A Delusion of Control: Loss of Agency in Modern Complex Systems

Despite the ancient mandate from Genesis, control of the earth and everything on it has eluded humanity for millennia. While it is estimated that less than 3 percent of the earth’s land surface remains “faunally intact”—defined as being habitat in which all expected species remain with proper abundance and interaction—the 97 percent that has been altered by humanity looks very little like “dominion” or control (Plumptre et al. 2021). Fisheries we thought we understood collapse, species we try to keep alive go extinct, and even infrastructure we thought to be robust is vulnerable to natural disasters (Finley 2011; Ludy and Kondolf 2012; Pievani 2014).

Mirroring the uncontrollability of species and the biosphere are the global systems that enable life in the modern world—systems such as energy, internet, finance, manufacturing and trade, transportation, telecommunication, health care and epidemiology, and water and agriculture, among others. These systems are complex and expansive in their own right but also combine, integrate, and interact with one another and the environment across various spatial and temporal scales. These connected systems create systems-of-systems that are even more complex, intractable, and, most importantly, largely uncontrollable.

The literature on the subject has long shown that humans can be quite bad at judging their level of “control” over a situation or an outcome (Langer 1975; Thompson et al. 1993). Games of chance are mistaken for those of skill, correlation is confused with causation, and cognitive biases toward optimism motivate people to misjudge their power to impact a result or mitigate a risk. The dangers of an “illusion of control”—the inclination to overestimate one’s ability to influence outcomes—are well-known as gamblers lose fortunes, financial systems face ruin, and military generals meet their Waterloo (Thompson 1999). What is less explored, however, are the consequences of the dissolution of this belief or illusion: what are the dangers associated with people seeing how little control they actually have? A sense of control—motivated by either illusory control or real control—is valuable for decision-making under uncertainty, and can inspire the optimism necessary to take risks and innovate (Thompson et al. 1993). Conversely, losing this sense of control can create debilitating risk aversion and anxiety, arresting the progress toward innovation and advancement (Ottesen and Grønhaug 2005; Harris and Osman 2012).

In this essay, we will begin with a description of a modern paradox in which positive metrics for global quality of life have increased but so, too, have rates of anxiety and depression (O’Leary 2021; Santomauro et al. 2021; WHO 2022a). We then discuss how modernity is enabled through a vast network of complex adaptive systems, which are self-organizing and operate beyond the control of any individual or entity. Next, we discuss the human need for (perceived) control and argue that fragilities in the opaque and inescapable systems we rely upon have tempered human agency in the modern world. We show how membership in—and reliance upon—these systems exposes us to perilous consequences of risk and contagion from others. While it may be comforting to continue the legacy belief that we—as individuals and as a society—are in control, the reality is that both the increasing understanding of risk and the awareness of intractable systemic effects have begun to expose this belief in control as a collective illusion. We posit that recent events highlight our helplessness and dispel the myth that either individual actors, or even those tasked with maintaining the functional operations of systems, are fully capable of pulling the strings. Finally, through the lens of complex adaptive systems and global systemic risk, we offer insights and governance strategies for managing resilience and mitigating systemic risk, with the goal of reducing the uncertainty that each of us faces and restoring greater levels of control.

While global events like the COVID-19 pandemic, the Russian invasion of Ukraine, and conflict in the Middle East have disrupted some positive trends in recent years (FAO et al. 2022; Schöley et al. 2022), there is little doubt that humans in the twenty-first century are far better off than at any other point in history (Pinker 2011; Roser 2020). Though not perfect, this century has seen—and is projected to see—the consistent setting of new high-water marks in metrics like literacy (World Bank 2022; World Economic Forum 2022), life expectancy (Heuveline 2022),¹ reduced infant mortality (Sharrow et al. 2023), per capita GDP (IMF 2023), food security (Zereyesus et al. 2022), and access to—and affordability of—telecommunication technologies (ITU 2021).

