Ten promises are kept.
Then one promise is broken.
Should trust fall by exactly one-eleventh?
No.
Trust does not count events as identical votes.
Trust violation weight is the amount a breach should change future reliance based on what the event reveals about competence, reliability, integrity, fairness, concealment, severity and the mechanism likely to produce future behaviour.
This is the third pillar beneath How Trust Works. The master owns betrayal and repair broadly. This page owns the update size: how much should one failure matter?
Quick Read
A trust breach should be weighted by diagnostic value, not counted mechanically. One accidental delay with immediate disclosure may say little about integrity and only modestly reduce trust in punctuality. One deliberate deception can reveal that the previous model of honesty was wrong, so it may deserve a much larger update even if many routine interactions went well before it. Severity matters, but so do intent, concealment, repetition, scope, surprise and whether the failure reveals a hidden process likely to recur. The opposite error is global collapse: treating one ordinary mistake as proof that nothing about the person or system is trustworthy. Strong trust updating asks what dimension failed, how surprising the event was under the old model, what future mechanism it exposes and which parts of the existing trust scope should actually change.
breach → identify dimension → compare with prior model → assess intent + concealment + severity + recurrence → estimate diagnostic value → shrink appropriate trust scope → raise verification where needed → collect new evidence
Routine Success and Diagnostic Success Are Different
Ten easy tasks go well.
One difficult task under pressure reveals a shortcut.
The difficult event can carry more information because it tests a condition the routine tasks never reached.
Trust evidence should therefore be weighted by:
- task similarity;
- pressure;
- stakes;
- opportunity for concealment;
- relevance to the trust dimension.
Accidental Error Is Not the Same as Deception
Scenario A:
A trusted student submits the wrong file, notices quickly, reports it and sends the correct file.
Scenario B:
A student knowingly submits copied work and claims it is original until evidence proves otherwise.
Both are failures.
The second carries much more information about integrity and concealment.
bad outcome severity and trust-diagnostic severity are not always the same thing.
Concealment Often Makes the Breach Heavier
An error occurs.
The actor reports it immediately.
Trust in competence may fall.
Trust in integrity may remain or even receive positive evidence from honest disclosure.
If the same error is hidden until discovery, two problems now exist:
- the original failure;
- the concealment mechanism.
The second can broaden the trust impact considerably.
Intent Changes Interpretation
Unintentional mistake:
may call for training or process repair.
Deliberate rule violation:
may indicate integrity or incentive problems.
But intent should not be guessed casually.
Separate what is known from what is inferred.
Repetition Changes Weight
One missed deadline.
Possibly noise.
Five missed deadlines under similar conditions.
Now reliability evidence is much stronger.
Repeated failure means the event is less plausibly an exception and more plausibly part of the generating mechanism.
Surprise Under the Old Model Matters
If someone has always been slightly late, another minor delay adds little new information.
If someone has a long record of transparent conduct and is discovered deliberately falsifying a critical record, the event is highly surprising under the previous integrity model.
Large surprise can justify a large trust update because the old model failed badly to predict the event.
One Breach Can Reveal Hidden Structure
A public process has worked smoothly for years.
Then one case reveals:
- there was no independent review;
- one person controlled approval and recordkeeping;
- exceptions were never logged;
- complaints had no return path.
The single visible failure can therefore reveal a systemic architecture problem that existed long before discovery.
Severity Still Matters
A forgotten low-stakes promise and a high-consequence safety breach should not receive identical updates.
Severity affects:
- how much reliance should shrink immediately;
- how much verification should increase;
- what repair is required;
- how slowly trust should rebuild.
How Risk Works remains the owner of consequence and exposure broadly.
Scope Determines How Far the Breach Travels
A student misses one homework deadline.
Trust in homework punctuality may fall.
Does trust in honesty also fall?
Not unless evidence connects the breach to honesty.
The first sibling, Trust Scope, owns the domain boundary.
Integrity Violations Often Travel Further Than Competence Errors
Competence error:
may be limited to one technical domain.
Integrity violation:
can affect several domains because honesty and rule adherence influence many forms of reliance.
But do not assume every mistake is an integrity failure.
The classification matters.
Fairness Violations Can Damage Institutional Trust Broadly
If comparable cases receive unexplained different treatment, people may update not only trust in one outcome but trust in the procedure itself.
That can affect future willingness to rely even in cases where personal outcomes are favourable.
OECD’s 2026 trust work continues to identify fairness, integrity, openness, reliability and responsiveness as key public-governance trust drivers.
Openness Changes Violation Weight
Organisation discovers an error and publishes:
- what happened;
- who was affected;
- what is still unknown;
- what is changing.
The original error still matters.
Transparent repair provides new evidence about responsiveness and integrity.
Opacity does the opposite.
A Breach Under Pressure Is Highly Informative
Integrity is easiest when interests align.
What happens when:
- money is at stake;
- reputation is threatened;
- nobody is watching;
- admitting the problem is costly?
Behaviour under conflicting incentives can reveal more about future reliability than routine low-pressure behaviour.
Do Not Let Outcome Bias Rewrite the Process
Bad process produces lucky good outcome.
Trust should not rise merely because damage was avoided.
Good process produces unlucky bad outcome.
Trust should not necessarily collapse if reasonable care and transparent repair were present.
Evaluate both outcome and mechanism.
Global Distrust Is Often an Overcorrection
One teacher makes one marking error.
Conclusion:
No teacher can be trusted.
The update is much broader than the evidence.
