# Complexity and Systems

*A density of the Field Guide to Psychological Safety · 40 articles and 69 papers · Tom Geraghty & Jade Garratt · revised 2026-10-25*

Complexity science, cybernetics, systems dynamics, ecology and safety science each use the word complex, and they converged rather than descended from one another: what each tradition actually said, which borrowed from which, and what that licenses for the way we work with organisations.

- **This page:** https://explore.psychsafety.com/densities/complexity/

## The landscape

Complexity is the word these traditions argue over, and they converged rather than descended from one another. Our [History of PS](https://explore.psychsafety.com/n/the-history-of-psychological-safety.md) makes that claim for psychological safety itself, convergence rather than descent, and the complexity literature wants the same reading: a mainline, a cybernetic valley beside it, systems dynamics, an ecological lineage with its own founding quarrel, safety science's systems turn and a management turn, each with its own vocabulary and its own cautions. Our way in is Tom's [Complexity](https://explore.psychsafety.com/n/complexity.md) newsletter: in large complex systems it is often impossible to isolate the effect of any single intervention, which can make impact feel unprovable and so discourage action, and the answer is not to pre-plan our way through but to probe, sense and respond, and to make small improvements to the organisational substrate knowing they made a difference to some.

**The mainline.** [Poincaré (1908)](https://explore.psychsafety.com/n/poincare-1908.md) is the headwater: deterministic laws under which prediction is nonetheless impossible, because differences in initial conditions too small to measure produce differences in outcome too large to ignore. [Weaver (1948)](https://explore.psychsafety.com/n/weaver-1948.md) named the object, organised complexity, the middle territory between problems of a few variables and problems of so many that averages suffice. [Simon (1962)](https://explore.psychsafety.com/n/simon-1962.md) supplied hierarchy in a sense wider than authority, and near-decomposability. [Anderson (1972)](https://explore.psychsafety.com/n/anderson-1972.md) accepted reductionism and refused its converse: at each level of complexity entirely new properties appear, and psychology is not applied biology; its worked cases are physics, and the bridge to organisations is ours. [Holland (1992)](https://explore.psychsafety.com/n/holland-1992.md) is where complex adaptive systems acquires its name and its canonical characterisation, from a computer scientist at Michigan and the Santa Fe Institute: parts that evolve, aggregate behaviour that emerges from interaction, and parts that carry internal models and anticipate, the property Holland singles out as separating the complex adaptive from the merely complicated. [Levin (1998)](https://explore.psychsafety.com/n/levin-1998.md) grounded the framework in ecosystems and tightened it: against the six-property checklists then fashionable, three mechanisms suffice, sustained diversity, localised interactions and an autonomous selection process, and the familiar six are their consequences. [Mitchell (2009)](https://explore.psychsafety.com/n/mitchell-2009.md), writing from inside the field, concludes that no single definition of complexity commands agreement, and treats that as an honest description of a young field rather than a defect.

**The adjacent valley.** Cybernetics is a distinct tradition, and the note on [Ashby (1956)](https://explore.psychsafety.com/n/ashby-1956.md) says so plainly: the book that proved the Law of Requisite Variety, only variety in the regulator can force down the variety of disturbance, is not an ancestor of complex adaptive systems thinking but part of a braided genealogy of converging resources, in which convergence from separate origins is arguably stronger evidence that an idea tracks something real than inheritance would be. [Wiener (1948)](https://explore.psychsafety.com/n/wiener-1948.md) named the field, control and communication treated as one subject in animal, machine or society. [Beer (1972)](https://explore.psychsafety.com/n/beer-1972.md) is one of the best used footpaths between the valleys, Ashby's law turned into variety engineering, a downstream application Ashby never made. [Gatlin (1972)](https://explore.psychsafety.com/n/gatlin-1972.md) is a second footpath almost nobody reading Juarrero or Cynefin has walked: Shannon information theory applied to DNA, and the source of the distinction between context-free constraints that bias outcomes and context-sensitive constraints that let a system do what no part could alone, taken up explicitly by [Juarrero (1999)](https://explore.psychsafety.com/n/juarrero-1999.md) and from Juarrero by Cynefin.

**Systems dynamics, planning and the state.** [Meadows (1999)](https://explore.psychsafety.com/n/meadows-1999.md) ranks twelve places to intervene in a system, from parameters, where almost all attention goes and there is very little leverage, up to the paradigm and the power to transcend it; this is systems dynamics rather than complexity science proper, a tradition that assumes the system can be surveyed and its structure mapped. [Repenning & Sterman (2001)](https://explore.psychsafety.com/n/repenning-sterman-2001.md) use the same tools on process improvement. [Rittel & Webber (1973)](https://explore.psychsafety.com/n/rittel-webber-1973.md) gave planning the wicked problem, with no definitive formulation, no stopping rule and no undo, so the planner has no right to be wrong. [Scott (1998)](https://explore.psychsafety.com/n/scott-1998.md) names the gaze that flattens, legibility and high modernism, with the German scientific forest, reduced to timber yield and dying a generation on, as the whole argument in miniature. Tom's articles on [queueing theory](https://explore.psychsafety.com/n/queueing-theory-slack-and-resilience.md) and the [Theory of Constraints](https://explore.psychsafety.com/n/the-theory-of-constraints.md) come from a different lineage again, Erlang's telephone exchange and Goldratt's The Goal.

