Source-linked evidence
Preserve where an observation came from, what it measured, and which interpretation boundary applies.
EVIDENCE OBSERVATORY
Evidence Observatory
A proprietary evidence platform for reconciling wearable, route, episodic, movement, transportation, and documentary sources without losing their identity or limits.
The Evidence Observatory exists because a credible mobility record needs more than a chart. It needs source provenance, compatible comparisons, visible samples, preserved missingness, scientific identity, and a consistent path from evidence to public explanation.
Why the platform exists
Keep Context Intact from Source to Review
Wearable and mobility records answer different questions. A route shows functional travel. Overnight recovery supplies longitudinal context. A targeted RRI session adds episodic specificity. An access record documents what happened in a public system.
The platform keeps those sources connected enough to review together while preventing them from being flattened into one overclaimed signal.
Within-person comparison
Organize comparable mobility and transportation contexts while retaining samples, dates, units, and missingness.
Reviewer-safe presentation
Publish bounded findings that can be inspected without exposing private raw records or asking a browser to recalculate science.
Evidence source roles
Longitudinal and Episodic Evidence Differ
Longitudinal context
WHOOP
Activity, strain, sleep, and overnight recovery patterns across time. WHOOP recovery HRV/RMSSD is not session HRV.
Functional travel
Strava
Activity, route, distance, duration, speed, and repeated mobility context.
Episodic evidence
Polar H10 and Kubios
Targeted RRI, HRV/RMSSD, ECG, acceleration, and movement evidence where available.
Source reconciliation relates compatible records; it does not make their fields interchangeable or fabricate values that a source did not provide.
Separate scientific identities
FSI, CSS, and Tensor v0.3
Each analytical identity asks a different question. The approved definitions below come from the synchronized publication resource.
Loading approved scientific identities…
Fractal Stability Index (FSI) is not a pain score.
Cohort Similarity Score (CSS) is not descriptive return direction and is not another name for FSI.
Tensor v0.3 is the frozen multidimensional model identity and is not reconstructed by this website.
Explainable evidence architecture
One Consistent Path to the Public Site
The public architecture is intentionally simple at the point of review, even though the underlying evidence work is rigorous.
- Source-linked evidenceWearables, routes, episodic sensors, movement context, and reviewed documents.
- Canonical scientific viewsValidated server-side evidence with authoritative formulas and cohort decisions.
- MCP evidence interfaceA consistent read-only presentation of approved resources and graph identities.
- Publication contractA destination-safe, versioned projection with units, samples, dates, limitations, sources, and hashes.
- Static website bundleSame-origin JSON synchronized atomically into the public-site repository.
- Render-only readerAccessible charts and tables display approved values without recomputing science.
Consistent evidence interface
What MCP Contributes
MCP provides one server-side interface for the Evidence Observatory's approved presentation. That keeps the publication exporter aligned with the same evidence resources used for review rather than creating a second scientific interpretation path.
Destination-safe publication
What the Contract Contributes
The publication contract selects only public-safe fields, validates destinations and hashes, and keeps websites from querying the warehouse, calling local services, or recalculating FSI, CSS, Tensor, burden, or cohort membership.
Auditability at the point of reading
Values Stay Beside Their Context
Provenance
Source labels and evidence scope identify what contributed to the public interpretation.
Samples
Sample counts remain metric-specific, including explicit small-sample qualifications.
Units
Displayed values retain their supplied units; unlike measures are not silently merged.
Missingness
Unavailable observations remain unavailable rather than becoming zero or being inferred.
Limitations
Interpretation is accompanied by what the source and comparison cannot establish.
Data-through date
The public record states how current the approved source projection is.
What the system supports
Research, Mobility, Legal Review, and Wearable Intelligence
The same evidence architecture can support different reviewers without pretending they ask the same question.
Mobility analysis
Compare functional output, observed burden, movement, and body coupling within a bounded case.
Research review
Inspect source roles, repeatability, samples, missingness, and reproducibility.
Legal and access review
Keep measured observations distinct from allegations, documents, findings, and legal interpretation.
Wearable intelligence
Explore context-preserving classification, multimodal evidence, within-person baselines, and human review.
Methodology and interpretation
What the System Does Not Claim
A wearable estimate, route record, personal account, medical document, agency decision, and scientific interpretation remain different evidence classes. The Observatory connects them for review without making them equally authoritative.
Within-person patterns can support investigation and individualized review. They do not establish diagnosis, universal causation, universal safety, or legal entitlement.
Canonical declaration: Within-person evidence model. Wearable signals estimate physiologic, musculoskeletal, movement, and transport burden; source artifacts and validation layers remain available for audit.