The human centricity illusion

1 Introduction–Human centricity as a moral credential

Human centricity has become one of the most frequently used, and at the same time one of the most diluted, terms in contemporary discussions on industrial development (Mattsson et al., 2026; European Commission and Innovation, 2021). In its most simple form, it is often described as putting the human at the centre and adapting work to the human. It sounds intuitive, axiomatic even, who would argue against? However, this is precisely why there is a risk for human centricity to be reduced to a slogan, a label that signals good intentions rather than concrete design and management principles. The question is if it is really as easy as just designing systems for the human? Because humans are practically already central to most production systems (Romero et al., 2016a; Romero et al., 2016b), just usually in a way which is neither sustainable nor very human-centric.

Human centricity is frequently invoked as a moral credential, a term that signals the right intentions. Yet this very vagueness enables a convenient misunderstanding across both practice and academia. While Industry 5.0 explicitly places worker wellbeing at the centre of industrial development, the practical operationalisation of this ambition remains less clear (European Commission and Innovation, 2021). Part of the reason, I would argue, is that human centricity is typically invoked without any underlying theory of work: it names a value, not a mechanism. Systems are presented as human-centric because they include humans, consult humans, or provide new tools to humans, while simultaneously remaining dependent on human compensation to function. The result is a discourse where the label spreads faster than the capability it is supposed to denote. This article argues that human centricity has become a widely adopted label across industry and academia, while the socio-technical commitments required to realise it are far less common. The novelty of the present article lies less in observing that humans compensate for system shortcomings, a point already made in earlier work (Bainbridge, 1983; Hollnagel, 2009), than in reframing human centricity itself, an explicit Industry 5.0 policy goal, as a question of whether systems consume or protect human adaptive capacity, and in distinguishing flexibility as competence from flexibility as compensation.

The inspiration taken from these traditions should be stated plainly. From the ironies of automation (Bainbridge, 1983), the argument takes the observation that automating the predictable parts of work leaves humans the harder residual tasks; from the efficiency–thoroughness trade-off (Hollnagel, 2009), that people continuously bridge the gap between demand and design; from the literature on invisible work (Star and Strauss, 1999), that this bridging goes largely unrecognised; and from first-order problem solving (Tucker and Edmondson, 2003), that quick fixes block the learning that would remove their cause. What the human centricity illusion adds is a diagnosis of how these well-known mechanisms are now rebranded: the compensation that earlier counted as a design failure is presented as evidence that the human is at the centre.

Without a doubt, the human worker’s greatest strength as a production resource is their flexibility. Humans are adaptive. They can deal with variation, interpret situations, understand general goals and principles, and improvise when conditions change. In a complex production system, they can also weigh goal conflicts against each other: quality to production rate, safety to efficiency, long term robustness to short term delivery demands. While machines are extremely efficient in the predictable and rule-based, they are often rigid when something unexpected happens. Humans, on the other hand, are built for the unexpected. Human intelligence is inherently situated in the real world, it is context-dependent and goal-oriented and cannot be reduced to purely internal cognitive processes, forming one of the main reasons for human adaptability and flexibility. This view aligns with resilience engineering, where human adaptation is often treated as a central source of successful performance under variable conditions (Hollnagel et al., 2006; Woods, 2018).

Meanwhile, this is where the paradox arises. Because if human flexibility is their greatest strength, would that not also mean that any system that wants to use its resources efficiently, should use this flexibility to the maximum? If we do not design with human adaptability as a prerequisite, would we not then lose the very point of having humans in production? This is a reasonable objection, and it leads to a very important insight: human centricity is not about eliminating flexibility. In fact, it is the very opposite. It is about understanding what type of flexibility we want humans to contribute with, and what flexibility should be carried by the system itself. Because resilience research also warns that adaptive capacity is finite and can be exhausted when systems repeatedly rely on local adaptation rather than designed robustness (Woods, 2018).

As a Perspective article, this paper develops a theoretically grounded conceptual argument about the human centricity illusion and formulates propositions for future empirical investigation, rather than presenting new empirical validation.

