Decision Quality in Safety: 5 Distortions That Hide Risk

A safety decision can be approved, documented, and still leave the exposure untouched. That is the uncomfortable test for leaders who want reliable protection rather than a tidy governance record. The question is not whether someone signed the form. The question is whether the decision changed the conditions that could produce serious harm.

HSE sets out leadership as a director-level responsibility, while ISO 45001:2018 specifies an occupational health and safety management system that connects policy, planning, operation, audit, and review. Neither source treats approval as the same thing as control. In practice, leaders need to examine the distortions that make a decision look complete before the risk is actually reduced.

Why decision quality matters more than decision volume

Decision quality in safety means that the right authority makes a decision using relevant evidence, defines the safe state, and verifies whether the intended control works in the field. A high volume of meetings can hide weak decision quality when the same exposure returns without a changed owner, barrier, or operating limit.

Andreza Araujo's work on safety culture places accountability close to daily operating choices, not only in policy statements. Her book Safety Culture: From Theory to Practice is useful here because it treats culture as something leaders make visible through repeated choices. A board that reviews 12 indicators but cannot identify who may stop a hazardous task still has a decision problem, not a reporting problem.

The five distortions below are not personality defects. They are recurring ways that organizations convert uncertainty into administrative comfort. Each one creates a different false signal, so the corrective action must match the distortion.

1. Approval is mistaken for risk reduction

Approval reduces risk only when it changes an exposure, a control, or the authority to intervene. A signed risk assessment can authorize work while leaving the original hazard, weak barrier, and production pressure exactly where they were.

This distortion appears when leaders count completed reviews instead of asking what changed after the review. A maintenance task may have a permit, a supervisor signature, and a toolbox talk, although the isolation point remains difficult to verify and the rescue route has never been tested. The paperwork is complete; the protective system is not.

ISO 45001 links operational planning with control of processes and continual improvement. That logic supports a simple leadership test: identify the decision, name the changed condition, and show evidence that the new condition exists. If the answer stops at approval, the organization has recorded intention rather than reduced risk.

Use a decision record with three mandatory fields. State the serious exposure, identify the control that must hold, and define the evidence that will close the decision. The evidence might be a field verification, an isolation test, a design change, or a time-bound restriction. It should not be another signature.

2. Ownership is confused with proximity

The person closest to the hazard is not automatically the person with authority to control it. Proximity helps leaders see the work, but ownership requires the power, resources, and escalation route needed to change the work.

A shift supervisor may notice that a temporary access route creates a fall exposure, while the engineering manager controls the redesign and the plant manager controls the production interruption. Calling the supervisor the risk owner without granting those decision rights turns accountability into blame.

HSE's leadership guidance emphasizes that boards and senior leaders retain responsibility for directing health and safety, even when tasks are delegated. That distinction matters because delegation should transfer defined actions, not erase the higher-level duty to provide resources and oversight.

A practical ownership test has four parts. Ask who can authorize the task, who can stop it, who can change the control, and who must be informed when the deadline cannot be met. If four different answers emerge and no escalation path connects them, the risk is distributed without being owned.

The four responsibilities of a risk owner make this distinction operational. They connect acceptance with action, verification, and escalation rather than treating a name in a register as proof of control.

3. Closure is mistaken for verification

Closing an action proves that an activity was marked complete. Verification proves that the exposure changed and that the control remains reliable under the conditions in which people work.

The difference becomes visible after an incident review. A corrective action may say that training was delivered to 38 employees by March 15, 2026. That is a completion fact. It does not show whether the equipment was redesigned, whether supervisors can recognize the failure mode, or whether the control still works during a night shift with reduced staffing.

James Reason's work on organizational accidents helps explain why a local correction can coexist with latent conditions elsewhere. A visible action can be closed while the system that allowed the exposure to persist remains intact. Leaders therefore need evidence at the barrier, not only evidence in the action tracker.

