Restricted versus general complexity: epistemological foundations for reflexive systems practice

A persistent tension runs through applied systems disciplines: as methodologies grow more sophisticated, their epistemological foundations remain surprisingly stable. Whether in systems engineering, operations research, information systems design or organisational systems analysis, practitioners deploy increasingly elaborate techniques whilst maintaining paradigmatic commitments that date to mid-twentieth-century cybernetics and systems theory. This tension becomes particularly acute when systems disciplines extend into domains characterised by what we might term “deep human dimensions”: meaning-making, social dynamics, ethical contestation and reflexive relationships between observers and observed.

Consider the contemporary landscape of applied systems work: artificial intelligence systems mediating employment, criminal justice and healthcare decisions (Benjamin, 2019; Crawford, 2021); climate adaptation systems requiring governance across unprecedented temporal and spatial scales (Haasnoot et al., 2013); critical infrastructure spanning national boundaries and cultural contexts; democratic information platforms shaping collective sense-making. In each domain, systems approaches emphasising decomposition, optimisation and technical control confront phenomena that resist such treatment not because methods are insufficiently refined, but because the phenomena themselves exhibit properties that methodological sophistication alone cannot adequately address.

The question this article pursues is not whether applied systems disciplines should incorporate participatory methods, acknowledge multiple perspectives or attend to ethical dimensions; few would dispute these aspirations. Rather, we ask whether the epistemological foundations underlying most systems practice permit these aspirations to be realised meaningfully, or whether they constitute what might be termed “sophisticated restricted complexity”: elaborations of method that leave fundamental paradigmatic commitments intact.

Edgar Morin's distinction between “restricted complexity” and “general complexity,” developed initially in La Méthode (1977–2004) and articulated through numerous publications (Morin, 1992, 2007, 2008), provides analytical resources for examining this question. Restricted complexity treats complexity as a problem to be overcome through methodological rigour. Complex problems are understood as merely complicated ones that can be managed through systematic decomposition, formalisation and control. The underlying assumption is that with sufficiently sophisticated techniques, complex systems can be fully specified, designed and controlled from an external observational position. This orientation seeks what Morin terms “blind intelligence”: an epistemology that maintains rigid separations between order and disorder, between observer and observed, between system and environment.

General complexity, in contrast, recognises complexity as an ontological condition with profound epistemological implications. At its etymological root, complexity derives from complexus: “that which is woven together” (Morin, 2008, p. 5). Certain features of complex systems, including emergence, recursivity and the entanglement of observer and observed, cannot be eliminated through better methodology but must instead be acknowledged and engaged. This is not a call for less rigour but for a different kind of rigour: one that integrates rather than excludes, that embraces rather than eliminates uncertainty and that treats knowledge itself as a system, open, recursive and dependent on its knower.

Second-order cybernetics, developed by von Foerster (1974, 2003) and others, complements Morin's framework by foregrounding the observer's constitutive role. Where first-order cybernetics studied observed systems, second-order cybernetics insists on studying observing systems, including the observer's role in constituting what is observed. As von Foerster (2003, p. 283) emphasised, the role of the observer is thereby appreciated and acknowledged rather than disguised, as had become traditional in western science. This insight transforms how we understand systems practice. The practitioner's disciplinary training, organisational position, cultural background and professional interests all shape what is perceived as relevant, what counts as a legitimate concern and what constitutes an acceptable solution. These are not distortions to be corrected through better methodology but constitutive features of any observing system.

This article makes three contributions. First, we demonstrate how to employ restricted and general complexity as an analytical tool for examining systems methodologies, identifying four characteristic patterns of restricted complexity and showing how they pervade practice despite methodological innovation. Second, we articulate connections between Morin's framework and second-order cybernetics, showing how these complementary perspectives together provide resources for epistemological transformation in systems practice. Third, we develop principles for what we term reflexive systems practice, grounded in general complexity and second-order cybernetics, with implications extending across applied systems disciplines.

Systems engineering serves as our detailed empirical case because its methodologies are highly codified in international standards, making epistemological assumptions explicit and examinable. However, the framework applies more broadly. Operations research confronts similar tensions when optimisation models encounter value conflicts (Ackoff, 1979). Information systems design faces parallel challenges when technical architectures embody power relationships (Winner, 1980). Organisational systems analysis struggles with comparable issues when intervention methodologies assume external consultant positions that actual practice belies (Schön, 1983). The restricted/general complexity distinction offers analytical resources for examining these patterns across disciplines.

The article proceeds in four movements. Section 2 establishes how restricted complexity manifests in systems engineering practice, examining core methodologies, softer innovations and their epistemological foundations. Section 3 presents general complexity and second-order cybernetics as alternative epistemological frameworks, developing their implications for boundaries, emergence and predictability. Section 4 articulates principles for reflexive systems practice. Section 5 concludes with theoretical contributions and a research agenda for epistemological transformation.

A clarification about scope: whilst this article focuses substantially on systems engineering, it addresses primarily the epistemological foundations codified in standards and professional literature rather than the full range of actual practice. Individual practitioners often transcend methodological limitations through tacit knowledge and practical wisdom (Schön, 1983). However, the institutional frameworks within which they operate (standards, procurement processes, liability regimes, educational curricula) remain grounded in restricted complexity, constraining what innovations can be sustained and legitimised. Our critique addresses these institutional frameworks as manifestations of broader epistemological patterns in applied systems disciplines.

Before examining how restricted complexity manifests in practice, we establish criteria for its identification. Morin's distinction is not simply descriptive but analytical; it provides a lens for examining how systems methodologies conceptualise and engage complexity. We propose that restricted complexity exhibits four characteristic patterns.

First, decomposition as fundamental strategy. Restricted complexity assumes that complex problems can be adequately addressed by breaking them into manageable components, analysing these systematically and recombining results. Whilst all analysis involves some decomposition, restricted complexity treats this as sufficient: the whole is understood as the sum of properly analysed parts.