1.1 The system shock absorber

Human centricity is often made out as adapting work to the human, but in many industrial systems, this is practically an illusion. Certainly, humans are put at the centre, but not as recipients of a better designed system, but rather as the component that is expected to compensate for system flaws. Operator flexibility and improvisational skills are used as an invisible buffer to deal with variations or disruptions, allowing the system to keep working despite poor robustness. When material is missing, when instructions are unclear, when planning is lacking, or when machines are acting inconsistently, the human is the one who absorbs the consequences by finding workarounds through tacit knowledge or making small adjustments that are never documented but are crucial for continued production. This resembles what the literature on invisible work describes as essential but often unrecognised activities that allow systems to function in practice (Star and Strauss, 1999), leading to increased risk and heightened workload. It works, but it does so at the expense of transparency, learning, and long-term robustness. From a resilience engineering perspective, such local adaptations may produce short-term stability while masking deeper system brittleness (Hollnagel et al., 2006; Woods, 2018). Human centricity becomes, not an expression of human dignity or development, but a way to outsource resilience and stability to the individual. This is why it is quite misleading to say that a system is human-centric based on the human being allowed to be flexible. Flexibility might be an expression of competence, but it can also be an expression of the system being poorly designed. A production system that demands improvisation to work in the everyday, is essentially designed to be dependent on human compensation. Meaning that the human is not a resource used for intelligence and development, but a resource used to absorb faults and correct flaws. This is where you can say that the human is already at the centre, but not as a recipient of adjustment, but as the system shock absorber. In this respect the situation is a contemporary instance of the classic ironies of automation (Bainbridge, 1983): automating the predictable, rule-based parts of the work leaves the operator with the harder residual task of compensating for what the automation cannot handle. It is also closely related to the efficiency–thoroughness trade-off (Hollnagel, 2009); the present argument extends that trade-off from individual performance to the design and governance of supposedly human-centric systems.

Consider a high-mix assembly area, daily production depends on a tight sequence of kitted parts arriving at each station. Variability is common: a supplier label mismatch causes a kit to be delayed; the digital work instruction contains an outdated torque value; and a semi-automated fastening tool occasionally fails its calibration check. None of these issues are catastrophic on their own, but together they create frequent micro-breakdowns. Operators keep throughput stable by improvising: they swap tasks across stations, borrow parts from nearby kits, “know” which torque value is safe despite what the instruction says, and restart or bypass tool checks to avoid stopping the line. The line appears stable in output metrics, and supervisors interpret this stability as evidence of a well-functioning system (Thorvald, 2011; Thorvald, 2021; Kolbeinsson et al., 2023; Thorvald and Mattsson, 2025).

Over time, these workarounds become normalised. The organisation does not see the underlying variability because it is absorbed locally and silently (Star and Strauss, 1999; Woods, 2018). Improvement time is repeatedly postponed when demand increases, since any slack is treated as inefficiency. The result is a classic human centricity illusion, humans may be at the centre, but primarily as the adaptive buffer that compensates for poor robustness. Resilience exists, but it is carried by individuals as workload, risk, and tacit coordination rather than designed into the system as maintainability, clear standards, and protected capacity for learning.

Two boundary conditions should be made explicit. First, the illusion does not require that an organisation explicitly claims to be human-centric. It arises wherever apparently stable, human-friendly operations rest on unrecognised human compensation, although an explicit claim makes the gap between rhetoric and design easier to see. Second, not every workaround is evidence of the illusion. Adaptation to genuine novelty is precisely the competence that humans should contribute in complex socio-technical systems (Hollnagel et al., 2006; Woods, 2018). The diagnostic mark of the illusion is instead repeated, unsupported, and unrecognised adaptation that has become structurally necessary for routine operations without triggering learning or redesign.