Trust should shrink in proportion to what the violation reveals.
Blind Forgiveness Is the Opposite Overcorrection
Serious deception occurs.
The explanation is accepted immediately because:
They have always been nice to me.
Affection is not enough to erase diagnostic evidence.
Trust can be compassionate without becoming evidentially blind.
Violation Weight Should Change Verification
Small reliability anomaly:
light temporary checking.
Serious integrity breach:
stronger independent verification and narrower scope.
The second sibling, Verification Budget, owns the checking response.
Responsibility Repair and Trust Updating Are Different Jobs
The person who failed may owe acknowledgement, restoration and mechanism repair.
Responsibility Repair owns that actor-side duty.
The receiver is not obligated to restore trust at the same speed.
Trust updating answers:
Given the new evidence, how much reliance is justified now?
Rebuilding Begins After Weighting
Before rebuilding trust, define what the breach changed.
- scope shrank where?
- verification rose where?
- which evidence remains valid?
- which claims must be retested?
The fourth sibling, Trust Rebuilding, owns how reliance can return from that new baseline.
A Practical Violation Weight Record
BREACH: Missed project handoff DIMENSION: Reliability, not proven integrity failure SEVERITY: Moderate; teammate lost 2 hours DISCLOSURE: Self-reported before discovery RECURRENCE: First event in 9 similar handoffs SCOPE UPDATE: Reduce trust in unsupervised high-pressure deadlines only VERIFICATION: Add midpoint checkpoint for next 3 projects REBUILDING: Return to lighter checks after 3 on-time high-pressure handoffs
A 30-Lens Trust Violation Audit
- Event: what happened?
- Dimension: competence, reliability, integrity, fairness?
- Severity: how large was consequence?
- Intent: accidental, reckless or deliberate?
- Evidence: what supports intent inference?
- Concealment: was failure hidden?
- Disclosure: self-reported or externally discovered?
- Recurrence: first or repeated?
- Pressure: what incentives were active?
- Surprise: how inconsistent with prior model?
- Routine history: what past successes remain relevant?
- Task novelty: was this outside proven scope?
- Scope: which domains are directly affected?
- Generalisation: does breach reveal deeper mechanism?
- Fairness: was procedure selectively applied?
- Openness: how transparent was response?
- Responsiveness: was harm contained quickly?
- Outcome bias: are we judging only the result?
- Process: was reasonable care present?
- Global distrust: is update too broad?
- Blind forgiveness: is update too small?
- Receiver: what vulnerability changed?
- Verification: what checking should rise?
- Responsibility: what does actor owe?
- Repair: did mechanism change?
- Rebuilding: what future evidence matters?
- Time: should trust recover slowly?
- Irreversibility: was damage permanent?
- Systemic signal: does one event expose architecture?
- World return: does the trust update match what the breach actually predicts about future reliance?
Laboratory 1: Same Outcome, Different Meaning
Compare three missed deadlines: accidental with immediate disclosure, negligent after repeated warnings, and deliberately concealed. Assign different trust updates and explain why.
Laboratory 2: Scope the Breach
Take one violation and list which trust dimensions it directly informs, which it only weakly suggests, and which remain untouched.
Laboratory 3: Surprise Test
Write the old trust model before revealing a hypothetical breach. Ask how probable that breach would have seemed. Larger surprise may justify examining whether the old model missed hidden structure.
For Primary Readers
If a friend forgets one pencil, that is different from a friend secretly taking your pencil and saying they did not. Both are problems, but they tell you different things about what to trust next.
For Secondary Readers
Compare accidental error, repeated unreliability and deliberate deception. Explain why trust should update by what the event reveals, not simply by counting good and bad outcomes.
For Advanced Readers
Model violation weight as diagnostic evidence about a latent trust-generating mechanism. Events deserve larger updates when they are highly improbable under the prior trust model, occur under revealing incentives, involve concealment or imply a broad integrity/process failure; ordinary low-information errors should produce narrower updates.
Common Misconceptions
- “Ten good actions cancel one bad action mathematically.” events differ in diagnostic value.
- “Any betrayal proves nothing can be trusted.” scope the update to what the evidence reveals.
- “Only outcome severity matters.” intent, concealment and process can carry major trust information.
- “An honest mistake should not affect trust at all.” competence or reliability evidence may still require updating.
- “Apology should restore trust immediately.” repair and future reliance are related but separate decisions.
Research Corridor
- OECD Survey on Drivers of Trust in Public Institutions 2026 Results — current evidence on public trust drivers including reliability, responsiveness, integrity, openness and fairness.
- OECD — Trust in Government.
- eduKateSG — How Trust Works.
- eduKateSG — Responsibility Repair.
Frequently Asked Questions
Why can one trust violation matter so much?
Because one event can reveal hidden information about integrity, concealment or process structure that many routine successes never tested.
Should one mistake destroy trust?
Not automatically. Determine what dimension failed, whether the event was accidental or repeated, and which future reliance claims the evidence actually affects.
Why does deception damage trust more than some errors?
Because deception can reveal an integrity mechanism that changes how future reports and promises should be interpreted across several situations, not just the original task.
Final Thought: Trust Learns More From Some Events Than Others
A trust violation should matter in proportion to what it teaches about the future—not according to a simple tally of how many good and bad moments came before it.
TRUST · FOUR PILLAR LEGS
Return to How Trust Works, or continue through Trust Scope, Verification Budget and Trust Rebuilding. Return to the How X Works Hub.