The core claim of the article is simple but consequential. All complex adaptive systems are complex systems, but not all complex systems are adaptive. A complex system, in the standard definition, is any system composed of many interacting components whose aggregate behavior cannot be understood by analyzing the parts in isolation. Its behavior is shaped by nonlinear interactions and feedback loops, meaning that small changes in one part of the system can have disproportionately large effects elsewhere, or vice versa. Such systems often operate far from equilibrium and exhibit emergent patterns: coherent structures or behaviors, such as storms or jet streams, that arise from local interactions but are not properties of any individual component. The authors stress that this notion of complexity contrasts sharply with systems that are merely complicated. A commercial airliner is complicated but can, in principle, be fully decomposed into subsystems whose behavior is specified by design. A turbulent fluid, a power grid under fluctuating load, or the global climate cannot be understood this way, because feedbacks, nonlinearities, and stochastic influences produce behavior that is more than the sum of the parts.

Crucially, complex systems do not have to be adaptive. Many paradigmatic complex systems in physics and engineering involve components that follow fixed rules and never learn. Molecules in a fluid, stars in a galaxy, or components in an electrical grid interact in intricate ways, but the rules of interaction are set by physical laws and engineering design. The system can show unpredictability, phase transitions, or critical phenomena, yet the components themselves do not change their behavior in response to outcomes. The authors offer the weather as the canonical example. Atmospheric dynamics are governed by fluid mechanics and thermodynamics, and interactions among air masses, heat flows, moisture, and topography generate emergent features such as storms. The system is open, nonlinear, and famously sensitive to initial conditions, which keeps forecasting challenging despite sophisticated models and data. But neither air parcels nor pressure systems adapt in the sense of learning: they do not revise their governing rules based on past forecasts or policy decisions.

A complex adaptive system adds a crucial feature to this picture. Drawing on John Holland’s definition, the authors describe a CAS as a system that involves many components that adapt or learn as they interact. In CAS, components, often called agents, change their state, strategies, or rules in response to experience. Adaptivity typically implies some form of memory or internal structure that can be updated, and it often leads to coevolution between agents and their environment. Examples abound in the life and social sciences. In an ecosystem, species evolve traits that affect and are affected by the traits of other species and by the abiotic environment. In an immune system, cells learn to recognize and respond to novel pathogens, with the repertoire of responses shaped by previous encounters. In economies and financial markets, firms and investors adjust strategies in response to prices, regulations, and each other’s behavior. Cities, too, are complex adaptive systems: households, firms, and institutions adapt to infrastructure, policies, and one another, producing shifting patterns of land use, mobility, and inequality.

Adaptation can also occur at multiple, nested levels. An agent at one level, such as a person, may itself be a complex adaptive system composed of interacting subsystems while simultaneously functioning as a component of a higher-level CAS, such as a city. In many biological and social CAS, this adaptivity is hierarchically organized: higher-level structures constrain and enable local behavior even as local interactions generate emergent system-level patterns. The authors trace the intellectual history of this distinction, noting that Walter Buckley introduced the phrase complex adaptive system into sociology in the 1960s, and that Holland and colleagues at the Santa Fe Institute later elaborated CAS as a unifying framework for systems with adaptive agents, from economies to ecosystems. In parallel, physicists and mathematicians advanced general theories of complex systems, including chaos, self-organization, and critical phenomena, often without explicit adaptivity. The CAS tradition, the authors emphasize, did not emerge as a mere correction to theories of self-organization; it developed around a distinct concern with how agents, populations, or institutions learn, remember, and revise behavior over time.

Why does this distinction matter in practice? The authors identify three consequences, the first being model choice. System dynamics models, aggregate representations of stocks, flows, and feedbacks popularized in the systems thinking community, are well suited to complex systems whose behavior is dominated by structural feedbacks rather than explicit agent-level adaptation. But they generally do not represent individual agents adapting their rules or strategies over time. When applied to systems where agent heterogeneity and learning are central, such as urban mobility, financial markets, or the spread of misinformation, purely aggregate models can miss key dynamics. The authors give a vivid example: an urban planner might construct a system dynamics model of vehicle flows and road capacity and conclude that adding a new road will reduce congestion. In reality, drivers adapt their routes, developers change land use, and the new capacity induces additional demand, potentially leaving congestion unchanged or even worse. An agent-based model that explicitly represents adaptive travelers and developers would be far better aligned with the system’s adaptive character.