Despite these trends and evidence that by many metrics, global quality of life in modern society is materially “better” than at any other point in human history, mental health has been on the decline worldwide (Goodwin et al. 2022; World Economic Forum 2018). This worrisome trend has developed over recent decades, and in the last several years alone, rates of depression and anxiety increased by over 25 percent (WHO 2022b). In the United States and elsewhere in the developed world, these symptoms have led to “deaths of despair,” whereby alcoholism, other addictions, and suicide have become leading causes of mortality for young working-class individuals (Case and Deaton 2015, 2021).

As we will show below, research identifies agency and a sense of control as critical ingredients to a population’s well-being. We propose that a contributing factor to this global decline in mental health—despite the quality-of-life improvements—is a growing perception of lost control and the conscious or unconscious realization that our autonomy, personal agency, and sense of security are threatened by the realities of globalized and interconnected modern life.

We posit that the erstwhile illusion of control has been attenuated by two factors. First, inaccurate perception of control—irrational overconfidence—may have declined as delusions have been replaced by knowledge, education, and humbling self-awareness of the limits of our own control (Bernstein 1996). Evolving understanding of probability, accelerating abundance of information and data, and the exponential growth of computing power have provided greater clarity around risk and uncertainty (Arrow 1996; Klir 2006; Lanzetta and Driscoll 1968; Peterson and Pitz 1988). This phenomenon is apparent as probability analyses have become ubiquitous and increasingly integral to modern advancements like navigation software, financial portfolio management, and meteorology, and even in less critical areas such as sabermetrics and other sports analytics (Markowitz 1952; Bar-Shalom, Li, and Kirubarajan 2002; Šarčević et al. 2021; Gallian 2010). Second, when we study the global systems that support our lives, we see that the locus of control has shifted outward from autonomy and autarchy into the system. As this devolution toward the system diminishes our actual level of control over our well-being, our rational sense of agency thus progressively declines. Current events provide frequent reminders that our growing dependence on the increasingly technological global systems-of-systems is undeniably reducing our individual agency (Newnam and Goode 2015; Dalcher 2007; Funtowicz and Ravetz 1994).

The benefits of modern society have been made possible, in large part, by the massive global complex adaptive systems (CAS) that enable the continuous flow of information, money, goods, services, and people around the world. These aforementioned systems of energy exploration and production, electricity transmission, computer networks, health care, food and water supplies, transportation networks, commerce, and finance, among others, facilitate our contemporary way of life but also introduce layers of uncertainty and unpredictability. While global interconnectivity is not new, the current “Digital Age of Globalization” is unique for the sheer scale, tight coupling, high levels of interdependence, and instantaneous characteristics of these modern networks (Sachs 2020).

As dynamic global systems grow in their numbers of nodes and links, they become CAS, which are self-organizing with collective behaviors that cannot be reduced to the sum of their individual components. Additionally, CAS are decentralized, are constantly evolving, and can behave in ways that are unpredictable and nonlinear (Holland 1992; Miller and Page 2007). New emergent properties take shape, with systemic characteristics like tipping points, feedback loops, contagion, cascades, and synchronous failures adding layers of complexity to the structure and dynamics of the intricate web that supports modern lifestyles (Centeno et al. 2023a).

Complexity has been a natural consequence of specialization of labor, which has enabled the assembly of increasingly smaller and more technical parts to create advanced technologies. Adam Smith’s description of the division of labor that allowed a pin factory to achieve greater output was a parable of increasing economic complexity (Smith 1776). As individual economic actors moved away from self-reliance to systems of barter, exchange, and commerce, full control over their well-being was gradually traded for the benefits of this greater productivity (Becker and Murphy 1992). Specialization and division of labor allowed for greater innovation and complexity of design, with parts and inputs now sourced through highly complex supply chains, and with efficiency managed increasingly toward the goal of just-in-time inventory (Yang and Sachs 2003; Ibrahim et al. 2021).