**The ecological lineage.** Ecology arrives with its own founding quarrel. [Tansley (1935)](https://explore.psychsafety.com/n/tansley-1935.md) coined ecosystem against Clements's superorganism, a community with a maturity and a proper climax state, and the origin is a standing caution: the concept that licenses ecological thinking about organisations was coined specifically to stop people treating a collective as an organism with a health and a journey to maturity. [Gleason (1926)](https://explore.psychsafety.com/n/gleason-1926.md) removed the destiny from the other flank, communities as transient overlaps of independently behaving populations, and was sidelined for decades for saying so. [Holling (1973)](https://explore.psychsafety.com/n/holling-1973.md) split resilience from stability; [Gunderson & Holling (2002)](https://explore.psychsafety.com/n/gunderson-holling-2002.md) and [Holling (2001)](https://explore.psychsafety.com/n/holling-2001.md) turned that into the adaptive cycle and panarchy, cycles nested across scales and coupled by revolt and remember; [Walker et al. (2004)](https://explore.psychsafety.com/n/walker-et-al-2004.md) tightened the vocabulary into resilience, adaptability and transformability, and [Folke (2006)](https://explore.psychsafety.com/n/folke-2006.md) mapped how the perspective grew. Beneath them sit the mechanisms: [Paine (1966)](https://explore.psychsafety.com/n/paine-1966.md) on keystones, [May (1972)](https://explore.psychsafety.com/n/may-1972.md) on stability, [MacArthur & Wilson (1967)](https://explore.psychsafety.com/n/macarthur-wilson-1967.md) on islands and [Scheffer et al. (2009)](https://explore.psychsafety.com/n/scheffer-et-al-2009.md) on the wobble before a tipping point. Tom's articles on the [adaptive cycle](https://explore.psychsafety.com/n/the-adaptive-cycle-and-self-organised-criticality.md) and [five ecological concepts](https://explore.psychsafety.com/n/five-ecological-concepts-for-working-in-organisational-change.md) carry these into organisations, and [Leopold (1949)](https://explore.psychsafety.com/n/leopold-1949.md), thinking like a mountain, is the way of thinking whose name Tom's workshop overview [Thinking Like an Ecologist](https://explore.psychsafety.com/n/thinking-like-an-ecologist-a-field-guide-for-organisations.md) echoes.

**Substrate.** [Darwin (1881)](https://explore.psychsafety.com/n/darwin-1881.md) is the substrate argument in its founding form: the ground itself an artefact, deposited by worms passing a substantial fraction of England's topsoil through their bodies, invisible precisely because it is continuous. [Jones, Lawton & Shachak (1994)](https://explore.psychsafety.com/n/jones-lawton-shachak-1994.md) coined ecosystem engineers for organisms that create, modify or maintain habitat, with effects that persist beyond the engineer; [Laland, Matthews & Feldman (2016)](https://explore.psychsafety.com/n/laland-matthews-feldman-2016.md) add inheritance, descendants inheriting altered worlds and not only genes; and [van Breemen (1995)](https://explore.psychsafety.com/n/van-breemen-1995.md) argues that Sphagnum builds the acidic, nutrient-poor, cold and anoxic conditions in which few other plants can flourish. Tom's article on the [organisational substrate](https://explore.psychsafety.com/n/the-organisational-substrate.md) sets down the definition, the accumulated physical and non-physical matter an organisation lives in and on, built continuously by everything living in it, so that culture is what currently grows in the substrate rather than the substrate itself; the piece on [rewetting](https://explore.psychsafety.com/n/rewetting-organisations.md) is the restoration metaphor, and [Looking At Things Properly](https://explore.psychsafety.com/n/a-practical-guide-to-looking-at-things-properly.md) the reading practice, indicator species, careful to distinguish itself from the organisational ecology of [Hannan & Freeman (1977)](https://explore.psychsafety.com/n/hannan-freeman-1977.md), which studies populations of firms under environmental selection. [Mars & Bronstein (2023)](https://explore.psychsafety.com/n/mars-bronstein-2023.md) test borrowings of this kind, and on their argument the substrate borrowing is sound when the concept is articulated against its ecological source rather than taken literally.