Second, external observability. Restricted complexity positions the analyst or practitioner as standing outside the system, capable of objective analysis and neutral design. The observer's role in constituting what is observed is treated as a methodological problem to be minimised rather than an epistemological fact to be engaged.

Third, controllability through method. Restricted complexity assumes that with sufficiently rigorous methodology, systems can be fully specified, their behaviour predicted and desired outcomes achieved through proper technique application. Uncertainty is treated as epistemic (stemming from incomplete knowledge) rather than ontological (intrinsic to complex systems).

Fourth, instrumental primacy. Restricted complexity privileges technical and economic objectives, treating other concerns (ethical, political, aesthetic) as constraints to be satisfied rather than as constitutive dimensions of systems themselves. Questions of meaning, purpose and value become matters of requirements specification rather than ongoing interpretive work.

Systems engineering emerged in the mid-twentieth century as a response to increasing technical complexity in aerospace and defence systems (Hughes, 1998). Its development paralleled and drew upon cybernetics and general systems theory, yet incorporated these frameworks in ways that exemplified restricted complexity's assumptions. Contemporary systems engineering, as institutionalised in international standards and professional handbooks, maintains this orientation despite decades of methodological innovation.

The ISO/IEC/IEEE 15288:2023 standard exemplifies restricted complexity's epistemological stance. It provides a comprehensive framework encompassing life cycle processes from stakeholder requirements through validation and disposal, assuming relatively stable environments, linear development trajectories and clearly defined objectives (ISO/IEC/IEEE, 2023). The standard's emphasis on requirements completeness, traceability matrices, baseline control and verification against predefined criteria reflects an underlying assumption that systems can be fully specified, designed and controlled from outside through structured processes (Blanchard and Fabrycky, 2011; INCOSE, 2023).

Consider requirements engineering, a core systems engineering activity. From a restricted complexity perspective, requirements are discoverable facts to be elicited through systematic stakeholder interviews, documented in specifications and verified against delivered systems. The INCOSE Systems Engineering Handbook acknowledges that the perception and definition of a particular system depends on an observer's interests and responsibilities (Walden et al., 2015, p. 18), yet treats this as a matter of perspective management rather than a fundamental epistemological challenge. The observer remains conceptually external to the system, able to adopt different viewpoints but not fundamentally implicated in the system's constitution.

Yet as Morin (1992) observes, complex systems possess “eco-auto-causality” or recursive causality, where the organising process elaborates the products, actions and effects necessary for its own generation. This fundamentally circular relationship resists the linear causality assumed by phase-gate development processes. When a healthcare information system is deployed, it does not simply serve predefined functions but transforms clinical workflows, power relationships among medical professionals and patients' experiences of care, which in turn reshape how the system is understood and used. Baseline control processes systematically fail to accommodate such recursive dynamics.

Systems engineering has not remained static. Soft systems methodology (SSM), developed by Checkland (1981), arose precisely from recognising the limitations of “hard” systems approaches when confronting messy, ill-defined problem situations involving multiple stakeholders with divergent worldviews. SSM's emphasis on learning rather than optimisation, its use of rich pictures and root definitions and its acknowledgment of multiple perspectives represent significant departures from traditional technical orientations. Similarly, critical systems thinking, as articulated by Jackson (2019) and Midgley (2000), developed methodological pluralism and boundary critique specifically to address earlier approaches' limitations.

More recently, several developments have addressed aspects of complexity: model-based systems engineering (MBSE) promises more agile, integrated approaches through formal modelling languages (Friedenthal et al., 2008); resilience engineering (Hollnagel et al., 2006) explicitly embraces uncertainty and adaptation; systems-of-systems engineering (Jamshidi, 2009) grapples with emergence across organisational boundaries. These represent genuine advances in capability.

Yet despite these innovations, mainstream systems engineering practice has largely resisted fundamental epistemological transformation. As Jackson (2000, p. 368) observes, whilst critical systems thinking recognised that early hard approaches were suitable for tackling certain well-defined problems but were found to have limitations when faced with complex problems involving people with a variety of viewpoints, systems engineering has continued to position itself primarily as a technical discipline. The incorporation of softer methods within systems engineering practice often occurs within an overarching framework that maintains restricted complexity's core assumptions.

This represents what might be termed “sophisticated restricted complexity”: an elaboration of methods that maintains rather than transforms the fundamental epistemological stance. MBSE, despite enabling more flexible modelling, typically maintains assumptions of decomposability and external observability. Models are treated as representations of systems “out there” rather than as performances that constitute systems through acts of distinction (Espejo and Reyes, 2011). Resilience engineering acknowledges uncertainty but often instrumentalises it, treating adaptive capacity as an additional system property to be designed rather than recognising adaptation as fundamentally emergent from stakeholder interactions that cannot be fully specified in advance.

Stakeholder engagement becomes a technique for requirements elicitation rather than genuine co-creation; participatory design methods are employed instrumentally to improve user acceptance rather than to fundamentally redistribute power in system definition; ethical considerations are addressed through checklists and impact assessments rather than through sustained ethical deliberation. As Checkland and Scholes (1990) observed, SSM is often misappropriated as simply another problem-solving technique rather than recognised as embodying a fundamentally different epistemology.

Second-order cybernetics offers resources for understanding why softer approaches have been incorporated without epistemological transformation. Traditional systems engineering positions the engineer as standing outside the system, capable of objective analysis and neutral design (INCOSE, 2023). Yet as Scott (1996) demonstrates through analysis of high-modernist planning schemes, this stance ignores how the practitioner actually participates within the system being designed.

This connects to a crucial insight from critical systems heuristics: system boundaries are not discovered through neutral analysis but are socially constructed through value-laden judgements (Churchman, 1979; Ulrich, 1983). Consider defining the “system” for urban transportation. Does it include only vehicles and infrastructure? Also air quality impacts? Climate effects? Social equity in access? Effects on urban form and community cohesion? Displacement of low-income residents through induced gentrification? Each boundary judgement privileges certain stakeholders and concerns whilst marginalising others.

