The Credible Safety
Evidence Framework™
Safety claims are not trusted because evidence exists. They are trusted because the evidence possesses characteristics that allow others to evaluate it independently. This framework identifies the four properties that separate credible safety evidence from evidence that merely exists.
The question this framework answers
Frameworks #1 through #3 addressed operational questions: what evidence matters, how it is generated from events, and how it demonstrates improvement. Framework #4 addresses the question that sits above all of them.
What characteristics must safety evidence possess before different stakeholders consider it credible?
This is not a question about whether safety improved. It is a question about whether the evidence supporting that claim is evaluable by someone who was not present when it was produced. That distinction matters enormously in practice.
Stakeholders do not trust safety claims because evidence exists. They trust safety claims because the evidence is traceable, verifiable, repeatable, and valid within a clearly defined operational context.
Two operators, one claim
Consider two autonomous vehicle operators making the same statement: we have reduced the likelihood of this incident recurring.
Operator A provides a chart showing fewer incidents this quarter.
Operator B provides the incident history, the scenario classification, the investigation findings, the corrective actions implemented, the validation results, the operational monitoring data after deployment, and documented evidence that recurrence has decreased within the affected scenario class.
Both organizations made the same claim. Only one made a credible claim. The difference is not confidence. The difference is not intent. The difference is the quality and structure of the evidence.
The four properties of credible safety evidence
These properties are not original to this framework. They reflect how mature safety-critical industries — aviation, medical devices, rail, nuclear — have consistently structured evidence for decades. What this framework does is synthesize them into a form directly applicable to autonomous vehicle operations.
Traceable
Can the safety claim be connected back to the operational event that produced it? Evidence should answer: which incident, which scenario, which investigation, which corrective action. Without traceability, improvement becomes a statement rather than a demonstrable outcome. Traceability is what allows an independent reviewer to follow the chain from claim to event without relying on the organization's word.
Verifiable
Can another competent party examine the evidence and reach the same conclusion? Verification may come from independent review, regulatory oversight, documented testing, or structured audit. Evidence that cannot be examined by someone outside the organization is not evidence in any meaningful sense. It is a claim dressed as evidence. Verifiability is what separates assurance from assertion.
Repeatable
Would the same process produce comparable evidence if applied again? Repeatability is not about reproducing the exact incident. It is about producing evidence through a consistent and disciplined process. This is why standardized investigation methods, documentation practices, and validation procedures matter — not as bureaucracy, but as the mechanism that makes individual evidence items part of a coherent evidentiary system.
Contextually Valid
Under what operating conditions does this evidence support the safety claim? No safety improvement applies universally. Every claim has boundaries: operating environment, weather, traffic density, ODD, software version, vehicle configuration. Context defines where the evidence remains meaningful and where it does not. Evidence without context is not wrong — it is incomplete, which in safety governance amounts to the same thing.
"None of these four properties measure safety. They measure the quality of the evidence supporting safety. That distinction is the most important thing this framework asserts."
ITquette Autonomous Lab · Framework #4Why organizations create trustworthiness, not trust
There is a subtle but important distinction that this framework insists on. Organizations do not create trust. They create trustworthiness. Trust is the decision another party makes after evaluating the evidence an organization has produced.
This means that no single organization — regardless of its size, its data, or its public communications — can generate trust unilaterally. Trust requires an evaluating party. And that evaluating party needs evidence that possesses the four properties above before it can make any rational trust decision.
Consider how aviation works. Nobody trusts aviation because Boeing says trust us. Trust in aviation exists because there is an ecosystem: the FAA, the NTSB, independent investigations, certification processes, mandatory reporting requirements, maintenance organizations, and manufacturers all operating within a shared evidentiary framework. No single organization creates that trust. The system creates it because the system produces evidence that is traceable, verifiable, repeatable, and contextually bounded.
Autonomous vehicles are at an earlier stage of that same journey. The regulatory infrastructure is forming. The evidentiary standards are being established. The organizations that build their operational evidence systems around these four properties now will be structurally better positioned when the ecosystem matures and the scrutiny intensifies.
How this framework connects to Itquette LLC's operational service
This framework is not only a research proposition. It is the conceptual foundation behind every deliverable Itquette LLC produces.
The incident intake record establishes traceability — connecting every subsequent piece of evidence back to a specific operational event with a specific vehicle, time, location, and ODD context. The structured investigation process produces evidence through a consistent method, supporting repeatability. The corrective action log and validation records provide verifiable documentation that an independent reviewer can examine. The monthly pattern report and regulatory notification packages bound all of it within clearly defined operational context.
Every service deliverable is designed to produce evidence that possesses all four properties — not because the deliverables are sophisticated in themselves, but because the framework they are built on demands it.
Where Framework #4 sits in the series
Frameworks #1 through #3 built the operational foundation: identifying what evidence matters, generating it from events, and using it to demonstrate improvement. Framework #4 sits above that foundation and asks whether the evidence being produced is the kind of evidence that others can actually evaluate.
The question Framework #5 will address follows naturally: once an organization has credible safety evidence, how does it transform that evidence into an operational safety assurance capability that regulators, insurers, city partners, and the public can rely on?
That is the final question in the series. And it is the question the industry most needs answered.
Related frameworks
The Safety Evidence Gap Framework™
Why safety improvement and safety acceptance are not always the same thing — and what sits between them.
The Scenario Intelligence Framework™
How operational events become reusable organizational learning rather than isolated records.
The Incident-to-Improvement Framework™
How organizations demonstrate that corrective actions produced measurable risk reduction.
Working through a safety evidence or documentation challenge?
ITquette Autonomous Lab is the research arm of Itquette LLC — a Phoenix-based AV safety operations company. If this framework connects with something your organization is actively working through, we welcome that conversation.