HTML Test

Instructions: Use the PDF toolbar to navigate, zoom, print, or download the document. Click the Pages/Thumbnail icon at the upper-left of the PDF toolbar to open or close the page thumbnail panel for quick navigation.

Full Text / Article Transcript

The Erosion of the Human Safety Margin

Albert N. Clark
Independent Author
Published: August 31, 2026
ASX Research Journal and Database
ISSN 3068-3351 (Online)
Place of Publication: Cadiz City, Philippines
Publisher: ASXResearch.org

Author Note

Albert N. Clark
Department of Aerospace Sciences, ASXResearch.org
ORCID iD: https://orcid.org/0009-0002-7348-4395
The author reports no conflicts of interest.
Correspondence concerning this article should be addressed to Albert N. Clark, Email: [email protected]

Abstract

This article examines whether aviation’s human safety margin is being systematically eroded by experience loss, staffing shortages, compressed training pipelines, automation dependence, fatigue, weakened supervision, and the normalization of degraded operating conditions. Rather than arguing that modern aviation professionals are inherently less capable than previous generations, the analysis focuses on the institutional and operational conditions that shape proficiency among pilots, air traffic controllers, maintenance technicians, military aviators, and other frontline personnel. Historical development, contemporary accident and incident evidence, workforce pressures, training trends, military safety experience, regulatory oversight, and organizational culture are evaluated alongside emerging responses from manufacturers, government agencies, and advanced technology programs. The article argues that aviation faces a dangerous transition in which qualification may increasingly be mistaken for experience and automation may be used to compensate for weakened human resilience instead of reinforcing it. The central conclusion is that artificial intelligence and advanced automation should strengthen an already competent human system rather than become substitutes for declining training depth, mentorship, judgment, and professional standards.

Keywords: aviation safety, human performance, professional standards

The Erosion of the Human Safety Margin

Commercial aviation has been quietly moving toward the single-pilot question for decades, long before artificial intelligence made the idea sound technologically fashionable.

The proposition that aviation’s human element is deteriorating deserves to be treated neither as nostalgia nor as a foregone conclusion. The evidence does not support the crude assertion that today’s pilots, controllers, mechanics, military aviators, dispatchers, firefighters, and ground personnel are simply less intelligent or less conscientious than their predecessors. It does, however, support a more disturbing diagnosis: in important portions of the aviation system, experienced people are retiring faster than institutional knowledge can be replaced; increasingly junior workforces are being asked to absorb sophisticated responsibilities quickly; staffing shortages and production pressures compress training and supervision; automation reduces opportunities to practice perishable skills; fatigue and workload remain stubbornly resistant to administrative solutions; and organizations can gradually become accustomed to operating closer to the edge because yesterday’s shortcut did not produce an accident. Sedlar et al. (2023) describe normalization of deviance as the gradual acceptance of departures from established standards after repeated exposure to those departures without an adverse consequence. That mechanism is a far more useful explanation than generational contempt. The erosion of the human safety margin is therefore not necessarily a decline in humanity’s biological capability; it is frequently the foreseeable product of deliberate decisions about staffing, scheduling, training, experience, cost, automation, organizational tolerance, and how much resilience may be removed before anybody notices. Aviation is extraordinarily safe precisely because generations before us built layers of human redundancy into it. The danger begins when those layers are quietly treated as excess capacity rather than safety infrastructure.

Figure 1
The Erosion of the Human Safety Margin: Pressures, Warning Signs, and the Path Toward Recovery in Modern Aviation
Note. Click image for full-size image or click here.

