Why Good Pilots Make Bad Decisions

There is a moment in many accident investigations where the question becomes almost unbearable. The crew was experienced. The aircraft was serviceable. The weather was manageable. The decision — the one that killed everyone on board — was one the crew had made correctly hundreds of times before. And this time they didn’t.

The temptation is to find something exceptional about that crew, that day, that flight. But the disturbing lesson of half a century of human factors research is that there was nothing exceptional. The failure was ordinary. The conditions that produced it exist in almost every cockpit, on almost every flight. Most of the time, we get away with it.

That is not a reassuring finding. But it is an honest one.

The Paradox of Expertise

High-skill professionals are not more immune to human factors errors than novices. In some ways, they are more vulnerable. Expertise automates. The more practiced you are, the more your execution migrates from conscious control to pattern recognition. This is efficient — it frees cognitive bandwidth for other things. It is also dangerous, because pattern recognition is selective. Your brain stops asking “what is actually happening here?” and starts asking “does this match what I’ve seen before?”

When the answer is yes — even approximately yes — you proceed. The mismatch between the familiar pattern and the actual situation can go unnoticed until it is operationally significant. By then, you’re already committed.

This is why a captain with 15,000 hours can fly a perfectly trimmed aircraft into terrain. Not despite their experience, but partly because of what experience does to attention. The scan was there. The cross-checks were performed. But the underlying mental model — the assumed picture of where they were and what was happening — was wrong. And no amount of technical proficiency corrects a flawed mental model.

What Human Factors Actually Is

“Human factors” gets used loosely, often as a polite synonym for pilot error. It is not that. Human factors is the scientific study of how humans interact with systems, environments, and each other under conditions of stress, fatigue, time pressure, and ambiguity. It asks why capable people systematically fail in predictable ways — and what design and training interventions can reduce those failures.

The field has given aviation a vocabulary that barely existed forty years ago: situational awareness, crew resource management, threat and error management, the authority gradient. These are not soft concepts. They are the result of meticulous accident analysis, simulator research, and field observation. They represent aviation’s genuine attempt to reckon with the fact that the most advanced aircraft ever built are still being operated by human beings, and human beings are not reliable in the ways that machines are.

The crucial reframe is this: human factors errors are not failures of character. They are failures of systems — cognitive systems, crew systems, organizational systems — operating at or beyond their design limits. Treating them as character failures is not just unfair. It prevents the correct intervention.

The Authority Gradient Problem

The 1977 Tenerife disaster — still the deadliest accident in aviation history — should have been survivable. The KLM Boeing 747 began its takeoff roll in restricted visibility, on a runway that was not clear, without a confirmed ATC takeoff clearance. The first officer had concerns. He voiced them. He was not heard. The captain pushed through.

What happened in that cockpit was not unusual. It was the authority gradient in its most lethal expression: a steep power differential between captain and junior crew that made it functionally impossible for critical information to reach the decision-maker in time to matter.

CRM was aviation’s response. Starting in the early 1980s, airlines began training crews explicitly in communication, assertiveness, and the shared responsibility for flight safety. The program has saved lives. It has also, for decades, been applied in ways that miss the deeper problem.

You can teach a first officer the phrase “Captain, I am concerned that…” You can role-play the scenario in a simulator until it feels natural. And then you can put that first officer in an actual cockpit with an actual captain who carries twenty years of seniority and the full weight of institutional authority, and watch the assertiveness evaporate. Not because the training failed. Because the training addressed behavior while the culture remained unchanged.

The authority gradient problem is not solved by telling junior crew to speak up. It is solved — partially, imperfectly, over time — by changing what captains do when they receive input they don’t want. By making psychological safety a measurable crew quality rather than a training module. By normalizing the challenge-response loop on routine flights so that it is available on the critical ones.

This is harder work. It cannot be done in a classroom.

How Situational Awareness Degrades

Situational awareness does not announce its departure. This is what makes its loss so consistently dangerous.

Mica Endsley’s three-level model — perception, comprehension, projection — is now standard in aviation human factors training, and it is useful precisely because it breaks the problem into stages. Most SA failures begin at Level 1: a failure to perceive something that was available to be perceived. Not a failure of vision or hearing. A failure of attention allocation. The information was there. It simply wasn’t processed, because attention was somewhere else.

From there, the cascade is predictable. Flawed perception produces flawed comprehension — you interpret the situation incorrectly, because you’re working from incomplete data. And flawed comprehension produces flawed projection — you model where things are going based on where you think they are. By the time you reach a decision point, you may be three levels removed from reality, all while feeling entirely confident that you know what’s happening.