2 Flexibility as competence vs. flexibility as compensation

A more mature view of human centricity must therefore distinguish between flexibility as competence and flexibility as compensation. In competence mode, human adaptability is used as a premium resource: to interpret goals under uncertainty, resolve goal conflicts, manage novelty, and contribute to learning and improvement. This aligns with cognitive systems engineering perspectives where human adaptability is treated as a central capability in joint cognitive systems operating under uncertainty (Rasmussen et al., 1994; Hollnagel and Woods, 2005; Woods and Hollnagel, 2006). In compensation mode, adaptability becomes an operating condition: humans absorb variability created by missing materials, unclear instructions, unstable equipment, or planning deficits through tacit workarounds and undocumented micro-adjustments, i.e., invisible work (Star and Strauss, 1999). The human centricity illusion emerges when organisations rhetorically emphasise the former while they structurally rely on the latter. The two modes should not be read as a strict binary but as ends of a continuum. The same adaptive act may express both expertise and compensation; its position on the continuum depends on whether the adaptation addresses genuine novelty or a recurrent, preventable problem, whether it is formally supported and recognised, whether it feeds organisational learning, and whether it accumulates workload and risk for the individual. The model implies that “being human-centric” is not primarily a matter of attitude, but of system design and governance. Incentives, slack, authority, and metrics determine whether flexibility becomes competence or compensation. In the first case, flexibility is a strategic capacity. In the second case flexibility is an operating condition, the issue not being human adaptability itself, but whether the socio-technical system is designed to preserve adaptive capacity or continuously consume it during routine operations (Hollnagel and Woods, 2005; Woods and Hollnagel, 2006).

To make this distinction more than a rhetorical device, it helps to state what it implies. If flexibility can function either as competence or as compensation, then human centricity becomes an empirical and managerial question. Under what conditions does a production system drift into relying on compensation, and what follows when it does? The following propositions synthesise insights from cognitive systems engineering, resilience engineering, and socio-technical systems thinking into a conceptual framing of human flexibility in industrial systems. They are not intended as empirically validated hypotheses, but as conceptual claims intended to support future empirical and managerial investigation. In this sense, their role is not to verify the human centricity illusion empirically in the present article, but to provide a structured conceptual lens through which future empirical studies can identify and examine it. Empirically, the illusion should leave observable traces: a high ratio of undocumented to documented deviations, a high frequency of workarounds, and freed-up capacity that is reabsorbed as output rather than protected for improvement. Such workarounds resemble the first-order problem solving described by (Tucker and Edmondson, 2003), which resolves the immediate disruption while preventing the second-order learning that would remove its underlying cause.

• In production settings characterised by high operational variability and limited system support, human flexibility tends to shift from competence to compensation even when the organisation explicitly aspires to be human-centric.
• As reliance on flexibility as compensation increases, more work becomes invisible in the form of workarounds, tacit coordination, and undocumented micro-decisions, which in turn reduces transparency and learning while increasing mental workload and operational risk.
• When slack is not protected and improvement work lacks explicit mandates, efficiency gains tend to be absorbed as increased output, which means that capacity that is said to be freed up rarely becomes sustained time for improvement, maintenance, or learning.
• Whether human flexibility is used primarily as competence or as compensation is shaped more by governance mechanisms than by human-centred intentions alone, including incentives, authority to stop or redistribute work, the legitimacy of buffers, and the choice of performance metrics.
• When resilience is achieved mainly through flexibility as compensation, resilience is effectively carried by individuals as workload, risk, and local improvisation, whereas when it is achieved through flexibility as competence, resilience becomes an organisational capability that can be stabilised, transferred, and improved.

Taken together, these propositions suggest that the central issue is not whether humans adapt, but whether adaptation functions as a strategic organisational capability or as a continuously required compensatory mechanism. This distinction resembles resilience engineering perspectives that differentiate between resilient system design and resilience achieved primarily through frontline adaptation (Hollnagel et al., 2006; Woods, 2018). The distinction may also be used diagnostically by asking which deviations are documented and which are silently absorbed, which workarounds recur often enough to indicate system brittleness, whether freed-up capacity is protected for learning or converted into more output, and whether operators have recognised authority to stop, redistribute, or redesign work. Further indicators include time spent compensating for missing materials or unclear instructions, the gap between formal procedures and how the work is actually done, and the balance between output-oriented and resilience-oriented performance indicators. Such indicators of the illusion should be kept apart from its consequences, such as stress, fatigue, elevated risk, and reduced organisational learning; the former identify the phenomenon, the latter measure its costs. A full operationalisation is beyond the scope of a perspective article and remains a task for future empirical work.

A truly human-centric system is not one that utilizes human flexibility the most, but the one that does not abuse it. Human adaptability should be a premium resource, it should work as a backup engine in the system, not as the main source of propulsion. A robust system is a system where normal operations happen without humans having to improvise to make every day work, and adaptive capacity should be treated as a strategic reserve within the socio-technical system rather than as a continuously exploited operational necessity (Woods, 2018; Hollnagel and Woods, 2005; Woods and Hollnagel, 2006). When flexibility is needed, it should be connected to something meaningful, solving problems or prioritizing and assessing risks, not constantly saving the flow from collapsing.