**Safety science's systems turn.** [Perrow (1984)](https://explore.psychsafety.com/n/perrow-1984.md) built normal accident theory on two properties of the system, interactive complexity and tight coupling; [La Porte & Consolini (1991)](https://explore.psychsafety.com/n/laporte-consolini-1991.md) answered with high reliability organisations, in which authority migrates with tempo and the person closest to the problem decides; [Rasmussen (1997)](https://explore.psychsafety.com/n/rasmussen-1997.md) reframed safety as dynamic control, in a vocabulary drawn from Gibson's ecological psychology; [Leveson (2004)](https://explore.psychsafety.com/n/leveson-2004.md) made safety an emergent property of the whole control structure; [Reiman et al (2015)](https://explore.psychsafety.com/n/reiman-et-al-2015.md) treat safety in high-hazard organisations as an emergent property of a complex adaptive system; [Shorrock et al (2014)](https://explore.psychsafety.com/n/shorrock-et-al-2014.md) distil systems thinking into ten principles; and [Woods (2018)](https://explore.psychsafety.com/n/woods-2018.md) sits between Holling's ecological resilience and everyday organisational life, resilience as a verb. [Read et al (2021)](https://explore.psychsafety.com/n/read-et-al-2021.md) trace the sixty-year migration from blaming the person to interrogating the system, and [ICAO (2026)](https://explore.psychsafety.com/n/icao-2026-srm-methods.md) audits an industry's own methods against Rasmussen's tenets, the risk matrix answering no to five of seven. Tom's articles on [normal accidents](https://explore.psychsafety.com/n/normal-accidents.md) and [resilience engineering](https://explore.psychsafety.com/n/psychological-safety-resilience-engineering.md), and Nick Drage's [Safe-to-Fail Wargames](https://explore.psychsafety.com/n/the-importance-of-safe-to-fail-wargames.md), carry this strand.

**Sense-making and the management turn.** [Stacey (1995)](https://explore.psychsafety.com/n/stacey-1995.md) brought complexity science to strategic management, strategy emerging from the self-organising interaction of people rather than from the plan, argued largely by analogy from the natural sciences. [Kurtz & Snowden (2003)](https://explore.psychsafety.com/n/kurtz-snowden-2003.md) introduced Cynefin in full, a sense-making framework rather than a categorisation model, grounded in action research and consultancy; [Snowden & Boone (2007)](https://explore.psychsafety.com/n/snowden-boone-2007.md) made it widely known, and [Pelrine (2011)](https://explore.psychsafety.com/n/pelrine-2011.md) read Agile as complexity management in disguise. [Morin (2007)](https://explore.psychsafety.com/n/morin-2007.md) names the argument the field is having, restricted complexity, the Santa Fe programme in which the observer stays outside and the search for law continues, against general complexity, in which complexity reaches back and damages the way we know; our own commitments sit on the general side, with [Cilliers (1998)](https://explore.psychsafety.com/n/cilliers-1998.md), [Juarrero (1999)](https://explore.psychsafety.com/n/juarrero-1999.md) and [Preiser et al (2018)](https://explore.psychsafety.com/n/preiser-et-al-2018.md). [Page (2015)](https://explore.psychsafety.com/n/page-2015.md) is the rigorous friend from the modelling side, and [Centola et al (2018)](https://explore.psychsafety.com/n/centola-et-al-2018.md) took the tipping point out of metaphor and into a controlled experiment.

## The key connections

**Emergence to substrate.** On our reading, four traditions arrive at the same place. We read [Anderson (1972)](https://explore.psychsafety.com/n/anderson-1972.md) as a warrant for refusing to treat a team or a culture as the sum of its members' attributes, and our [Core Principles](https://explore.psychsafety.com/n/our-core-principles.md) frame psychological safety as an emergent property of conditions, not a programme, a metric or an individual attribute, shaped by structural power and the substrate of norms and history that precede any interaction. [Levin (1998)](https://explore.psychsafety.com/n/levin-1998.md) supplies mechanism: aggregation and hierarchy emerge from local interaction and, once formed, constrain what follows, and nonlinearity produces the path dependency by which a system's history is written into its present. [Juarrero (1999)](https://explore.psychsafety.com/n/juarrero-1999.md) gives it a grammar, enabling constraints that create possibility rather than merely limiting it, which is precisely what a psychologically safe environment is, not an absence of structure but a structure that makes speaking possible; [Juarrero (2023)](https://explore.psychsafety.com/n/juarrero-2023.md) has coherence induced by enabling constraints rather than forceful impacts, which is the substrate argument in philosophical form. [Hogenkamp et al. (2026)](https://explore.psychsafety.com/n/hogenkamp-et-al-2026.md) move the target of intervention from the individual to the transaction, because the same trait reads differently in different settings, which we read as the substrate argument arriving from developmental science.

**Variety, noise and voice.** The cybernetic valley gives the voice problem a formal floor, by readings downstream of the sources rather than their own. Read through [Ashby (1956)](https://explore.psychsafety.com/n/ashby-1956.md), a narrow and homogeneous group of decision-makers cannot regulate a varied organisation, not because that is unfair but because it is arithmetically impossible, and an organisation whose only sanctioned channel is the leader has a regulatory variety of one. Read through [Wiener (1948)](https://explore.psychsafety.com/n/wiener-1948.md), the things people decline to say upward are not an absence of signal but noise in the channel, and a system full of noise cannot hold its shape. [Beer (1972)](https://explore.psychsafety.com/n/beer-1972.md) makes the organisational reading in the book itself: the algedonic channel, an alarm that bypasses the hierarchy when routine reporting fails, which we read as psychological safety's central problem stated as an engineering requirement in 1972. [Leveson (2004)](https://explore.psychsafety.com/n/leveson-2004.md) says the same in control terms, a structure whose upward feedback is suppressed by fear has already failed, and Tom's articles on the [OODA loop](https://explore.psychsafety.com/n/john-boyd-and-the-ooda-loop.md), distorted inputs corrupting the Orient step, and the [Theory of Constraints](https://explore.psychsafety.com/n/the-theory-of-constraints.md), information flow as the bottleneck, carry it into practice. [Meadows (1999)](https://explore.psychsafety.com/n/meadows-1999.md) puts information flows and rules high among the leverage points, and [Conway's Law](https://explore.psychsafety.com/n/psychological-safety-conways-law.md) says the communication structure gets built into whatever the organisation makes.