Historically, aviation learned professional competence through blood. Early pilots possessed enormous manual skill but operated in a system with primitive navigation, weak weather forecasting, little standardization, minimal human-factors knowledge, and almost no institutional protection against poor judgment. As technology improved, aviation progressively professionalized the human side as well: instrument standards, recurrent checking, type ratings, maintenance certification, formal air traffic control, crew resource management, military standardization, dispatch systems, accident investigation, and eventually Safety Management Systems. That progression matters because modern safety was never produced by better airplanes alone; it resulted from increasingly disciplined people operating inside increasingly disciplined organizations. Kelly and Efthymiou (2019), examining 50 controlled-flight-into-terrain accidents across commercial, military, and general aviation, found recurring decision errors, skill-based errors, communication deficiencies, planning problems, distraction, complacency, and fatigue despite decades of technological advancement. The implication is uncomfortable: technology can eliminate old failure modes while leaving human vulnerability perfectly capable of creating new ones. The historic answer was recurrent proficiency, supervision, standardization, and layered defense. When those defenses are weakened, the system can regress without the aircraft themselves becoming any less sophisticated. This is especially important in military aviation, where Li and Harris (2013) analyzed 523 military aircraft accidents and identified judgment and decision-making deficiencies along with systemic training weaknesses. Their findings reject the convenient fiction that an accident attributed to “pilot error” begins and ends with the person holding the controls. A deficient pilot can be an individual problem; a pattern of deficient pilots is an organizational product.

Commercial flight operations illustrate the paradox particularly well. Airlines increasingly recruit through heterogeneous pipelines—traditional civilian time-building, sponsored cadet programs, military transition, university programs, and accelerated pathways—while simultaneously operating aircraft whose automation makes routine line flying extraordinarily stable. Chan et al. (2025) found that pilots’ initial training backgrounds produce persistent differences in how they perceive and attribute accident causal factors, including how readily they identify latent organizational conditions. That does not make one pipeline inherently unsafe, but it does demonstrate that equivalent licenses do not create identical professional perspectives. Meanwhile, proficiency itself is perishable. Ebbatson et al. (2010) found significant relationships between recent manual-flying experience and transport-pilot manual-control performance, confirming that automation can preserve operational efficiency while reducing opportunities to exercise skills that suddenly become crucial when automation degrades. This creates the possibility of a highly credentialed pilot who is superbly competent at managing the normal automated system but less practiced at recovering when that system ceases behaving normally. The appropriate response is not romantic advocacy for hand-flying every transport leg; it is evidence-based recurrent training deliberately designed around improbable, compound, automation-degraded situations. The workforce pressure intensifies the problem. Boeing’s 2026 Pilot and Technician Outlook projects a global requirement for roughly 674,000 new commercial pilots and 728,000 new maintenance technicians over the next twenty years. Boeing’s 2025 outlook explicitly described a period of workforce “juniority” and stressed competency-based training, digital technologies, mixed reality, artificial intelligence, and machine learning as methods for increasing training effectiveness. When an industry must replace that many experienced professionals while simultaneously expanding, experience compression ceases to be an anecdote and becomes a strategic safety problem.

Air traffic control may be the clearest demonstration that competence cannot be separated from the environment in which competent people are required to perform. Zamarreño Suárez et al. (2024), reviewing 374 studies of controller workload, concluded that workload management, assessment, prediction, complexity, and human-centered operational conditions remain fundamental to air-traffic safety. Li et al. (2023) demonstrated that rotating rosters produce accumulated controller fatigue capable of degrading cognitive performance and showed that scientifically designed fatigue interventions can improve resilience. These are not accusations that controllers are becoming careless; they demonstrate that highly trained professionals remain biological organisms whose performance can be degraded by staffing patterns and scheduling. The FAA itself acknowledges a longstanding staffing problem. Its 2026–2028 Controller Workforce Plan reports that 2,028 trainees were hired during FY2025 while 1,460 people were lost through Academy attrition, training failures, retirement, resignation, promotion, and other separation, yielding a net workforce gain of 568; the agency plans progressively larger hiring classes through FY2028. That hiring surge is necessary, but hiring is not synonymous with certification, and certification is not synonymous with seasoned judgment. Every accelerated pipeline creates an unavoidable interval during which veterans are disproportionately responsible for training newcomers while simultaneously operating the system. The real safety question therefore is not whether standards have formally been lowered; it is whether the operational environment provides enough time, mentorship, staffing, recovery, and repetition for new controllers to attain the practical depth that experienced controllers once accumulated under different workforce conditions.

