FOUNDATION / RESEARCH PROGRAM

Trust should be inspectable, not imposed.

TrustXE is an open scientific and engineering program for representing evidence, provenance, uncertainty, contradiction and trust reasoning without creating a centralized authority of truth.

Explore the trust model Review what is proven

Core thesis No universal truth score.

Expose dimensions, evidence and uncertainty instead of hiding reality behind a single opaque number.

Scientific posture Hypotheses must be falsifiable.

TrustXE separates philosophy, specification, experiment and validated evidence.

Human boundary Contestability is architecture.

Consequential assessments should be explainable, challengeable and historically traceable.

AN OPEN INITIATIVE FOR HUMANITY

For a more trustworthy and better world.

TrustXE starts from a human question before it becomes a technical one: how can we preserve the freedom to think, disagree and discover truth while making manipulation, forgery and artificial certainty easier to inspect?

INITIATIVE

by Igor Schiavinatto Pires

This project is an invitation to scientists, engineers, journalists, institutions, builders and citizens to create a common language for evidence and uncertainty. Not to decide truth for humanity, but to make the path toward justified trust visible, inspectable and contestable.

We do not seek an algorithm that tells humanity what to believe.
We seek infrastructure that helps humanity understand why something deserves belief,
what remains uncertain, and how that conclusion can be challenged.

OUR VISION

A shared epistemic layer for the digital world.

We envision information that can travel with evidence, provenance, transformations, uncertainty and assessment history — readable by people and machines without requiring a single central authority.

WHY OPEN

Trust cannot be credible if its foundations are hidden.

The concepts, models, benchmarks, failures and governance of TrustXE should be open to inspection. A global trust substrate must be challengeable by the same world it intends to serve.

THE TRUSTXE MANIFESTO

We are building the ability to question with better evidence.

We believe the next information era will not be solved by creating another authority that declares what is true. It will require systems that preserve origin, expose evidence, quantify uncertainty, reveal dependency between sources and keep disagreement visible.

We believe authenticity is not truth, popularity is not corroboration, confidence is not certainty, and reputation is not evidence. These concepts must remain separate if machines are going to participate responsibly in human knowledge.

We seek an interoperable and scientifically testable trust layer that any responsible application can use — from journalism and research to AI agents, public institutions and platforms such as RiscaFake.

And we accept that TrustXE itself must always be open to revision. A project about trust earns legitimacy only by exposing its own assumptions, limitations, conflicts and mistakes.

01 / PHILOSOPHY

Evidence before authority.

TrustXE begins with a simple premise: trust is not a permanent property of a person, institution or statement. It is a contextual assessment supported by evidence, provenance and explicit uncertainty.

ANTI-GOAL

Trust infrastructure must never become surveillance infrastructure.

No social credit. No hidden behavioral scoring. No universal personal reputation label. No political or ideological ranking. No black-box punitive decision presented as truth.

02 / CORE MODEL

A claim travels with its epistemic state.

The TrustXE substrate keeps raw evidence, lineage, inference and application policy separate. That separation is what makes an assessment auditable.

Artifact

→

Claim

→

Evidence Graph

→

Assessment

→

TrustState

TrustState(c, x, t)experimental

Evidence Graph

Claim

Evidence A

Evidence B

Source

Artifact

Assessment

Relationships such as supports, contradicts, derived_from, attests and supersedes remain explicit.

03 / SCIENCE

Ideas become research only when they can fail.

TrustXE treats its mechanisms as hypotheses. Results must expose assumptions, datasets, metrics, limitations and replication status.

H-001

Evidence Independence

One observation copied 100 times should not behave like 101 independent observations.

Naive

101

TrustXE

≈1

H-002

Contradiction Reasoning

Conflict should be classified before it is resolved: logical, temporal, definitional, methodological, contextual or provenance-related.

logicaltemporalcontextualprovenance

H-003

Temporal Validity

Old evidence is not automatically false. Validity must depend on the claim, domain and temporal context.

EXECUTABLE SCIENCE

Research Engine v0

The repository already contains a deterministic Python research engine, synthetic benchmark generation, reproducibility manifests, unit tests and CI.

$ trustxe-research b001 --derivatives 100
visibleEvidence:      101
naiveNEff:            101.0
observableNEff:       dependency-aware
rootIdUsed:           false
reproducibility:      manifest + graph hash

04 / ARCHITECTURE

Protocol beneath product.

TrustXE is designed to remain a reusable substrate. RiscaFake is one potential application and validation environment, not a dependency of the core.

ApplicationsRiscaFake · AI agents · media · institutions

Policy / Decisioncontext-specific thresholds and actions

Trust Reasoningdependency · contradiction · temporal validity · calibration

Evidence Graphclaims · evidence · actors · assessments · disputes

Provenance & Integritylineage · hashes · attestations · transformations

AcquisitionAPIs · archives · sensors · human input · external systems

TrustXE owns

semantics, evidence/provenance models, trust dimensions, research methods, interoperability and conformance.

Applications own

operational thresholds, business policy, analyst workflow, user experience and consequential decisions.

05 / VALIDATION BOARD

What are we actually claiming?

This board exists to prevent philosophy, hypothesis, specification and evidence from being confused with one another.

NEXT FRONTIER

Latent evidence families and uncertainty intervals

The next scientific step is to infer probable evidence families from observable signals and represent effective evidence as a distribution rather than a falsely precise point estimate.

P(root family | observations) → distribution(N_eff)

06 / SOURCE OF TRUTH

Read the work behind the interface.

The website is an interpretive view. The repository remains the auditable source of truth.