**Efficiency against resilience, in every tradition.** [Perrow (1984)](https://explore.psychsafety.com/n/perrow-1984.md): the losses in slack, redundancy and generalist understanding that would make a system safer are the losses that make it worth building. [Holling (2001)](https://explore.psychsafety.com/n/holling-2001.md): efficiency, connectedness and accumulated capital rise together through the front loop and make the system progressively more rigid and brittle, an accident waiting to happen. [Levin (1998)](https://explore.psychsafety.com/n/levin-1998.md): heavily managed systems with simplified structures imposed from outside are not really complex adaptive systems at all, which is exactly what makes them fragile. [Woods (2018)](https://explore.psychsafety.com/n/woods-2018.md): adaptive capacity is routinely traded away in the name of efficiency. [Hannan & Freeman (1977)](https://explore.psychsafety.com/n/hannan-freeman-1977.md): generalism means holding excess capacity, which outsiders will often mistake for waste. [Patterson & Wears (2015)](https://explore.psychsafety.com/n/patterson-wears-2015.md) show the trade in progress, a pharmacist's fluent adaptations making a brittle system look fine from a distance. Tom's articles on [efficiency versus resilience](https://explore.psychsafety.com/n/efficiency-vs-resilience.md), the Irish Potato Famine and Nokia read through Ashby, and on [queueing theory](https://explore.psychsafety.com/n/queueing-theory-slack-and-resilience.md), a wait-time curve that stays flat and then turns sharply upward around 80 per cent utilisation, with slack not idle time as long as what fills it is interruptible, make the organisational case.

**Thresholds, drift and the signal in the variance.** [Holling (1973)](https://explore.psychsafety.com/n/holling-1973.md) names domains of attraction separated by thresholds, and says the important point is not how stable a system is within its domain but how likely it is to move into another. [Scheffer et al. (2009)](https://explore.psychsafety.com/n/scheffer-et-al-2009.md) say how you might tell: critical slowing down, rising variance and autocorrelation, so the information is in the variance rather than the mean, and a dashboard showing a steady average can be accurate and uninformative right up to the collapse. [Rasmussen (1997)](https://explore.psychsafety.com/n/rasmussen-1997.md) describes the drift towards the boundary, invisible because defences are redundant. [Paine (1966)](https://explore.psychsafety.com/n/paine-1966.md) shows removal as regime shift rather than subtraction, [Centola et al (2018)](https://explore.psychsafety.com/n/centola-et-al-2018.md) a convention flipping once a committed minority crosses roughly a quarter of the group, [Bak (1996)](https://explore.psychsafety.com/n/bak-1996.md) avalanches of all sizes from the same slowly driven dynamics, and Tom's article on [safety-organised criticality](https://explore.psychsafety.com/n/safety-organised-criticality-socy.md) carries that to organisational failure. [Kelso (2009)](https://explore.psychsafety.com/n/kelso-2009.md) is the hard case, a signal present in all fourteen daily reports of the week before and structurally indistinguishable from the thousands like it, and [Pauly (1995)](https://explore.psychsafety.com/n/pauly-1995.md) the slow one, each generation of fisheries scientists accepting the depleted stock it found as the baseline.

**Probe, do not plan.** [Holland (1992)](https://explore.psychsafety.com/n/holland-1992.md) grounds the probe-and-adapt stance: complex adaptive systems never reach a stable optimum, so change is continual adaptation rather than the execution of a plan. [Kurtz & Snowden (2003)](https://explore.psychsafety.com/n/kurtz-snowden-2003.md) make it a method, probe-sense-respond in the complex domain, and warn that codifying yesterday's emergent pattern into tomorrow's procedure guarantees meeting a pattern it cannot handle; [Pelrine (2011)](https://explore.psychsafety.com/n/pelrine-2011.md) finds the same loop at the heart of Scrum, and from [Rittel & Webber (1973)](https://explore.psychsafety.com/n/rittel-webber-1973.md) we draw the reason it needs voice: if you cannot get the answer right in advance, you need people able to say when it is going wrong. Tom's articles on [safe-to-fail experiments](https://explore.psychsafety.com/n/safe-to-fail.md), [experiments, bets and probes](https://explore.psychsafety.com/n/experiments-bets-and-probes.md) and [everything as an experiment](https://explore.psychsafety.com/n/everything-is-an-experiment.md) build the vocabulary, [Guardrails & Failure](https://explore.psychsafety.com/n/psychological-safety-84-guardrails-and-failure.md) bounds it, since not everything is safe to fail, and [Gersick (1988)](https://explore.psychsafety.com/n/gersick-1988.md) supplies a timing: a team's framework is set implicitly in its first meeting, and the midpoint is the one moment this evidence says it is reliably open to revising it.