Similarly, the benefits of economies of scale and technological advancement have shifted greater power into the hands of a small number of large companies. For example, it is estimated that as of 2023, Taiwan Semiconductor Manufacturing Company (TSMC) supplied up to 56 percent of the global semiconductor market (International Data Corporation 2023), and some estimates of Apple’s US market share of smartphones are as high as 55 percent (Counterpoint Research 2023). In the financial sector, the consolidation of banks and the moral hazard of government bailouts has increasingly concentrated US deposits, with the top four banks accounting for 24 percent of total assets of financial institutions (Statista 2023).

These seemingly abstract concepts of specialization and complexity had harsh concrete manifestations during the COVID-19 pandemic. Global demand created disruptions in supply chains for personal protective equipment (PPE) at the time of greatest need (Cohen and Rodgers 2020). Similarly, supply chain shocks from lockdowns created a shortage of semiconductor chips that held up the manufacturing of automobile assembly lines (Ewing and Boudette 2021), while millions were affected by job loss and business closures (Naseer et al. 2023).

These systemic effects on both health and commerce—all of which were outside the scope of control for those most affected—began with a localized health crisis in China and quickly propagated around the world, creating cascades and contagions through the global systems-of-systems. Beyond the pandemic, examples of systemic global disruptions are alarmingly frequent and costly (WEF 2023). A ship run aground in the Suez Canal delays global cargo for weeks (Lee and Wong 2021), an uncharacteristically cold February in Texas leaves millions without power (Flores et al. 2023), and a dictator’s decision to invade a neighboring country sends food and energy prices soaring around the world (Hendrix 2022).

In short, life in the modern Digital Age of Globalization means that individuals (whether they want to or not) find themselves at the mercy of systems and sprawling institutions that are impossible to fully understand, much less fully control. As described by Anthony Giddens, “no one can completely opt out of the abstract systems involved in modern institutions,” and we find ourselves putting more and more trust in esoteric “experts” who may or may not have things in hand (Giddens 1990, 84). Giddens goes on to liken modern life to a ride “aboard a careering juggernaut” that does not take orders or direction and can hardly be influenced by its passengers (Giddens 1990, 53).

Whether the ability to direct this juggernaut exists or not, there is little doubt that belief in our own agency and control over our environment is critical for human well-being (Chorpita and Barlow 1998; Kay et al. 2009; Langer and Rodin 1976). Research suggests that humanity is hardwired to seek out a feeling of control (Leotti, Iyengar, and Ochsner 2010) and shows that when outcomes feel governed by randomness or chaos instead of by deliberate action, fears from the inherent uncertainty increase levels of depression and anxiety with potentially long-term consequences for mental and physical health (Tullett, Kay, and Inzlicht 2015).

As the world gives us more reasons to doubt possible control, some seek comfort in conspiracy theories (van Prooijen and Acker 2015; van Prooijen and Douglas 2017). A conspiracy that someone—even if that someone is a bad actor—has a plan and the ability to execute it can be reassuring. As individuals become more affected by global systems and distant events beyond their personal agency, a conspiracy theory offering explanatory power and determinism is more attractive than the confusion of complexity and randomness. These theories thrive and grow in popularity during periods when feelings of helplessness, anxiety, and loss of control increase as well—essentially replacing one illusion of control for another (Douglas et al. 2019).

It is clear that humans do better when we believe we have agency (Taylor and Brown 1988). As noted above, however, we each depend on a system-of-systems—composed of the CAS that provide our daily necessities—over which we have very limited control. As fragilities in these systems become exposed, the belief in our agency becomes harder and harder to maintain. Global events like the financial crisis of 2008–9, COVID-19, climate change, and geopolitical conflict are all stark reminders of how little control we have over dramatic shifts that profoundly impact our daily lives. Like Ulrich Beck’s description of a blade of grass in Bavaria, whose life “ultimately comes to depend on the making and keeping of international agreements,” we find ourselves increasingly at the mercy of forces far beyond our abilities to influence (Beck 1992, 23). For the vast majority of us, everyday necessities like food, water, electricity, credit cards, and even smartphones rely upon networks that we couldn’t “fix” if they were to be disrupted. Those of us who depend on these modern systems are beginning to realize that we might not be in the driver’s seat. More worrisome still is the question (to continue the metaphor): is there anyone in the driver’s seat? And even if there were, is there any way of knowing whether the steering wheel itself is even connected at all?