The same logic applies in epidemiology and public health. Simple compartmental models of disease spread treat individuals as homogeneous and non-adaptive: they do not change contact patterns or behavior in response to risk or policy. In actual epidemics, people adapt; they reduce contacts, alter compliance, and respond to information. Treating an epidemic as a non-adaptive complex system may be adequate for early, short-term forecasting, but for long-term policy evaluation, the authors argue, a CAS perspective and models that incorporate behavioral responses are critical. The distinction also shapes how evidence is found and interpreted. The authors note that in the resilience literature, differences in terminology, including whether papers used complex systems, complex adaptive systems, or social-ecological systems, affected which studies were retrieved in systematic reviews. Important work on adaptive ecosystems could be overlooked if search terms were restricted to complex systems in a narrow sense, while treating all complex systems as adaptive risks importing results from physical systems into domains where agent behavior fundamentally changes the dynamics.

The third consequence concerns intervention itself. In a non-adaptive complex system such as a power grid, interventions change flows and feedbacks, but the components do not change their rules in response. Unintended consequences such as blackouts or cascading failures may still occur, but there is no need to worry about deliberate strategic responses by the system’s elements. By contrast, most policy-relevant systems are complex and adaptive: economies, cities, governance networks, and social-ecological systems all contain agents that respond to incentives and information. Pathogens evolve resistance, financial actors find loopholes, and communities reorganize in response to planning decisions. This adaptive response is what distinguishes CAS from non-adaptive complex systems and makes them both fascinating and difficult to govern. As Colander and Kupers argue in work cited by the authors, complexity-aware public policy should focus less on top-down control and more on creating conditions under which adaptive agents self-organize toward desirable outcomes. Nel and Taeihagh likewise highlight the soft underbelly of complexity science adoption in policymaking: non-technical challenges such as institutional incentives, capacity, and trust can make or break attempts to use complexity-informed tools.

The authors are careful to note that the CS/CAS distinction should not be read as forcing a rigid either/or ontology or a single preferred method. Many research problems require complementary descriptions: complex systems models may be appropriate for structural, physical, or network constraints, while CAS models are needed to represent learning, strategic adjustment, or evolution. The distinction is therefore methodological as much as semantic; it helps researchers decide when complementary or hybrid approaches, combining system dynamics, network science, and agent-based modeling, are necessary, especially for socio-technical systems that mix physical infrastructure with adaptive agents. As large-scale data on human behavior, ecosystems, and infrastructure proliferate, the authors see an opportunity to test when non-adaptive approximations suffice and when explicitly modeling adaptation is essential. Education and training in complexity science, they add, should explicitly address the CS/CAS distinction and its practical implications for students, policy analysts, and decision-makers.

Ultimately, the message of the Perspective is one of conceptual honesty. Complexity science now sits at a critical juncture: its concepts and tools are increasingly institutionalized, with dedicated schools, research centers, and journals, yet the rapid uptake of complexity thinking in applied domains risks diluting its meaning. For some problems, adopting a complex-systems lens means paying attention to feedback, nonlinearity, and emergent patterns in systems whose components do not adapt. For others, especially in life and social domains, complexity thinking must include adaptivity, learning, evolution, and, in some cases, goal-directed and self-regulating forms of behavior. Collapsing these cases under a single, vague notion of complexity risks disappointment and misdirected effort. Using the language of complex systems and complex adaptive systems with care, the authors conclude, can make complexity thinking a more precise, powerful, and trustworthy guide for action.

Subject of Research: Distinction between complex systems and complex adaptive systems

Article Title: Complex systems vs. complex adaptive systems: why the difference matters

Article References: Shutters, S. T., & Dhein, K. (2026). Complex systems vs. complex adaptive systems: why the difference matters. Frontiers in Complex Systems, 4, Article 1808634. https://doi.org/10.3389/fcpxs.2026.1808634

Image Credits: AI Generated

DOI: 10.3389/fcpxs.2026.1808634

Keywords: complexity science, complex adaptive systems, agent-based modeling, system dynamics, policy, emergence, Complex, systems, adaptive, difference, matters, scientific research