## The tensions and disagreements

**Is an organisation a complex adaptive system?** The traditions divide, and so do we. [Holland (1992)](https://explore.psychsafety.com/n/holland-1992.md) is the source text for the now-routine claim, and it has to be read critically: the account is thoroughly computational, built from rule-following agents, and its transfer to organisations can quietly re-mechanise the very thing it means to de-mechanise. [Juarrero (1999)](https://explore.psychsafety.com/n/juarrero-1999.md) is the reason a human complex system is not merely a complicated one: agents with intentions are constituted by constraints and can change them, so the rule-following agent is the wrong model for a team. The usual caution on [Levin (1998)](https://explore.psychsafety.com/n/levin-1998.md) is that human organisations are not literally under natural selection, so the transfer of keystones and autonomous selection to a world of intention, power and culture is suggestive rather than exact; [Hannan & Freeman (1977)](https://explore.psychsafety.com/n/hannan-freeman-1977.md) insist, from the other side, that their models are analogues, not metaphors; [Mars & Bronstein (2023)](https://explore.psychsafety.com/n/mars-bronstein-2023.md) caution that human intention sets humans apart from species whose interactions emerge through natural selection. [Morgan (2006)](https://explore.psychsafety.com/n/morgan-2006.md) holds that all theory is metaphor, which illuminates and hides, and our [History of PS](https://explore.psychsafety.com/n/the-history-of-psychological-safety.md) says we use ecological thinking not as metaphor but as analytical lens, a stance in tension with Morgan's premise. Tom's articles on [sociotechnical theory](https://explore.psychsafety.com/n/sociotechnical-theory.md) and [resilience engineering](https://explore.psychsafety.com/n/psychological-safety-resilience-engineering.md) take the strong position, organisations as complex adaptive systems, and [Page (2015)](https://explore.psychsafety.com/n/page-2015.md) is the discipline on it: complexity is a specific set of features that demand different models and different evidence, not a vague gesture at interconnectedness.

**Restricted or general.** [Morin (2007)](https://explore.psychsafety.com/n/morin-2007.md) draws the line, and most arguments in the complexity literature, over whether complexity can be tidily defined, whether cybernetics belongs in the ancestry, whether the epistemological turn is rigour or drift, are two people standing either side of it without naming it. The restricted side is not a straw man. [Holling (2001)](https://explore.psychsafety.com/n/holling-2001.md) rejects the view that complexity is simply what exceeds understanding and argues for requisite simplicity, living systems structured by a small number of controlling variables; [Page (2015)](https://explore.psychsafety.com/n/page-2015.md) wants formal models and computational methods; [Meadows (1999)](https://explore.psychsafety.com/n/meadows-1999.md) assumes the system can be surveyed, which [Cilliers (1998)](https://explore.psychsafety.com/n/cilliers-1998.md) says a complex system will not permit, in a position commentators have called radically different from Holland's rule-based route to emergence; and [Levin (1998)](https://explore.psychsafety.com/n/levin-1998.md) shows what restricted rigour is for, dissolving the Gaia superorganism through established mechanisms rather than appeal to whole-system processes. [Greenhalgh & Papoutsi (2018)](https://explore.psychsafety.com/n/greenhalgh-papoutsi-2018.md) carry the general claim into method: the honest study of a complex human system will look messier, more contextual and more theory-laden than the tidy trial its critics demand.

**Can safety be engineered into a complex system?** [Perrow (1984)](https://explore.psychsafety.com/n/perrow-1984.md) says the accidents are normal; read against Perrow, [La Porte & Consolini (1991)](https://explore.psychsafety.com/n/laporte-consolini-1991.md) frame the central dispute in safety science, whether human agency, culture, design, training and migrating authority, can compensate for the structural pressures of complexity and tight coupling, and present what they saw as anomalous data rather than a finished theory. [Leveson (2004)](https://explore.psychsafety.com/n/leveson-2004.md) treats safety as a control problem and models the whole control structure, an engineering formalism that sits awkwardly with interpretive accounts of culture and power; [Beer (1972)](https://explore.psychsafety.com/n/beer-1972.md) engineers regulation for viability where the complexity tradition doubts the system can be specified. [Reiman et al (2015)](https://explore.psychsafety.com/n/reiman-et-al-2015.md) split the difference, compliance suited to the ordered parts of the work and counterproductive for the complex parts, and [Rasmussen (1997)](https://explore.psychsafety.com/n/rasmussen-1997.md) prescribes not fighting deviations but making the boundary of safe operation visible and helping people work near it. [Stachowski, Kaplan & Waller (2009)](https://explore.psychsafety.com/n/stachowski-kaplan-waller-2009.md) complicate the comfortable reading from the other end: the control-room crews that handled a simulated crisis best were less inclusive and more directive, and used the protocol as a tool rather than letting it drive them.