Aircraft maintenance presents the same phenomenon in a less visible form. Maintenance failures rarely appear before the public as dramatic evidence of a deteriorating workforce; they are buried in inspection findings, repeat discrepancies, paperwork, deferred defects, installation errors, communication failures, and organizational pressures until one eventually survives every barrier. Aktas and Kagnicioglu (2023) found that maintenance technicians’ safety behavior is strongly influenced by safety leadership and safety climate, and that technicians can be less willing to report unsafe conditions involving their own teams. That finding is important because competence without reporting courage is an incomplete safety defense. Tyagi et al. (2023) similarly found that learning from past maintenance events depends critically on safety communication and on organizations contextualizing and evaluating lessons rather than merely possessing accident information. In other words, a maintenance organization can technically “know” about a previous failure and still fail to learn from it. The coming manpower problem makes that vulnerability harder to dismiss. Boeing’s forecast of hundreds of thousands of new technicians implies a huge transfer of tacit knowledge from retiring mechanics and inspectors to a younger workforce just as fleets, avionics, composites, software, engines, and maintenance analytics become more complex. The danger is not young mechanics; it is an industrial system tempted to treat certification as a substitute for mentorship and throughput as a substitute for craftsmanship. A mechanic who knows which manual paragraph applies is necessary. A mechanic who has seen a subtle defect before, recognizes when the paperwork does not make sense, stops a job under schedule pressure, and can explain why something merely “looks wrong” represents a different level of safety capability. That knowledge is expensive to build and very easy to lose.

Military aviation deserves special scrutiny because the hypothesis of universal decline encounters both supporting and contradictory evidence there. Li and Harris’s earlier findings demonstrate that training deficiencies and poor tactical decision-making can become systemic military accident factors, but current U.S. Army data show why careless generalization would be scientifically dishonest. Army Aviation reported five Class A flight mishaps in FY2025 compared with fifteen in FY2024, producing a manned Class A flight mishap rate of 0.66 per 100,000 flying hours—described by the Army as the third-best rate in its recorded history. More revealingly, the Army specifically attributed part of the improvement to closing a training gap involving unanticipated right yaw and loss-of-tail-rotor-effectiveness events that had emerged in FY2023 and FY2024. That sequence is almost a laboratory demonstration of the argument advanced in this paper: proficiency can erode, the erosion can manifest operationally, and a competent institution can identify the deficiency, restore training, and improve the outcome. Military aviation therefore should not be portrayed simply as another decaying institution. It is better understood as an environment where operational tempo, personnel turnover, complex aircraft, combat priorities, and training-resource constraints can expose proficiency gaps quickly—and where aggressive standardization can sometimes repair them just as quickly. The lesson for civil aviation is profound. Human standards do not inevitably decline; they decline when organizations fail to measure them, fail to recognize weak signals, or tolerate degraded performance because the mission is still being completed. The most dangerous sentence in aviation remains some variation of, “We have been doing it this way and nothing has happened.”