There are well-documented precursors to SA breakdown: channelized attention (fixation on a single task to the exclusion of the broader environment), workload saturation, interpersonal conflict within the crew, and fatigue. Each of these reduces the bandwidth available for environmental scanning. Each of them is routinely present on commercial flights.

What protects against it is habit. Regular, structured cross-checks. Verbalization of the shared mental model. Explicit callouts when something doesn’t fit. The crew that periodically asks “what’s our current picture?” — out loud, as a genuine question rather than a ritual — builds a form of collective SA that individual awareness cannot replicate.

Decision Fatigue at Altitude

The neuroscience here is not complicated, but the implications are underappreciated. Decision-making draws on a pool of cognitive resources that depletes with use. Early in a shift, decision quality is high. The prefrontal cortex — the part of the brain responsible for deliberative reasoning, risk assessment, and impulse control — is well-fuelled and operating normally. Later, as fatigue and glucose depletion accumulate, it is not.

What fills the gap is heuristic thinking: pattern matching, gut feeling, satisficing rather than optimizing. Heuristic thinking is fast and often good enough. It is also biased in systematic ways. It over-weights recent information, under-weights base rates, and is disproportionately influenced by how a question is framed. Under fatigue, you are more susceptible to confirmation bias, more likely to anchor on the first data point you receive, and significantly more likely to continue with an existing plan rather than revise it.

This matters for aviation in a very specific way. Long-haul crews are making their most consequential decisions — approaches, diversions, go-around calls — at the end of flights, when cognitive resources are most depleted. The regulatory answer is rest requirements and fatigue risk management. These help. But no rest requirement eliminates end-of-duty cognitive degradation. The more durable answer is pre-briefed decision criteria: commitments made early in the flight, when thinking is clearest, that constrain decisions made later, when it is not.

This is what the approach briefing is actually for. Not the recitation of speeds and frequencies. The pre-commitment to what will trigger a go-around, what weather minimum is a hard stop, what technical issue ends the flight. The captain who briefs this explicitly, and holds the crew to it, is doing something cognitively sophisticated: outsourcing a late-flight decision to an earlier, clearer version of themselves.

What CRM Gets Wrong About Human Nature

The core assumption of first-generation CRM training was essentially optimistic: give people the right communication tools, and they will use them. Teach the concepts, run the scenarios, issue the certificate. Behavior will change.

Thirty years of evidence suggests this is insufficient. Not wrong — communication training does produce real improvements in crew coordination. But it mistakes the symptom for the disease. The symptom is crews who don’t communicate well under pressure. The disease is organizational cultures that reward compliance over candor, that make assertiveness feel unsafe, that value schedule completion above all else.

CRM training teaches individuals. Culture shapes collectives. And individuals in a culture that punishes speaking up will not use their communication tools in the moments that count, regardless of what they learned in the simulator.

The research on psychological safety — Amy Edmondson’s work in healthcare, organizational studies in aviation and nuclear power — is consistent on this point: performance improves when team members believe they can raise concerns without punishment, and it degrades when they don’t. Psychological safety is not a feeling. It is a measurable property of teams, and it predicts error rates with uncomfortable reliability.

What CRM gets right is the vocabulary. What it often gets wrong is the mechanism for change. You cannot train your way out of a cultural problem.

What Actually Helps

The evidence points to three things that make a genuine difference.

First, individual psychological awareness. Knowing your own biases — particularly plan continuation bias and get-there-itis — doesn’t eliminate them, but it creates a small but real window of deliberate reflection. Pilots who can catch themselves in “I really don’t want to divert” and ask “is that a decision or a preference?” are operating at a different level than those who can’t name what’s happening.

Second, crew culture deliberately built around challenge and inquiry. Not assertiveness as a technique, but genuine curiosity as a crew norm. Captains who model uncertainty — who say “I’m not fully comfortable with this, what do you see?” — create the conditions for junior crew to contribute. This has to happen on routine flights for it to be available on critical ones. The crew that only speaks up in emergencies has never practiced speaking up.

Third, pre-briefed decision criteria treated as binding commitments. The go-around decision is not made on final approach; it is made during the approach briefing. The diversion trigger is not negotiated at minimum fuel; it is set before departure. This is not rigidity. It is the recognition that your future self, under pressure, is less reliable than your current self, with time and information to think clearly.

None of this is complicated in theory. All of it is difficult in practice. The human factors that kill people in cockpits are not exotic. They are the ordinary limitations of human cognition operating in conditions it was never designed for. Recognizing them — clearly, honestly, without the flattery of thinking you’re somehow exempt — is where the work begins.