
This is not just software. It's a different way of knowing.
- GitHub Issues: Bug reports and feature requests
- Discussions: Questions and ideas
- Email: sam@kindpathcollective.org
MIT License
Built with care by the KindPath Collective 🌍
A non-reductive epistemology encoded in Python