The Potomac River collision of January 29, 2025, is therefore valuable precisely because the final investigation does not support the simplistic story that one bad controller or one incompetent pilot killed 67 people. The NTSB concluded in January 2026 that systemic failures involving FAA helicopter-route design, inadequate review of available safety data, failure to act on previous recommendations, overreliance on visual separation, and deficiencies in Army safety monitoring contributed to the collision. The NTSB also determined that the DCA tower was below target staffing but had sufficient personnel available that evening to staff the helicopter and local-control positions separately; the decision to combine positions was not caused by insufficient staffing, and controller fatigue was not identified as causal. That distinction is essential to an article about declining standards: evidence must be allowed to destroy a convenient accusation. The March 22, 2026, LaGuardia collision between Air Canada Express flight 8646 and an airport rescue vehicle is equally sobering but must be treated differently because the investigation remains ongoing. Two pilots died and dozens were injured, but no responsible researcher can presently convert that tragedy into proof of controller, ARFF, airport-management, or flight-crew incompetence. Muecklich et al. (2023), however, show why the event belongs within the broader inquiry: their examination of 87 ground-operation accidents and incidents found situational-awareness deficiencies and failures to follow procedures among prominent human-factor themes. The proper use of LaGuardia is therefore not to prejudge blame; it is to ask why modern airport surface operations remain capable of producing catastrophic conflicts despite surveillance, radios, procedures, training, lighting, and decades of runway-safety work.

The deeper problem lies in institutional culture, because individual proficiency exists inside organizations that teach people what deviations will actually be tolerated. Teperi et al. (2023) found that mature human-factors practice becomes most effective when it evolves from isolated compliance activities into an organization-wide mindset embedded in management, supervision, work planning, and daily operations. Pratama and Caponecchia (2025), reviewing aviation safety across national cultures, likewise found that communication, teamwork, decision-making, power distance, individualism, and uncertainty avoidance can shape safety performance in ways that cannot be captured by technical qualification alone. Taken together, these findings expose the weakness of reducing professional standards to certificates and checkrides. A pilot may satisfy a recurrent check yet work in a company where schedule pressure subtly punishes conservative decisions. A controller may be fully certified yet operate in a facility where routinely combined positions have become normal. A mechanic may hold every required authorization yet learn that stopping a departure over an ambiguous discrepancy attracts more management scrutiny than signing it off. A military crew may meet published proficiency requirements while losing exposure to demanding scenarios because flying hours are scarce. In each case the formal standard remains intact while the practical standard quietly moves. That is how a high-reliability organization can become a high-risk organization without ever announcing the transition. The corrosion is incremental, administratively defensible, and usually invisible until an accident investigation reconstructs years of decisions that each looked individually reasonable.

Manufacturers are not ignoring this problem, although their response is understandably framed as training modernization rather than remediation of declining humanity. Airbus has incorporated competency-based training and assessment into ab initio programs and Evidence-Based Training into recurrent instruction, emphasizing technical and nontechnical competencies rather than mere completion of prescribed maneuvers. Boeing likewise identifies competency-based training as the foundation of its workforce-training strategy and expects artificial intelligence, machine learning, mixed reality, and digital learning systems to play increasingly important roles as the industry absorbs enormous numbers of new personnel. These developments are significant because they represent a shift from asking whether a trainee completed a syllabus to asking whether the trainee demonstrated durable competence. Chan et al. (2025) make that distinction particularly relevant by demonstrating that training histories can produce persistent differences in safety perception even after pilots enter common operational environments. Ebbatson et al. (2010) provide the complementary warning that competence itself can decay when it is not exercised. The future training system therefore cannot be merely faster because the workforce shortage is large; it must become more diagnostic. Simulator telemetry, eye tracking, adaptive scenarios, AI-generated variability, individualized remediation, maintenance augmented reality, and competency analytics could allow instructors to identify weaknesses that a fixed sequence of exercises misses. Technology can help replace scarce instructional capacity, but it cannot be allowed to become a mechanism for manufacturing credentials faster. If the industry uses AI to accelerate people through training rather than to deepen their competence, it will have automated precisely the wrong part of the problem.