Engineers who define the system as “vehicles and infrastructure” are not discovering a pre-existing system but enacting a particular system through their professional gaze, one that systematically excludes those experiencing transportation as a determinant of housing security or community integrity. Traditional systems engineering treats boundary definition as a technical activity within the System Definition process (INCOSE, 2023), thereby naturalising these boundaries, making them appear as objective features of reality rather than contingent products of specific social, political and institutional arrangements. As Ulrich and Reynolds (2010) demonstrate, determining system boundaries involves implicit choices about who counts as a stakeholder, what concerns are deemed relevant and whose values are privileged. Malaina (2014) extends this critique to computational models, arguing that they ignore the nature of any complex system as constructed by the observer and are unable to explain the sociohistorical construction of the agents and the social structure in which they are located.

The fourth pattern, instrumental primacy, manifests in systems engineering's emphasis on optimisation, efficiency and technical solutions. The INCOSE Systems Engineering Handbook states that practitioners work in the context of designing for affordability and performance (INCOSE, 2023, p. xxi), revealing an underlying assumption that technical and economic optimisation constitute primary values. Habermas (1971, 1984) identified this pattern as the dominance of instrumental reason and its tendency to colonise other domains of human activity. Instrumental rationality, oriented toward efficiency and control, represents what Habermas termed the “technical interest,” one of three fundamental cognitive interests guiding human inquiry. The practical interest in mutual understanding and the emancipatory interest in freedom from domination represent equally legitimate but fundamentally different orientations. When every concern must be translated into measurable objectives and optimisable parameters in order to gain legitimacy, the discipline systematically excludes forms of knowledge and value that cannot be captured in these terms.

Whilst this section has focused on systems engineering, similar patterns appear across applied systems disciplines. Operations research exhibits comparable tendencies: problems are modelled as optimisation exercises; the analyst stands outside the problem situation; multiple perspectives are incorporated through multi-criteria decision analysis that assumes preferences can be adequately captured in utility functions (Ackoff, 1979). Information systems development confronts analogous challenges. The traditional systems development lifecycle assumes requirements can be systematically specified and success objectively verified, yet as Winner (1980) demonstrated, technical architectures embody political arrangements, and as Friedman and Nissenbaum (1996) showed, interface designs encode values. Organisational systems analysis similarly struggles with the analyst's position: Schön (1983) demonstrated that the “technical rationality” model fails to capture how professionals actually work in complex situations characterised by uncertainty, value conflicts and unique circumstances. These patterns suggest that restricted complexity is not specific to systems engineering but reflects broader epistemological commitments across applied systems disciplines.

Edgar Morin's theory of general complexity offers not another complexity theory to be added to existing frameworks but a fundamental reconceptualisation of complexity itself. Where restricted complexity treats complexity as a methodological challenge, general complexity recognises complexity as an ontological condition that transforms what knowledge can be and how it must be pursued. Complex systems are not structures of discrete parts with predetermined roles, amenable to analysis through decomposition and recombination. Rather, they are dynamic configurations of interdependencies, contradictions and emergent properties that cannot be disentangled without fundamentally transforming what is being studied. The act of analytical separation itself becomes an intervention that must be acknowledged rather than a neutral operation that reveals pre-existing structure.

Morin's work merits particular attention in systems science for several reasons. His sustained engagement with cybernetics, particularly second-order cybernetics, and general systems theory provides direct bridges to systems science's intellectual foundations whilst simultaneously offering radical critiques of how these foundations have been interpreted. His concept of “self-eco-organisation” specifically addresses the relationship between autonomy and environmental embeddedness, offering resources for understanding systems that are simultaneously self-organising and contextually embedded (Morin, 2008). And his epistemological framework explicitly engages the relationship between knowledge and the knower, providing foundations for reflexive practice that second-order cybernetics demands.

Central to Morin's critique is his concept of “blind intelligence” (intelligence aveugle): the epistemology that maintains rigid separations between order and disorder, observer and observed, system and environment. This intelligence is “blind” not because it lacks sophistication but because its very sophistication prevents it from seeing what it excludes. Morin distinguishes between two orientations towards rigour itself. Restricted complexity pursues rigour through elimination: eliminating ambiguity through precise definition, eliminating uncertainty through better data and eliminating the observer's influence through methodological controls. General complexity pursues rigour through integration: integrating multiple perspectives while acknowledging their incommensurability, embracing uncertainty while maintaining analytical precision, acknowledging the observer's participation while sustaining systematic inquiry.

Four interconnected principles are particularly relevant to systems practice.

The dialogical principle acknowledges that complex systems involve elements that are simultaneously antagonistic and complementary. Order and disorder are not opposed states but interdependent aspects of organisation. Life itself exemplifies this: living systems maintain order through constant metabolic activity that is itself disordered at molecular levels; they achieve stability through processes of continual transformation. For systems practice, this implies that conflicts, contradictions and tensions are not problems to be resolved but features to be engaged. When stakeholders hold genuinely incommensurable perspectives on a system, this incommensurability cannot be eliminated through better communication or integration methods; it must be sustained, worked with and acknowledged as constitutive of the system itself.

The recursive principle recognises that complex systems exhibit circular causality: products and effects are simultaneously causes and producers of what produces them. Social systems exemplify this profoundly: individuals produce society through their interactions, whilst society produces individuals through socialisation. Neither can be taken as foundational; each produces the other in an ongoing recursive process. For systems science, this challenges linear models of causation that underlie much systems analysis. Interventions enter recursive loops rather than acting on external causes, transforming intervention from external action to participation in ongoing processes.

The hologrammatic principle suggests that not only is the part in the whole but the whole is inscribed in the part. Each element of a complex system contains information about and is shaped by the whole, yet the whole emerges from interactions among parts. This principle has profound implications for system boundaries and levels of analysis, suggesting that local and global are not separate scales but mutually inscribed: understanding any part requires understanding the whole it participates in producing, yet the whole exists only through its parts.