Scholarship on civilizational collapse shows that this increasing loss of control in complex societies is a recurring phenomenon, which can precipitate fatal consequences for civilizations (Gibbon 1788; Tainter 1988, 1995). Recent studies of this historical pattern have addressed how increasing connectivity and interdependence of previously autonomous agents initially contribute to the growth phase of civilization, while late-stage fragilities and collapse are often associated with unwieldy complexity, excessive levels of dependence, and the eventual inexorable loss of agency and control (Bookstaber 2023; Sandberg 2023; Tainter 2014). This loss of control is often masked by the illusion that tomorrow will be just like yesterday, and ultimately contributes to diminished resilience and the unraveling of societies (Centeno et al. 2023b). Throughout history, the illusion of being able to control complex adaptive systems has allowed the confidence and optimism to take the risks necessary for societies to grow and flourish (Heunks 1998; Rigotti, Ryan, and Vaithianathan 2011). Yet the hindsight of civilizational collapse exposes this illusion as a dangerous delusion, much like the hubris of Shelley’s Ozymandias.

What can be done to establish critical aspects of control—genuine empowerment and meaningful management—of the global systems that we rely upon daily? The first important tool we discuss is governance: critical to the sustainable functioning and operation of complex systems. Like a governor on an engine, which limits speed within tolerance levels, governance is synonymous in many ways with control and can be essential to establishing the parameters within which the structure and dynamics of CAS take shape.

Given that even one CAS alone is far too dynamic and unpredictable to fully control, the prospect of governing an interconnected network of multiple such systems is even more intractable. Effective governance in this context is considered to be the most challenging type of “wicked problem,” where goals are difficult to define, progress is elusive to test, and solutions that prove effective in one scenario cannot be widely applied more generally to others (Rittel and Webber 1973; DeFries and Nagendra 2017; Heifetz 1994). The wickedness of this challenge—and what we know about CAS—means that it will never be possible to restore full confidence and individual agency within these unwieldy systems, but some governance strategies and principles may help provide greater capacity for adaptation as well as systemic reliability, stability, and resilience.

Often overlooked in discussion of governance is that responsible self-governance—greater control in the form of risk management by the members of the systems—provides an important foundation for systemic resilience (Espinosa 2022). As individuals and entities, the self-organizing nodes linked together in the network benefit from systemic efficiency, trading the control, autonomy, and risks of self-reliance for the advantages, efficiencies, and other risks of dependence upon the system (Ostrom 2015; Espinosa and Guzman 2015). As interdependent nodes take on greater systemic risk, the potential for a dangerous positive feedback loop develops: weak nodes create a fragile system, and a fragile system further threatens the stability of nodes. Vulnerabilities appear, as exogenous shocks have greater potential to trigger the cascading failure of nodes and links, precipitating systemic collapse (Crucitti, Latora, and Marchiori 2004; Ash and Newth 2007).

Both the financial crisis of 2008–9 and the COVID-19 pandemic exposed the ways in which many individuals and entities participating in systems had abdicated their bottom-up governance, trading their fiduciary obligations of prudence for the benefits of efficiency and risk (Bender et al. 2022; Quétel et al. 2022). The safety and security of stockpiling reserves have been supplanted by the convenience of modern systems. Households are increasingly managed with just-in-time inventory: ordering a battery to be shipped the same day or having meals delivered to the doorstep (Kassas and Nayga 2021). In this way, households have followed businesses in opting for the convenience and financial benefits of on-demand purchasing, just-in-time inventory, and consumer debt (Vermeulen 2015). This shift has led to a decline in resilience in both companies and households, which live at the margin by forgoing reserves, redundancies, buffers, diversification of suppliers, and rainy-day funds (Lusardi, Schneider, and Tufano 2011), drawing us closer to the apocryphal adage that “civilization is three meals away from anarchy.”