**Resilience of what, for whom.** [Holling (1973)](https://explore.psychsafety.com/n/holling-1973.md) argues for managing for resilience, and the caution is that the definition is descriptive, so an undesired state can persist too; [Folke (2006)](https://explore.psychsafety.com/n/folke-2006.md) is explicit that resilience is not inherently good, the lock-in trap, so the question is always resilience of what, to what, and for whom, and records that the social-ecological framework has been criticised for under-theorising power and politics. [Holling (2001)](https://explore.psychsafety.com/n/holling-2001.md) names the rigidity trap, a perverse resilience that entrenches a maladaptive system because novelty is smothered and its inventors ejected. The individual bounce-back that management and wellbeing writing calls resilience is, on the readings of both [Holling (1973)](https://explore.psychsafety.com/n/holling-1973.md) and [Folke (2006)](https://explore.psychsafety.com/n/folke-2006.md), engineering resilience wearing the ecological vocabulary, closer to what Holling called stability, and it can move the burden of a brittle system onto the people inside it. We read [Hogenkamp et al. (2026)](https://explore.psychsafety.com/n/hogenkamp-et-al-2026.md) as the same challenge from neurodiversity, masking as high-cost regulation and the least safe members bearing burdens the environment created, and our [Core Principles](https://explore.psychsafety.com/n/our-core-principles.md) put power and its unequal distribution at the heart of the work.

**How much connection, and of what kind.** [May (1972)](https://explore.psychsafety.com/n/may-1972.md) shows a large randomly connected system almost certainly unstable once complexity exceeds a critical value, with examples favouring interactions arranged in blocks, a corrective to the idea that more connection is always better; [Simon (1962)](https://explore.psychsafety.com/n/simon-1962.md) makes a neighbouring point, near-decomposability as what makes a complex system analysable at all; [Perrow (1984)](https://explore.psychsafety.com/n/perrow-1984.md) makes coupling the danger, and Tom's article on [normal accidents](https://explore.psychsafety.com/n/normal-accidents.md) notes that measures intended to improve resilience can add complexity and therefore risk. Against them, read organisationally, [MacArthur & Wilson (1967)](https://explore.psychsafety.com/n/macarthur-wilson-1967.md) give an isolated team receiving fewer new practices and losing the ones it has faster, a property of position rather than merit, and Tom's articles on [ecotones](https://explore.psychsafety.com/n/psychological-safety-59-ecotones-the-edge-effect.md) and the [organisational fabric](https://explore.psychsafety.com/n/the-organisational-fabric-of-psychological-safety.md) argue that the richest exchange happens at the interfaces and that isolated high-performing bubbles cannot share. [Dunbar's number](https://explore.psychsafety.com/n/psychological-safety-82-dunbars-number-and-team-size.md) and [Conway's Law](https://explore.psychsafety.com/n/psychological-safety-conways-law.md) are the design constraints in between. The terms do some of the reconciling: May's result turns on the number, strength and arrangement of interactions, MacArthur and Wilson's on variety, which [Ashby (1956)](https://explore.psychsafety.com/n/ashby-1956.md) says is not merely a shame to lose.

**Stage-shaped ideas, and oversold ones.** [Gunderson & Holling (2002)](https://explore.psychsafety.com/n/gunderson-holling-2002.md): the adaptive cycle is a heuristic that organises observation rather than a validated law, applied loosely enough in management writing to explain any outcome after the fact, which is the failure mode of every stage-shaped idea; [Tansley (1935)](https://explore.psychsafety.com/n/tansley-1935.md) and [Gleason (1926)](https://explore.psychsafety.com/n/gleason-1926.md) remove the licence for maturity journeys at the source; [Gersick (1988)](https://explore.psychsafety.com/n/gersick-1988.md) broke the stage model of group development with eight real teams; and Tom's [Six-Stage Critique](https://explore.psychsafety.com/n/the-four-stages-of-psychological-safety-a-six-stage-critique.md) turns this on the Four Stages, borrowing Snowden's reading of Box, all models wrong but not all in the same way, to argue that a stage model applied to a complex phenomenon manufactures a false sense of control. [Scheffer et al. (2009)](https://explore.psychsafety.com/n/scheffer-et-al-2009.md) and [Bak (1996)](https://explore.psychsafety.com/n/bak-1996.md) show the other way a complexity idea can be oversold: early warning signals are contested and can fail entirely, Bak's models are deliberately oversimplified toys whose reach rests on universality, which Bak calls a hope, and [Levin (1998)](https://explore.psychsafety.com/n/levin-1998.md) flagged the self-organised-criticality and edge-of-chaos claims as stimulating but still without substantial empirical support, a caution that bears on Tom's article on [safety-organised criticality](https://explore.psychsafety.com/n/safety-organised-criticality-socy.md).