Define verification before assigning the action. For a guarding problem, verification may require a physical inspection, a functional test, and a review after 30 days of operation. For a work-practice change, it may require observations across 3 shifts and a conversation with the people who perform the task. The method should test the control, not the existence of the record.

The checks that separate paper completion from risk reduction provide a useful companion framework for this review.

4. A dashboard is mistaken for a decision system

A dashboard supports a safety decision only when each signal has a threshold, an owner, a response time, and a defined escalation. Without those elements, the dashboard can describe exposure without changing what leaders do next.

Numbers create a strong impression of control because they compress complex work into a small space. A monthly report may show 0 lost-time injuries, 97 percent training completion, and 12 overdue actions, yet still fail to reveal that a critical barrier has been bypassed for 6 consecutive shifts. Precision in presentation does not guarantee precision in protection.

ISO 45001 requires organizations to monitor, measure, analyze, and evaluate relevant performance, but measurement is only useful when it informs action. A leader should be able to point from a red signal to the person who must decide, the condition that triggers intervention, and the evidence needed to release the restriction.

The thresholds that make critical-risk exposure look acceptable are especially important for executive reviews. They show why a dashboard can be accurate and still fail to protect people when the metric is not connected to a decision right.

5. Consistency is mistaken for control reliability

A control is reliable when it continues to protect people as work conditions change, not simply when the same instruction appears in every procedure. Consistency in documents can conceal variation in equipment, staffing, weather, workload, contractor capability, and recovery time.

Leaders often reward standardization because it makes comparison easier. The risk emerges when a standard method is treated as proof that every context has the same margin for error. A confined-space entry with a trained team, a tested rescue system, and stable ventilation is not equivalent to one with a delayed rescue vehicle and a temporary access arrangement, even when both use the same form.

The ILO's occupational safety and health management guidance connects management systems with prevention and improvement. That principle requires leaders to revisit assumptions when the work changes, rather than using prior approval as permanent permission.

Test reliability at the point where the operating context changes. Review the control after a 4-hour delay, a 2-person staffing reduction, a contractor handover, a temporary repair, or a change in sequence. These are not arbitrary stress tests. They are ways to discover whether the control depends on conditions that the organization has stopped protecting.

For a broader leadership routine, the weekly safety decision review helps compare the approved plan with field evidence before mismatch becomes normal.

How leaders can repair distorted safety decisions

Repair starts by separating five questions that are often compressed into one approval meeting. What is the serious exposure? Which control must hold? Who has authority to change it? What evidence will verify it? What happens when the safe state cannot be achieved?

HSE explains why effective health and safety performance comes from the top, which means senior leaders must examine their own decision patterns rather than asking only for more frontline compliance. A leader who repeatedly accepts open actions, vague owners, or untested controls is teaching the organization what approval really means.

Andreza Araujo's experience across more than 250 cultural transformation projects supports a practical conclusion: safety culture becomes credible when the response to risk is visible, consistent, and connected to operating decisions. The credential does not replace evidence. It reinforces the importance of building routines in which evidence can change the decision.

Start with one recurring review, one critical exposure, and one verification standard. Review the same case after 7 days, 30 days, and 90 days, because time reveals whether the control survived the original attention. If the exposure remains, escalate the decision instead of refreshing the language around it.

The leadership test is what changes after the meeting

A safety meeting has value only when it changes a condition that matters. Approval, ownership, closure, dashboards, and consistency are useful only when they lead to a verified safe state and a clear response when that state cannot be maintained.

Leaders should therefore leave each critical-risk review with five explicit outputs: the exposure, the required control, the accountable authority, the verification evidence, and the escalation deadline. If any output is missing, the decision is not ready to be treated as protection.

That discipline is more demanding than collecting signatures, but it is also more honest. A decision system earns trust when people can see what changed, who can intervene, and how the organization responds before a weak signal becomes a serious event.

To develop this discipline across your organization, explore Andreza Araujo's safety culture work and connect leadership routines with the decisions people face every day.