It is important to note, however, that systemic risk and endogenous failures can manifest even when all individual nodes are prudently self-governed and careful about their own individual exposures. The principle of emergence tells us that systems might produce outcomes predicated not on the actions of their individual parts but rather on their interactions (Centeno et al. 2015). For example, we might monitor the quality of apples to make sure no bad fruit infects the entire basket, but perhaps the problem is not an individual rotten apple but the very co-presence of so many apples that precipitates spoilage. A systemic view of risk management, therefore, is critical for long-term stability, and complete dependence on bottom-up strategies will not effectively restore resilience and the confidence of control that comes with it (Chambers et al. 2019).

At the other end of the governance spectrum, top-down approaches can be valuable in certain contexts. For example, in “tragedy of the commons” scenarios such as natural resource depletion, electrical grid blackouts, and water scarcity, the aggregation of self-interested individual choices can overwhelm systemic integrity, requiring oversight. Top-down governance is also relevant within systems that prioritize safety, such as air traffic control and nuclear power. Intervention can take various forms, such as designing architecture of systems for robustness and resilience, actively monitoring and regulating the stability of systemic structure and dynamics, stress testing, and directing the recovery following systemic failure. Examples of top-down systemic interventions include stock markets with “circuit breakers’” to stop feedback loops and panics (Sifat and Mohamad 2019), electrical grids implementing rotating power outages to manage load (Menati and Xie 2021; Wang, Hong, and Li 2021), and disaster recovery by FEMA (Olshansky and Johnson 2014). Top-down strategies can also be effective at identifying and making use of “leverage points” to make significant change to system behavior through minor intervention (Meadows 2008). A danger here, however, is that once these levers are identified, there is substantial risk that they could be pushed for change in the wrong direction (Forrester 1971, 1968).

In this way, top-down intervention can create potential for significant unintended consequences that reduce resilience and erode a sense of individual agency and control. Within the financial system, for example, the possibility of top-down governance intervention—in the form of government bailouts—has incentivized institutions to abdicate bottom-up self-governance as banks take larger risks to achieve economies of scale and global competitiveness (Andersen and Jensen 2022; Mishkin 2006). These selective governance interventions have inadvertently concentrated market share in the hands of a smaller number of large “too big to fail” banks, as assets are siphoned away from smaller banks where depositors face the risk of not being bailed out (Eufinger, Gorostiaga, and Richter 2023; Davila, Burge, and Hammen 1997). The moral hazard of rescues and bailouts inspired Nassim Taleb to argue that one of the pillars of a “black-swan-robust world” is “no socialisation of losses and privatisation of gains” (Taleb 2009).

The challenge of navigating the wicked problem of complex systems governance requires bespoke management solutions to promote and maintain stability and resilience. Research has advocated for the concept of “adaptive governance” as a potentially effective solution that provides dynamic flexibility while balancing the trade-offs and conflicts between bottom-up and top-down, small and large scales, and local and global coordination (Walker and Salt 2012; Carpenter et al. 2009). Evolving from the fields of ecosystem and resource management, the theory of adaptive governance emphasizes nimble, multilevel, and dynamic management that incorporates new information, which feeds back to adjust policy (Folke et al. 2005; Boyd and Folke 2012). Recognizing the various flexible forms and applications of adaptive governance, Chaffin, Gosnell, and Cosens note:

In contrast, nonadaptive governance styles are fully rigid, unresponsive to change, and struggle to learn from the past and the present. “Indeed, nonadaptive governance of a dynamic system with changing thresholds,” write Walker and Salt, “is bound to fail” (Walker and Salt 2012, 131).