The military and DARPA are already exploring the next stage, although not because either organization publicly describes its mission as compensating for declining pilot quality. DARPA’s Air Combat Evolution program used the X-62A VISTA to develop trusted human–machine collaboration in combat aviation, explicitly studying how pilots calibrate trust in autonomous systems while authority shifts between human and machine. The project is crucial to this discussion because it suggests where the erosion of the human safety margin may eventually lead: not to removal of professional standards, but to machine systems that monitor, cross-check, advise, and assume selected functions when human workload or capability becomes the limiting factor. DARPA’s work is therefore adjacent to the problem rather than evidence that the Pentagon has concluded its aviators are becoming inferior. Likewise, there is no documented public program as of August 31, 2026, showing Elon Musk, Tesla, xAI, or SpaceX directly addressing pilot, controller, mechanic, or aviation-workforce proficiency. Musk’s companies are technologically relevant to autonomy and AI, but attaching his name to this human-performance problem without evidence would make the paper more fashionable and less accurate. The documented actors are less glamorous and more important: FAA workforce planners, NTSB investigators, military safety organizations, aircraft manufacturers, airlines, universities, training providers, unions, airport operators, and human-factors researchers. The approaching question is not whether AI will “save” bad humans. It is whether aviation will use increasingly powerful machine intelligence to preserve human excellence—or use it as an excuse to tolerate less of it.

Federal, state, county, and local governments occupy different but overlapping parts of this equation. The FAA establishes and enforces the principal federal certification framework for pilots, mechanics, controllers, carriers, and certificated airports, while the NTSB investigates accidents and issues recommendations but does not regulate. State, county, municipal, and public-authority airport owners nevertheless control important parts of the operational environment: airport staffing, rescue and firefighting, snow and ice operations, wildlife mitigation, vehicle control, construction, surface inspections, emergency planning, and increasingly airport SMS. Under 14 C.F.R. Part 139, airport operators—including cities and counties—must satisfy federal requirements governing runway safety, markings, lighting, ARFF, fueling, emergency response, and other functions; the FAA notes that roughly 520 U.S. airports currently operate under Part 139 certification. LaGuardia illustrates the shared-governance problem neatly: the airport is owned by New York City and operated by the Port Authority of New York and New Jersey, while federal authorities regulate significant aspects of aviation safety and the NTSB investigates the 2026 runway collision. The legal consequences of declining standards can therefore spread rapidly. Airlines and maintenance organizations face negligent hiring, training, supervision, retention, and operational-control allegations; manufacturers may face design or failure-to-warn claims; airport operators may face liability arising from surface operations and emergency response; governments may encounter statutory and sovereign-immunity questions; and individual professionals can face certificate action or criminal exposure in exceptional cases. Civil litigation also has a powerful discovery function: training records, staffing rosters, fatigue data, safety reports, internal warnings, simulator performance, maintenance communications, and rejected recommendations can transform what management once described as an unforeseeable accident into documented organizational knowledge.

Where this is going depends upon whether aviation admits what the evidence actually says. There is no credible basis for declaring that an entire generation of aviation professionals is incompetent, and the Army’s recent safety improvement is an important counterexample to any such lazy conclusion. There is, however, substantial evidence that aviation is entering a dangerous transition in which extraordinary numbers of experienced people must be replaced while operational complexity, automation, traffic density, maintenance sophistication, military mission demands, and organizational interdependence continue to increase. The human safety margin can disappear even while every person involved technically meets the minimum standard. It disappears when “qualified” quietly substitutes for experienced, when recurrent training becomes an administrative event, when automation starves professionals of meaningful practice, when supervisors are too thinly spread to mentor newcomers, when fatigue is scheduled instead of managed, when deviations become normal because they worked yesterday, and when technology is expected to compensate for institutional decisions that technology did not cause. Aviation’s response must therefore be almost ruthlessly conservative about competence: more meaningful training rather than merely more training throughput; deliberate preservation of manual and cognitive proficiency; protected mentorship; data-driven identification of weak skills; stronger maintenance reporting cultures; scientifically managed controller fatigue; recurrent exposure to compound failures; aggressive military lessons-learned systems; and technological assistance designed to reinforce rather than replace professional mastery. The tragedy would not be that artificial intelligence becomes better than humans at parts of aviation. The tragedy would be that humans deliberately allow their own safety margin to erode because the machines appear capable of catching them. If AI eventually becomes aviation’s knight in shining armor, it should be because it adds another layer to an already excellent human system—not because we dismantled the human system and then begged the machine to save what remained.