The principle of auto-eco-organisation recognises that complex systems are simultaneously autonomous and dependent, self-organising yet environmentally embedded. The boundary between system and environment is not a passive interface but an active production, continuously enacted through the system's operations. As Varela et al. (1974) demonstrated in autopoietic theory, which Morin engages extensively, the fundamental question becomes not how systems adapt to pre-existing environments but how they specify the environments relevant to their organisation.

Morin insists that general complexity demands reconceptualising knowledge itself as a system: open, recursive and dependent on its knower. This epistemological turn distinguishes Morin's complexity from other complexity theories that maintain conventional epistemological assumptions whilst acknowledging complexity in phenomena.

Knowledge cannot be separated from the knower. Every act of knowing involves a subject who observes, distinguishes, categorises and interprets. This subject brings conceptual frameworks, disciplinary training, linguistic structures, cultural assumptions and practical interests that shape what can be observed and how it can be understood. Moreover, knowledge itself exhibits recursive organisation. Our knowledge shapes how we observe, which shapes what we know and which shapes how we observe. Scientific paradigms, in Kuhn's (1962) sense, exemplify this: they provide frameworks that make certain observations possible whilst rendering others literally unthinkable, and these observations in turn reinforce the paradigm. Breaking such recursive loops requires what Morin terms “epistemological reflexivity.”

Significantly, Morin argues that systems theory itself, despite intentions to overcome reductionism, often reproduces it at a higher level. By seeking to explain wholes rather than parts, systems theory maintained the explanatory logic of reduction; it simply shifted the level of analysis. General systems theory, Morin argues, must become reflexive, acknowledging its own limitations and the observer's role in system constitution.

Second-order cybernetics, developed by Heinz von Foerster, Humberto Maturana, Francisco Varela and others, complements Morin's framework by foregrounding the observer's constitutive role in system description. Von Foerster's formulation is direct: first-order cybernetics is the cybernetics of observed systems, whilst second-order cybernetics is the cybernetics of observing systems (von Foerster, 2003, p. 285). This shift transforms cybernetics from a science of control and communication in systems “out there” to a science that includes the observer in its descriptions.

Several key insights follow. Observation is fundamentally an act of distinction: drawing a boundary that separates something from its background, creating a distinction between system and environment. Following Spencer-Brown's (1969) Laws of Form, which von Foerster engaged extensively, the observer draws this distinction; it does not exist prior to observation. Different observers, drawing different distinctions, constitute different systems. What counts as “the system” depends on the distinctions an observer draws, which depend on the observer's purposes, interests and conceptual frameworks.

Second-order cybernetics insists on including the observer in the domain of description. But observers are themselves systems described by other observers (including themselves). This creates recursive loops that cannot be eliminated but must be engaged. Maturana and Varela's (1980) work on autopoietic systems develops this insight: living systems are organisationally closed (self-producing through recursive operations) whilst structurally open (exchanging matter and energy with environment). Knowledge is not a representation of an independent reality but an effective action that allows an observer to maintain organisation in a domain of interactions.

Von Foerster (1992) argued that second-order cybernetics leads necessarily to questions of ethics because it reveals that fundamental decisions about how to observe cannot be made through observation alone. When observation itself is recognised as constitutive, the question becomes: how should we observe? This cannot be answered objectively because any answer would require a prior observational stance. Von Foerster developed what he termed an “ethics of ethics”: not a set of rules to apply but an orientation towards enabling others' ethical capacity. This reframes ethics from application of predetermined principles to cultivation of spaces for ethical reflection and dialogue.

Luhmann’s (1995) social systems theory represents the most sustained operationalisation of second-order cybernetics for social analysis. Drawing on von Foerster's concept of observation as distinction and Spencer-Brown's (1969) calculus of indications, Luhmann developed a comprehensive framework in which social systems are constituted through communication rather than through the actions of individual subjects. In his account, second-order observation becomes a defining characteristic of modern functionally differentiated society: social systems maintain themselves by observing how other systems observe (Leydesdorff, 2006). Valentinov (2017) has traced this development from Wiener's first-order cybernetic concern with feedback and complexity to Luhmann's reconceptualisation of feedback as self-referential observation, demonstrating how the transition from first- to second-order cybernetics entails not merely including the observer but fundamentally reconceiving what systems are: operationally closed, self-referential networks of communication that specify their own boundaries through distinction.

This trajectory has been extended in several directions within Kybernetes. Lepskiy (2018) has traced the evolution from first-order cybernetics (observed systems under classical rationality) through second-order cybernetics (observing systems under non-classical rationality) to what he terms third-order cybernetics: the study of self-developing reflexive-active environments under post-non-classical rationality. This developmental arc resonates with our argument: restricted complexity corresponds to the epistemological commitments of first-order cybernetics, whilst general complexity aligns with the reflexive, participant-centred orientation that second- and third-order cybernetics demand. Espejo and Lepskiy (2021) have pursued this further in developing what they call ontological cybernetics, combining organisational cybernetics with reflexive-active environments to address social responsibility. Their work demonstrates that the bridge from cybernetic theory to applied practice, which our framework also seeks to build, is an active concern within the Kybernetes community.

Luhmann's framework also illuminates a productive tension for our argument. His theory has been criticised for insufficient attention to normative and emancipatory dimensions (Habermas, 1984), precisely the domains that our fourth and fifth principles (navigating multiple rationalities and cultivating ethical deliberation) seek to foreground. Recent Kybernetes scholarship has engaged this tension directly. Laursen et al. (2022) introduced a special issue on moral communication observed through social systems theory, demonstrating that Luhmannian analysis can examine moral phenomena without itself performing moral judgement, yet leaving open the question of how normative engagement is possible from within a systems-theoretic framework. Our contribution addresses this gap: where Luhmann's framework provides powerful descriptive resources for understanding how social systems observe and reproduce themselves, reflexive systems practice complements this with the ethical and participatory dimensions that his framework systematically brackets.