## What this means for practice

- **Read the habitat before the indicator.** [Looking At Things Properly](https://explore.psychsafety.com/n/a-practical-guide-to-looking-at-things-properly.md): context precedes reading, succession governs interpretation, presence is not establishment, and once the indicator becomes the target it stops being an indicator. [Scheffer et al. (2009)](https://explore.psychsafety.com/n/scheffer-et-al-2009.md) say to watch the variance and the recovery time rather than the mean, [Pauly (1995)](https://explore.psychsafety.com/n/pauly-1995.md), read our way, to distrust a baseline set on the day you arrived, and [Carson (1962)](https://explore.psychsafety.com/n/carson-1962.md) that the signal may be an absence, the birds not singing.
- **Shape the substrate, not the outcome.** Tom's article on the [organisational substrate](https://explore.psychsafety.com/n/the-organisational-substrate.md): before the new practice, the redrawn org chart or the rebranded values, ask what ground they are expected to thrive in, since culture-change programmes addressing only culture have the approximate effectiveness of shouting at a lawn. By [Scott's (1998)](https://explore.psychsafety.com/n/scott-1998.md) lights, installing safety as a programme scored with one legible number is high modernism in the small; the piece on [rewetting](https://explore.psychsafety.com/n/rewetting-organisations.md) is the posture, removing the structures that suppress voice and letting the system self-organise; and [Be soft to each other, and hard on the system](https://explore.psychsafety.com/n/be-soft-to-each-other-and-hard-on-the-system.md) is the sticker.
- **Keep slack and variety, and expect to be told they are waste.** [Hannan & Freeman (1977)](https://explore.psychsafety.com/n/hannan-freeman-1977.md) say the capacity that saves you later is mistaken for waste now, and [Perrow (1984)](https://explore.psychsafety.com/n/perrow-1984.md) that it is the loss of it that makes the system worth building. Tom's article on [queueing theory](https://explore.psychsafety.com/n/queueing-theory-slack-and-resilience.md) bounds the claim: in stable, low-variance environments running near full utilisation carries little cost, and the real danger is believing an environment is stable when it is not.
- **Act by probes, with guardrails.** Tom's articles on [safe-to-fail experiments](https://explore.psychsafety.com/n/safe-to-fail.md) and [experiments, bets and probes](https://explore.psychsafety.com/n/experiments-bets-and-probes.md) give the vocabulary and insist it matters, since the frame shapes how risk is assessed and failure understood; [Guardrails & Failure](https://explore.psychsafety.com/n/psychological-safety-84-guardrails-and-failure.md) says not everything is safe to fail and that guardrails reduce the stakes of speaking up by containing consequences; [Kurtz & Snowden (2003)](https://explore.psychsafety.com/n/kurtz-snowden-2003.md) warn that in sense-making sessions large power differences have to be dismantled first or people simply watch what the boss does; and [Safe-to-Fail Wargames](https://explore.psychsafety.com/n/the-importance-of-safe-to-fail-wargames.md) rehearse under pressure where that costs less than learning from real failure.
- **Name the tradition you are borrowing from.** [Mars & Bronstein (2023)](https://explore.psychsafety.com/n/mars-bronstein-2023.md): rigour comes from tracing a metaphor back to its source, with collaborators from that discipline. The caution on [Ashby (1956)](https://explore.psychsafety.com/n/ashby-1956.md): variety is a precise term whose popular management use drifts into a loose synonym for diversity of background, which argues something quite different. [Morgan (2006)](https://explore.psychsafety.com/n/morgan-2006.md): every theory illuminates and hides. [Mitchell (2009)](https://explore.psychsafety.com/n/mitchell-2009.md): no single definition of complexity commands agreement, so meet confident claims about what complexity really is with that posture. Tom's [Critique of Personality Profiling](https://explore.psychsafety.com/n/the-fallacy-of-applying-complicated-models-to-complex-problems-aka-why-personality-profiling-is-bs.md) is what it looks like when a complicated framework is mapped onto a complex problem.

## What the map does not yet hold

The mainline is thin beyond its anchors. [Holland (1992)](https://explore.psychsafety.com/n/holland-1992.md) and [Mitchell (2009)](https://explore.psychsafety.com/n/mitchell-2009.md) are the Santa Fe voices with nodes of their own; Arthur is cited only inside [Levin (1998)](https://explore.psychsafety.com/n/levin-1998.md), as is Kauffman, Lorenz only inside [Poincaré (1908)](https://explore.psychsafety.com/n/poincare-1908.md), and Gell-Mann and Prigogine not at all. The network science and machine learning developments of the decade after 2009, which Mitchell's survey stops before, have no node; [PS & Artificial Intelligence](https://explore.psychsafety.com/n/psychological-safety-90-artificial-intelligence.md) is about AI in the team, not about complexity science. On the systems side, [Meadows (1999)](https://explore.psychsafety.com/n/meadows-1999.md) and [Repenning & Sterman (2001)](https://explore.psychsafety.com/n/repenning-sterman-2001.md) carry systems dynamics alone: Forrester, von Bertalanffy's general systems theory and Checkland's soft systems are absent, and Senge appears once, in passing, in [Learning Methods & Orgs](https://explore.psychsafety.com/n/psychological-safety-49-learning-methods-learning-organisations.md). Stacey is held at [1995](https://explore.psychsafety.com/n/stacey-1995.md) and in the matrix that Tom's [Complexity](https://explore.psychsafety.com/n/complexity.md) newsletter draws on; nothing covers Stacey's later work. Complexity leadership theory is absent.