References

Aktas, E., & Kagnicioglu, C. H. (2023). Factors affecting safety behaviors of aircraft maintenance technicians: A study on Civil Aviation Industry in Turkey. Safety Science, 164, 106146. https://doi.org/10.1016/j.ssci.2023.106146

Chan, W. T.-K., Li, W.-C., & Braithwaite, G. (2025). Pilots’ training backgrounds affecting the attribution of event causal factors and airline safety management. Journal of Air Transport Management, 125, 102786. https://doi.org/10.1016/j.jairtraman.2025.102786

Ebbatson, M., Harris, D., Huddlestone, J., & Sears, R. (2010). The relationship between manual handling performance and recent flying experience in air transport pilots. Ergonomics, 53(2), 268–277. https://doi.org/10.1080/00140130903342349

Kelly, D., & Efthymiou, M. (2019). An analysis of human factors in fifty controlled flight into terrain aviation accidents from 2007 to 2017. Journal of Safety Research, 69, 155–165. https://doi.org/10.1016/j.jsr.2019.03.009

Li, W.-C., & Harris, D. (2013). Identifying training deficiencies in military pilots by applying the Human Factors Analysis and Classification System. International Journal of Occupational Safety and Ergonomics, 19(1), 3–18. https://doi.org/10.1080/10803548.2013.11076962

Li, W.-C., Zhang, J., & Kearney, P. (2023). Psychophysiological coherence training to moderate air traffic controllers’ fatigue on rotating roster. Risk Analysis, 43(2), 391–404. https://doi.org/10.1111/risa.13899

Muecklich, N., Sikora, I., Paraskevas, A., & Padhra, A. (2023). The role of human factors in aviation ground operation-related accidents/incidents: A human error analysis approach. Transportation Engineering, 13, 100184. https://doi.org/10.1016/j.treng.2023.100184

Pratama, G. B., & Caponecchia, C. (2025). Examining the influence of national culture on aviation safety: A systematic review. Journal of Safety Research, 92, 317–330. https://doi.org/10.1016/j.jsr.2024.11.020

Sedlar, N., Irwin, A., Martin, D., & Roberts, R. (2023). A qualitative systematic review on the application of the normalization of deviance phenomenon within high-risk industries. Journal of Safety Research, 84, 290–305. https://doi.org/10.1016/j.jsr.2022.11.005

Teperi, A.-M., Paajanen, T., Asikainen, I., & Lantto, E. (2023). From must to mindset: Outcomes of human factor practices in aviation and railway companies. Safety Science, 158, 105968. https://doi.org/10.1016/j.ssci.2022.105968

Tyagi, A., Tripathi, R., & Bouarfa, S. (2023). Learning from past in the aircraft maintenance industry: An empirical evaluation in the safety management framework. Heliyon, 9(11), e21620. https://doi.org/10.1016/j.heliyon.2023.e21620

Zamarreño Suárez, M., Arnaldo Valdés, R. M., Pérez Moreno, F., Delgado-Aguilera Jurado, R., López de Frutos, P. M., & Gómez Comendador, V. F. (2024). Understanding the research on air traffic controller workload and its implications for safety: A science mapping-based analysis. Safety Science, 176, 106545. https://doi.org/10.1016/j.ssci.2024.106545

This document is preserved as part of the ASX Academic Paper Repository, an open-access collection of academic research in aviation, transportation, safety, management, logistics, and related disciplines.

Please Support Open-Access Aviation Safety Research

Your support helps us do everything we can to educate the people flying, and save their lives.

Donation Options