Morin's general complexity and second-order cybernetics developed largely independently, emerging from different intellectual traditions (French complexity studies and Continental philosophy vs Anglo-American cybernetics), yet they converge on fundamental insights. Their complementarity provides particularly powerful resources for systems science.

Morin addresses ontology; cybernetics addresses epistemology. General complexity reconceptualises what complex systems are: configurations exhibiting recursivity, emergence and dialogical principles. Second-order cybernetics reconceptualises how we know systems: through observations that are themselves systematic operations by observing systems. Together, they transform both what systems are understood to be and how they can be known. Morin emphasises emergence; cybernetics emphasises distinction. These are complementary: systems emerge through interactions, but what counts as the relevant system depends on distinctions drawn by observers. Morin provides critique; cybernetics provides method. General complexity offers sustained critique of simplifying paradigms; second-order cybernetics provides methodological resources: the observer as necessary inclusion in system description, recursion as fundamental operation, distinction as primary act.

Kauffman's (2005) development of von Foerster's eigenform concept provides formal grounding for this complementarity. By demonstrating that objects are tokens for eigen behaviours, stable forms produced through recursive processes. Kauffman shows that what appears as a pre-existing entity is itself a product of the observer's recursive engagement. For the problem of enacted boundaries that is central to our argument, eigenform theory provides a specific mechanism: system boundaries are not static demarcations but eigenvalues, stable patterns that emerge through recursive interaction between an observing system and its domain. This reconfigures the boundary question from one of correct placement to one of understanding how recursive processes stabilise particular distinctions whilst rendering others invisible. Scott (2004), in his historical account of the development of second-order cybernetics, has shown how this trajectory from observation of systems to observation of observation was latent in cybernetics from its inception, suggesting that the restricted complexity we diagnose represents not merely a philosophical limitation but a failure to follow through on cybernetics' own foundational insights. This reinforces the claim that systems are enacted through observation rather than existing independently of it and connects Morin's ontological insights about recursivity with second-order cybernetics' formal treatment of self-reference.

Perhaps most fundamentally, both frameworks insist that the stance of external observation is untenable. Morin argues that knowledge itself is a system dependent on its knower; von Foerster insists that observers are always already part of what they observe. Both conclude that practitioners cannot stand outside systems but must acknowledge their participation within them.

An important qualification is warranted. Whilst general complexity and second-order cybernetics move decisively beyond restricted complexity's separation of observer and observed, they do not fully dissolve the subject–object structure. Morin's principles still function partly as categories describing properties of complex systems; second-order cybernetics includes the observer but retains the observer as a system relating to other systems. Adjacent traditions push further towards fully relational epistemologies. Fuenmayor's (1991a, b) interpretive systemology, grounded in phenomenological hermeneutics, argues that systems thinking must be understood as an interpretive activity embedded in historically constituted horizons of meaning rather than as observation conducted by a subject upon an object. Escobar (2018) develops relational ontologies in which entities do not precede their relations but emerge through them. Barad's (2007) concept of intra-action similarly proposes that phenomena are produced through mutual constitution rather than interaction between pre-existing entities.

These perspectives suggest that the move from restricted to general complexity, whilst necessary, may represent an intermediate position. A fully relational epistemology would recognise that neither practitioner nor system exists prior to the relational configuration producing both. This article occupies such an intermediate position deliberately: sufficiently beyond restricted complexity to enable genuine epistemological transformation, whilst retaining enough of the practitioner-as-participant structure to remain actionable for applied systems disciplines. The tension between relational depth and practical applicability is acknowledged here as a productive frontier for future work.

Both frameworks fundamentally reconceptualise system boundaries. In restricted complexity, boundaries are treated as discoverable features of systems, identifiable through systematic analysis. From the perspective developed here, boundaries are enacted through observation rather than discovered. Every act of system definition involves drawing distinctions that are simultaneously descriptive and constitutive. Ulrich's (1983) critical systems heuristics develops this insight methodologically, providing questions that make boundary judgements explicit: whose interests are served? Who is included as stakeholder and who excluded? Morin's hologrammatic principle adds further depth: boundaries are not simply drawn around systems but are produced by systems themselves through their autopoietic operations (Maturana and Varela, 1980). This has three implications for practice: boundary judgements must be recognised as ethical and political acts; boundaries are provisional and revisable; and multiple boundaries can be drawn around the same situation, constituting genuinely different systems.

Regarding emergence, restricted complexity typically treats it as an epistemic limitation: properties not yet predictable from knowledge of components. From the perspective developed here, emergence is ontological. Complex systems exhibit properties that cannot in principle be predicted from analysis of components, because the properties emerge through interactions that transform what components are. Von Foerster (1984) distinguished between trivial and non-trivial machines. Trivial machines are synthetically deterministic; non-trivial machines are analytically indeterminate: their output depends on their history, they can change their own rules of operation, and their behaviour cannot be fully predicted even with complete knowledge of structure. Living systems, social systems and most sociotechnical systems are fundamentally non-trivial. Treating them as trivial machines constitutes a category error. Practice must therefore be reoriented towards what might be termed “navigation”: developing capacity to respond to emergence, learning through interaction and maintaining adaptability. Haasnoot et al. (2013) term this “adaptive pathways” in climate adaptation planning: rather than producing master plans based on deterministic forecasts, adaptive approaches work through flexible strategies that evolve based on monitoring, stakeholder input and changing conditions.

Finally, both frameworks transform how knowledge and expertise are understood. Restricted complexity treats knowledge as objective understanding of systems, with expertise consisting of mastery of methodologies. General complexity and second-order cybernetics insist that knowledge is situated, contextual and dependent on the knower. This resonates with Schön's (1983) critique of “technical rationality.” The practitioner's participation is not a problem to be minimised but a resource to be engaged. Moreover, if knowledge is situated and emergence irreducible, then local knowledge, practical experience and diverse perspectives become epistemologically necessary, not merely politically desirable.