While perhaps simple to discuss in theory, adaptive governance strategies can prove difficult to implement in the real world (Schultz et al. 2015). One reason is that governance systems have themselves become large, unwieldy, and complex, making the prospect of using one complex system to manage others an example of what Galaz calls a challenge of “double complexity” (Galaz 2011). Another challenge relates to how much current models of adaptive governance rely on human relationships and trust (Schultz et al. 2015). Despite these difficulties, the principles of adaptive governance and management are critical to consider if any comfort and confidence are to be restored in the unpredictable and nested complexity of the modern world.

Dependence on the CAS of our modern world has, in many ways, diminished our ability to actually control our outcomes and well-being. Further, our rising awareness of these intractable systems has shined a light on our inefficacy, shattering the illusion that we individually have agency to meaningfully affect our globalized and interconnected present. The illusion that we as individuals—or even as a society—can have dominion over the runaway juggernaut that is our modern complex systems-of-systems is a dangerous optimism bias. However, the anxiety that comes from supplanting this illusion with a diminished sense of control and autonomy—and the resignation that the world is unpredictable, chaotic, and random—is an unsatisfying alternative.

In this paper, we connect the negative consequences of feeling out of control to the reality that our CAS-enabled modernity has eroded human agency. The specialization of our modern world certainly necessitates the concession of some levels of control—we cannot each be our own heart surgeon, car mechanic, pilot, farmer, etc.—but an awareness of our position within the larger system and an increased understanding of how complexity creates vulnerabilities in our lives can be empowering. Through our study of complexity and systemic risk, we can begin to understand the ways in which our autonomy has diminished, and how we might work to reclaim a sense of control.

We illustrate how complete dependence on top-down strategies can make society vulnerable to unintended consequences. We will never understand the systems-of-systems well enough to know exactly which of the countless levers to pull, how far, and for how long, or what intervention is exactly appropriate at any given time. Conversely, a focus on only bottom-up approaches can be limiting and ineffective at taking advantage of larger systemic benefits. We illustrate how many of these challenges are addressed by adaptive governance strategies, which incorporate a more nuanced and flexible approach to these complex dynamic systems. We argue that while certainly not a “silver bullet” to solve all issues of agency and control in CAS, elements of adaptive governance—which should incorporate perspectives from multiple levels of the system, incentivize resilience over efficiency, and foster more confidence in the longevity of shared systemic structures and dynamics—can help guide our modern systems toward more stability. Designing governance paradigms that employ the lens of complexity, global systemic risk, and the understanding of inexorable interdependence will allow both individuals and policy-makers to more effectively navigate the complex systems-of-systems that define globalization, restoring a greater sense of agency and control in an uncertain world.

The authors have no competing financial or nonfinancial interests to declare. Miguel A. Centeno is the Global Perspectives section editor for Global Epistemologies: Concepts, Methodologies, and Data Systems. He was not involved in the review process for this article.

Miguel A. Centeno is Musgrave Professor of Sociology at Princeton University and Executive Vice Dean of Princeton University’s School of Public and International Affairs. He is founder and co-director of the Princeton Institute for International and Regional Studies (PIIRS) Global Systemic Risk research community.

Peter W. Callahan studied history at Princeton University and earned his MS in geography and environmental studies from the University of New Mexico. He is a researcher at Princeton’s PIIRS Global Systemic Risk research community, where his scholarly interests include the study of socio-ecological systems, historical systemic risks, sustainable development, and renewable energy policy and technology.

Paul A. Larcey is co-director of the PIIRS Global Systemic Risk research community at Princeton University. Larcey’s work with the UK’s innovation agency focuses on key emerging technologies including life sciences, quantum technologies, and AI. He has worked in corporate research, venture capital, and global industrial sectors at board and senior levels and studied engineering, materials science, and finance at London, Oxford, and Cambridge Universities.

Thayer S. Patterson is coordinator and a founding member of the PIIRS Global Systemic Risk research community at Princeton University. Following his studies in economics and mechanical engineering at Yale, and finance at Princeton’s Bendheim Center for Finance, his research has focused on the causes and consequences of catastrophic systemic risk.