Two of our own articles use a name that the work they describe did not have. Tom's article on [sociotechnical theory](https://explore.psychsafety.com/n/sociotechnical-theory.md) presents the Tavistock work of the 1950s as the framework for understanding organisations as complex adaptive systems, a name the idea acquired in [Holland (1992)](https://explore.psychsafety.com/n/holland-1992.md); the founding paper, [Trist & Bamforth (1951)](https://explore.psychsafety.com/n/trist-bamforth-1951.md), sits outside this region and speaks of a social substrate and of an emergent property of the relationship between work design and people, not of complex adaptive systems. Tom's article on [resilience engineering](https://explore.psychsafety.com/n/psychological-safety-resilience-engineering.md) describes a field focused on how complex adaptive systems cope with surprise; [Woods (2018)](https://explore.psychsafety.com/n/woods-2018.md) is the one node here that speaks for that field by name, and the field's founding texts are not nodes. Hollnagel's FRAM appears only inside [ICAO (2026)](https://explore.psychsafety.com/n/icao-2026-srm-methods.md). Drift into failure is named in [Rasmussen (1997)](https://explore.psychsafety.com/n/rasmussen-1997.md), which credits Dekker with developing it, and in [Schröder-Hinrichs et al (2012)](https://explore.psychsafety.com/n/schroder-hinrichs-et-al-2012.md), which credits Rasmussen, but has no node of its own; the two descs disagree. Loose coupling in the organisational sense is named once, unattributed, in [Simon (1962)](https://explore.psychsafety.com/n/simon-1962.md), and [Woods et al. (1994)](https://explore.psychsafety.com/n/woods-et-al-1994.md) use the phrase in another sense; Weick's nodes on the map are on sensemaking, collective mind and high reliability, not on coupling.

## Explore the density

### Start here

1. [Science and Complexity](https://explore.psychsafety.com/n/weaver-1948.md) (Weaver, 1948)
2. [The Architecture of Complexity](https://explore.psychsafety.com/n/simon-1962.md) (Simon, 1962)
3. [More Is Different](https://explore.psychsafety.com/n/anderson-1972.md) (Anderson, 1972)
4. [Complex Adaptive Systems](https://explore.psychsafety.com/n/holland-1992.md) (Holland, 1992)
5. [Ecosystems and the Biosphere as Complex Adaptive Systems](https://explore.psychsafety.com/n/levin-1998.md) (Levin, 1998)
6. [An Introduction to Cybernetics](https://explore.psychsafety.com/n/ashby-1956.md) (Ashby, 1956)
7. [Complexity and Postmodernism: Understanding Complex Systems](https://explore.psychsafety.com/n/cilliers-1998.md) (Cilliers, 1998)
8. [Dynamics in Action: Intentional Behavior as a Complex System](https://explore.psychsafety.com/n/juarrero-1999.md) (Juarrero, 1999)
9. [Restricted Complexity, General Complexity](https://explore.psychsafety.com/n/morin-2007.md) (Morin, 2007)
10. [The New Dynamics of Strategy: Sense-Making in a Complex and Complicated World](https://explore.psychsafety.com/n/kurtz-snowden-2003.md) (Kurtz & Snowden, 2003)
11. [Resilience and Stability of Ecological Systems](https://explore.psychsafety.com/n/holling-1973.md) (Holling, 1973)
12. [Panarchy: Understanding Transformations in Human and Natural Systems](https://explore.psychsafety.com/n/gunderson-holling-2002.md) (Gunderson, 2002)
13. [Normal Accidents: Living with High-Risk Technologies](https://explore.psychsafety.com/n/perrow-1984.md) (Perrow, 1984)
14. [The Organisational Substrate](https://explore.psychsafety.com/n/the-organisational-substrate.md)
15. [Complexity](https://explore.psychsafety.com/n/complexity.md)

### Trails through this region

- [Complexity: a long and technical ascent](https://explore.psychsafety.com/trails/complexity-descent/) (17 stops)
- [Thinking Like an Ecologist — Workshop Companion](https://explore.psychsafety.com/trails/thinking-like-an-ecologist/) (8 stops)

### Everything in the region

- [Complexity & Systems](https://explore.psychsafety.com/themes/complexity/)
- [Ecological Thinking](https://explore.psychsafety.com/themes/ecological/)
- [Complexity & Systems](https://explore.psychsafety.com/themes/complexity/)
- [Ecological Thinking](https://explore.psychsafety.com/themes/ecological/)

### On the interactive map

- https://explore.psychsafety.com/?theme=complexity
- https://explore.psychsafety.com/?theme=ecological
- https://explore.psychsafety.com/?mode=papers&theme=complexity
- https://explore.psychsafety.com/?mode=papers&theme=ecological