What does systems practice look like when grounded in general complexity and second-order cybernetics rather than restricted complexity? This question cannot be answered through methodological prescription alone. The inadequacy of restricted complexity is paradigmatic rather than methodological; new techniques deployed within unchanged epistemological frameworks simply reproduce existing limitations in more sophisticated forms.

What follows is an articulation of principles for reflexive systems practice: practice that acknowledges complexity as ontological, treats observation as constitutive and embraces the practitioner's participation as epistemologically necessary. Three clarifications frame what follows. First, reflexive practice is not an addition to existing practice but a transformation of it. Second, reflexive practice is not opposed to rigour but pursues rigour differently: through integration rather than elimination, acknowledgment rather than control, dialogue rather than optimisation. Third, reflexive practice does not provide certainty but offers something arguably more valuable: epistemological resources for engaging complexity with honesty about what can and cannot be achieved.

Practitioners are participants within the systems they engage, not external observers capable of objective analysis and neutral design. This follows directly from second-order cybernetics' insistence that observers cannot stand outside what they observe. Acknowledging participation transforms practice in several ways. It makes reflexivity methodologically necessary: practitioners must continually examine their own assumptions, question whose interests are served by particular framings and remain alert to how their positions enable certain observations whilst constraining others. It transforms stakeholder engagement from a technique for gathering information into an ethical imperative emerging from recognition that practitioners have no privileged access to system definition. And it requires acknowledging that practice itself is an intervention that transforms what is being studied: design activities do not act on pre-existing systems but participate in ongoing processes of system formation.

System boundaries are not discovered facts but contingent enactments that require ethical and political justification, not merely technical rationale. This follows from both general complexity's insistence that systems are enacted through observation and second-order cybernetics' recognition that observation is fundamentally distinction-drawing. Boundary judgements must be made explicit and subject to deliberation: what distinctions are being drawn? Why these rather than others? Whose perspectives are included and whose excluded? What concerns are deemed relevant and what dismissed as out of scope? Practitioners must acknowledge that multiple boundaries can legitimately be drawn around the same situation, constituting genuinely different systems. A transportation system defined by engineers as “vehicles and infrastructure” is genuinely different from one defined by community organisers as “mobility justice, including effects on housing security and community integrity.” Both are valid enactments; the question is not which is correct but what purposes and values each serves. And boundary work is continuous rather than completed at project initiation.

Uncertainty in complex systems is irreducible, arising from emergence, recursivity and the impossibility of complete knowledge rather than from incomplete information or inadequate methodology. Treating uncertainty as ontological transforms practice from control to navigation. It requires reconceptualising planning through what Haasnoot et al. (2013) term adaptive pathways: flexible strategies that evolve based on monitoring, stakeholder input and changing conditions. Plans become hypotheses to be tested through implementation rather than blueprints to be executed. It requires designing for adaptability rather than optimality, developing monitoring and response capabilities for emergent phenomena and cultivating epistemic humility: accepting that some consequences cannot be foreseen and some aspects of system behaviour will emerge unpredictably.

Technical and economic optimisation, whilst important, constitute only one legitimate form of rationality. Drawing on Habermas's (1971) distinction among cognitive interests, practice must also engage practical wisdom concerning meaning and value, and emancipatory reasoning concerning power and justice. This requires recognising that not all legitimate concerns can be translated into measurable objectives and optimisable parameters. Questions of meaning, purpose, dignity and justice resist quantification not because we lack adequate metrics but because they are fundamentally qualitative. Practice needs spaces where stakeholders can articulate what matters in their own terms, express incommensurable values without forced reconciliation and deliberate about meanings that technical language cannot capture. It requires attending to power relationships: whose knowledge counts as expertise? Whose concerns are treated as legitimate? And it requires practitioners to develop capacities beyond technical expertise, including facility with ethical reasoning, social analysis and interpretive understanding. Crucially, this navigation of multiple rationalities does not presuppose a single standpoint from which integration is accomplished. Each rationality produces its own positions, meanings and criteria of validity. The task is not to synthesise these into a unified framework but to sustain their productive tension, enabling practitioners to move responsibly among different modes of reasoning without reducing any to the terms of another.

Ethical concerns cannot be addressed through specialty processes or impact assessments but must be woven throughout practice as continuous deliberation. This follows from von Foerster's (1992) ethics of ethics: when observation is recognised as constitutive, questions of how to observe cannot be answered objectively. Ethical deliberation requires moving from applied ethics (applying predetermined principles) to deliberative ethics (creating conditions for ongoing reflection about what matters and why). Questions such as “Whose interests does this design serve?” and “What values are embedded in this technical choice?” cannot be answered definitively at project initiation but must be revisited as understanding evolves. It requires creating institutional arrangements that support deliberation: cross-disciplinary teams, formal mechanisms for raising ethical concerns, psychological safety and organisational memory capturing how value conflicts were navigated. And it requires acknowledging that some ethical tensions cannot be resolved through clever design but must be sustained through governance.

These principles manifest differently across systems disciplines. We offer brief illustrations showing how reflexive practice transforms core activities.

In requirements and problem formulation, reflexive practice reconceives these as an ongoing dialogical process of meaning-making among participants with different perspectives, values and forms of expertise. In systems engineering, requirements documents should include explicit sections documenting boundary judgements (what was excluded and why), source stakeholders for each requirement (making power dynamics visible) and requirement evolution. In operations research, problem formulation should document whose definition of the problem is adopted, what alternative formulations were considered and what concerns are rendered invisible by the chosen formulation. In information systems development, requirements gathering should be reconceived as participatory design where users are recognised as co-designers rather than sources of information.

In design and architecture, reflexive practice treats design decisions as simultaneously technical and ethical choices. Architecture documents should include “reflexive annotations” making explicit the values embedded in technical choices. Design reviews should ask: whose work does this design make easier and whose harder? What assumptions about users' capabilities does it embed? In urban planning, infrastructure designs should document not merely technical specifications but whose mobility needs are prioritised and what communities bear costs.