3 Protected adaptive capacity is not waste

Is this not fundamentally the same narrative that Lean has already preached for decades? Lean proponents have long spoken about freeing up resources through improvements in efficiency and using these resources for various improvements or quality assuring processes. You might argue that the whole idea surrounding using the human potential is already there. And yet, organisational reality often moves in the opposite direction. When improvements create new resources, these margins are frequently absorbed as increased output rather than protected as capacity for maintenance, learning, or improvement. Time for improvement becomes time for production, and slack is increasingly interpreted as inefficiency rather than as a prerequisite for long-term robustness and resilience. Classical interpretations of the Toyota Production System similarly emphasise continuous improvement, worker involvement, and problem solving as integral parts of production rather than as auxiliary activities (Liker, 2004).

This observation is central because it shows that human centricity cannot be reduced to values or ambitions alone. It is fundamentally a question of governance and socio-technical system design. From this perspective, operational behaviour is shaped less by stated intentions than by incentives, performance measures, authority structures, and the legitimacy of slack and buffers within the production system (Rasmussen et al., 1994; Hollnagel and Woods, 2005; Woods and Hollnagel, 2006). The argument is not directed against Lean as a philosophy of learning, problem solving, continuous improvement, and respect for people. Rather, the critique concerns a common organisational drift in which efficiency gains are captured as increased utilisation, while the capacity required for learning, maintenance, and improvement is not protected. When organisational success becomes narrowly associated with utilisation and short-term output, adaptive human performance risks becoming a substitute for robust system design rather than a complement to it. Under such conditions, both Lean and human centricity risk becoming rhetorical projects: we speak about liberating human potential while simultaneously constructing systems that depend on continuously consuming it.

4 Discussion and conclusions

The decisive question is not how we define human centricity in words, but how we operationalise it in system logic. For human centricity to be more than a fashion statement, it needs to be institutionalised. In other words, human centricity must be embedded in the operational logic of the socio-technical system rather than remaining a rhetorical aspiration. There must be formal mechanisms to protect human adaptive capacity. That means not just encouraging improvements, but actually setting aside time, mandates, and incentives for it. It means that operators must have the opportunity to stop or redistribute work without being punished by production KPIs. It means that management KPIs must assess more than output, they must assess robustness, resilience, stability, and long term sustainability. And it means that a buffer must not be considered inefficient, but a design parameter. These prescriptions echo the principles of high-reliability organizations (Weick and Sutcliffe, 2001): sensitivity to operations, deference to expertise, and a preoccupation with robustness rather than mere output. They also reflect the way the cost–efficiency gradient drives operators toward the boundaries of safe performance (Rasmussen, 1997).

In this sense, human centricity is not a question of being nice to people, but a question of system architecture and governance. The human factor cannot be merely a buzzword in a policy, it must be part of how the production system is managed. Otherwise, human flexibility will always be drawn into the operational logic and consumed where it is the least meaningful, in constant compensation for variations and flaws.

The human centricity illusion is also sustained by organisational power dynamics. Frontline workers may carry the consequences of variability, but they do not always have the authority to redefine standards, slow down production, allocate improvement time, or challenge performance metrics. In such cases, compensation becomes structurally normalised: the organisation benefits from human adaptability while the costs are experienced locally as workload, stress, risk, and invisible coordination. These costs are not incidental. Sustained compensation erodes wellbeing, narrows the space for skill development and meaningful work, and shifts accountability for system deficiencies onto those least able to remedy them.

The conclusion is clear. Yes, human flexibility is the most important reason for why they are still a central resource in production. But a human-centric system is not a system that maximizes the use of flexibility in everyday operations. It is a system that makes flexibility a strategic resource, not a constant emergency solution. The human should not have to save the process every day. They should be able to exercise judgment when really needed. Human centricity is therefore not putting the human at the centre, but about designing systems that do not require the human to stand there to take the blow.

A truly human-centric system is one where resilience is designed, not improvised.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Francesco Facchini, Politecnico di Bari, Italy

Reviewed by: Salvatore Digiesi, Politecnico di Bari, Italy

Rashesh Vaidya, Nepal Open University, Nepal