In verification and evaluation, reflexive practice reconceives these not as confirmation against predetermined criteria but as inquiry into what actually emerged through implementation, including unintended consequences and effects on marginalised groups. Post-deployment reviews should occur at regular intervals. Systems should include mechanisms for affected parties to contest operation, not merely report defects. Evaluation becomes learning rather than mere judgement.

In risk and uncertainty management, reflexive practice expands beyond technical and economic risks to include epistemic risks (arising from boundary judgements that excluded relevant considerations), power risks (shifts in power relationships harming vulnerable groups) and meaning risks (transformations in how people understand themselves or relationships). These cannot be quantified probabilistically but must be deliberated qualitatively.

Realising reflexive practice requires transformations extending beyond individual practitioners to educational systems, organisational arrangements and professional standards.

Systems education must develop capacities for reflexive engagement alongside technical expertise: integrating science and technology studies examining how technologies embody values; teaching critical systems thinking and boundary critique as core content; developing studio-based courses where students practise reflexive documentation on real projects; and assessing not only technical proficiency but capacity to articulate assumptions, identify excluded stakeholders and engage productively with value conflicts.

Organisations must create structures supporting reflexive practice: cross-disciplinary teams bringing diverse perspectives; formal mechanisms for ethical deliberation integrated throughout development rather than relegated to review boards; adaptive funding and development cycles that respond to emergence rather than rigid milestone progression; and organisational memory systems capturing not only technical decisions but boundary judgements, value conflicts and how these were navigated.

Professional standards must evolve to legitimise reflexive practices: requiring documentation of boundary judgements as core artefacts; mandating stakeholder analysis that explicitly identifies excluded groups; including reflexive review as required lifecycle activity; providing guidance on navigating value conflicts; and recognising uncertainty as intrinsic rather than treating it solely as risk to be mitigated.

Three objections warrant response. First, does acknowledging participation undermine objectivity? This objection assumes objectivity requires the observer's separation from what is observed, but second-order cybernetics demonstrates this is impossible. Acknowledging participation does not undermine rigour but pursues it differently: through systematic examination of how position shapes observation rather than pretending position can be eliminated. The alternative, assuming objectivity whilst actually participating, produces systematic blindness to how one's position shapes outcomes.

Second, is this too abstract to guide actual practice? Section 4.3 provided concrete illustrations. However, we maintain that epistemological transformation must precede detailed methodological prescription. New techniques within unchanged epistemological frameworks reproduce existing limitations. The relationship between epistemology and methodology is recursive: epistemology guides methodology, which in turn tests and refines epistemology.

Third, does embracing uncertainty endanger safety-critical systems? This confuses epistemic humility with recklessness. Acknowledging that systems cannot be fully predicted does not mean abandoning safety practices; it means adopting practices better matched to reality. Safety-critical systems benefit from reflexive approaches: explicitly documenting what assumptions safety arguments depend on, maintaining vigilance for emergent hazards and designing for graceful degradation. Illusions of complete control may increase danger by discouraging vigilant monitoring and adaptive response.

This article has developed Morin's distinction between restricted and general complexity as an analytical framework for examining epistemological foundations in applied systems disciplines. Our contribution operates at several levels.

First, we have demonstrated how to employ restricted and general complexity as an analytical tool. We showed that restricted complexity exhibits four characteristic patterns (decomposition, external observability, controllability through method and instrumental primacy) and demonstrated through detailed examination of systems engineering standards and innovations how restricted complexity pervades practice. Crucially, we showed that methodological innovations, including SSM, critical systems thinking and contemporary developments, have often been incorporated without transforming epistemological foundations. This reveals “sophisticated restricted complexity”: elaborations of method that leave paradigmatic commitments intact.

Second, we have articulated systematic connections between Morin's general complexity and second-order cybernetics. These frameworks developed largely independently yet converge on fundamental insights: complexity as ontological rather than merely epistemic, observation as constitutive rather than neutral and participation as unavoidable rather than eliminable. Together they constitute epistemological resources for transformation: Morin addresses what complex systems are, whilst second-order cybernetics addresses how we know systems.

Third, we have developed principles for reflexive systems practice grounded explicitly in these frameworks. Previous calls for reflexive practice in systems work (Schön, 1983; Checkland and Scholes, 1990; Midgley, 2000) have operated primarily at pragmatic or methodological levels. By grounding reflexivity in explicit epistemological frameworks, we provide theoretical justification for why reflexive practice is necessary (not merely desirable) and what it fundamentally entails. The five principles, acknowledging constitutive participation, treating boundaries as enacted, engaging uncertainty as ontological, integrating multiple rationalities and cultivating ethical deliberation, constitute a coherent epistemological stance rather than an ad hoc collection of practices.

This work extends critical systems thinking (Jackson, 2019; Midgley, 2000) by providing explicit epistemological grounding. Critical systems thinking has developed primarily through pragmatic engagement with methodological pluralism and boundary critique. By connecting these developments to Morin's general complexity and second-order cybernetics, we provide theoretical foundations that strengthen critical systems thinking's claims whilst subjecting them to epistemological scrutiny.

We have also extended second-order cybernetics by demonstrating its implications for applied systems practice in concrete detail. Much second-order cybernetics literature remains abstract or philosophical (von Foerster, 2003; Glanville, 2007). By showing how observer participation, distinction-drawing and recursion manifest in actual systems methodologies, we bridge theory and practice in ways that make second-order cybernetics more accessible to applied systems disciplines.

Our framework complements but differs from complexity science traditions (such as those of the Santa Fe Institute) that focus on computational modelling of complex adaptive systems, often maintaining restricted complexity's assumptions about external observability. General complexity insists on including the observer in analysis. Our contribution is showing how general complexity provides epistemological foundations that complexity science's methodologies often lack.

We also acknowledge that interpretive and relational traditions within systems thinking have pursued critiques of subject–object dualism that parallel aspects of our argument. Fuenmayor's (1991a, b) interpretive systemology developed a phenomenological critique of systems methodology that anticipates several concerns raised here, particularly regarding the observer's embeddedness in interpretive horizons. Relational ontologies (Escobar, 2018) and agential realism (Barad, 2007) offer resources for pushing beyond the intermediate position this article occupies. Our contribution is not to claim that the critique of restricted complexity is itself novel in the broadest sense, but rather to operationalise Morin's specific distinction as a diagnostic framework applicable to codified systems methodologies, and to connect this diagnosis with second-order cybernetics in ways that yield actionable principles for reflexive practice.

The contribution developed here can be situated more precisely within the debates that characterise Kybernetes as a venue for second-order cybernetics and social systems theory. Roth and Sales (2025), in their inaugural editorial, articulate the journal's orientation as pursuing a shift from viewing systems as objects of control to engaging them as sites of observation and participation. The restricted/general complexity distinction provides analytical vocabulary for diagnosing precisely where this shift has and has not occurred in applied systems disciplines. Our four patterns of restricted complexity (decomposition, external observability, controllability through method, instrumental primacy) identify the specific epistemological commitments that sustain the object-of-control stance, whilst the principles of reflexive practice articulate what the observation-and-participation stance requires in applied contexts.

Several strands of recent Kybernetes scholarship converge on the terrain this article addresses. Espejo (2015) and Espejo and Lepskiy (2021) have demonstrated how organisational cybernetics and reflexive-active environments can operationalise second-order cybernetic principles for applied practice. Lepskiy (2018) has mapped the evolution from first- through second- to third-order cybernetics, tracing developments in reflexivity that parallel our movement from restricted to general complexity. The growing body of Luhmannian scholarship in the journal (Bombaerts, 2023; Laursen et al., 2022; Leydesdorff, 2006) demonstrates sustained engagement with questions of observation, distinction and moral communication. Umpleby (2005) has pursued connections between von Foerster's constructivism and scientific practice that anticipate our concern with reflexive engagement. What has been largely absent from these debates is Morin's specific contribution: a systematic diagnostic framework that identifies where and how restricted complexity persists despite explicit second-order cybernetic commitments. Our article brings this diagnostic resource into dialogue with the Kybernetes community's ongoing work on observation, reflexivity and applied cybernetics.

The framework opens multiple research directions organised around four streams.

Theoretical development should include comparative epistemological analysis: how does system dynamics relate to restricted/general complexity? What epistemological stance does the viable system model assume? How do network approaches treat observation and participation? The restricted/general complexity distinction itself requires further refinement: are there intermediate positions? Can restricted complexity be appropriate for certain problem domains whilst general complexity is necessary for others? Integration with critical realism (Bhaskar, 1975), actor-network theory (Latour, 2005), feminist epistemology (Haraway, 1988) and Indigenous knowledge systems (Cajete, 2000) would enrich general complexity and these related approaches. Of particular importance is deeper engagement with relational epistemologies (Barad, 2007; Escobar, 2018) and interpretive systemology (Fuenmayor, 1991a, b), which offer resources for moving beyond the intermediate position this article occupies towards a fully relational understanding of systems practice.

Methodological development should operationalise the five principles: what specific practices realise each principle? What tools, techniques and facilitation approaches support reflexive practice? Assessment frameworks for evaluating reflexive capacity in individuals, teams and organisations are needed. Boundary critique methods for time-constrained professional contexts require development, as do approaches for reconciling adaptive methods with institutional requirements for planning and accountability.

Empirical investigation should include comparative case studies examining how systems initiatives embody restricted vs general complexity, longitudinal studies comparing reflexive and traditional approaches, participatory action research collaborating with practitioners experimenting with reflexive methods, and cross-cultural studies challenging the framework's potential Western biases.

Disciplinary and educational research should examine what educational approaches effectively develop reflexive capacity, how professional norms and reward structures shape viable epistemological stances, how systems science can engage more productively with science and technology studies and critical social theory, and what happens when teams include members with different epistemological commitments.

A work arguing for reflexive practice must acknowledge its own position and limitations. We have argued for general complexity and second-order cybernetics as superior to restricted complexity, yet this argument reflects epistemological commitments that could themselves be examined. Our detailed case focused on systems engineering, and more extensive empirical work across multiple disciplines would test and refine claims. This work draws primarily on Western systems thinking traditions, and the framework retains a residual subject–object structure that relational and interpretive traditions (Barad, 2007; Escobar, 2018; Fuenmayor, 1991a, b) have more thoroughly problematised; how these frameworks relate to non-Western and Indigenous knowledge systems represent both a limitation and a vital research direction. We have articulated principles but provided limited detailed methodological guidance, reflecting our emphasis on epistemological foundations whilst recognising that future work must develop detailed methods. These acknowledgements do not undermine the argument but exemplify reflexive practice: acknowledging position, recognising limitations and remaining open to revision.

Systems thinking faces both unprecedented opportunities and fundamental challenges. Systems increasingly mediate fundamental aspects of human life: algorithmic systems shape employment, criminal justice and healthcare; digital platforms structure social interaction; smart infrastructure governs urban life. When systems have such pervasive influence, questions about how they are developed become inseparable from questions about power, justice and human flourishing. The gap between restricted complexity's assumptions and the actual nature of these phenomena grows increasingly apparent.

General complexity and second-order cybernetics provide resources for engaging these challenges. They offer not certainty about outcomes or prescriptions for action, but epistemological frameworks that acknowledge complexity honestly, treat observation as constitutive and recognise participation as unavoidable. They enable systems thinking that works with complexity rather than against it, that acknowledges limitations whilst maintaining analytical rigour and that integrates multiple forms of knowledge whilst sustaining critical inquiry.

Morin (2007, p. 22) writes that we must abandon the architectural metaphor of foundations in favour of a musical metaphor of construction in movement. This captures both the challenge and the possibility of epistemological transformation. We cannot construct new foundations from a position of certainty. Rather, we must proceed through ongoing reconstruction, acknowledging uncertainty whilst maintaining direction, engaging complexity whilst sustaining rigour and transforming practice whilst learning through practice.