Sometimes, only sometimes, tweeting helps you think.
Over the past few months, I have been trying to make sense of the language of complexity used in contemporary analysis. Read a few paragraphs, and you inevitably run into 'emergence', 'initial conditions', 'contingency' and several adjacent distinctions pointing to a certain formulation of reality - often emphatically explained, but tentatively questioned.
Emergence and contingency represent fundamentally different ways of understanding how things come to be the way they are. The distinction between these two concepts has long structured how we think about complex systems and historical change.
Emergence describes how higher-order patterns reliably arise from lower-level interactions, seemingly inevitable once certain conditions are met. Consider how birds flock: no individual bird is in charge, yet complex, coordinated movements emerge from each bird following simple rules about maintaining distance from its neighbors. Or think about how consciousness emerges from billions of neurons firing in patterns - while we don't fully understand it, there's something seemingly inevitable about how awareness arises from sufficient neural complexity. Markets show similar emergent properties: the way prices settle at an equilibrium emerges from countless individual buying and selling decisions, without any central coordination.
Contingency, by contrast, captures how specific historical accidents shape outcomes in ways that could easily have been different. The classic example is the QWERTY keyboard layout - originally designed to prevent typewriter keys from jamming, it persists in our digital age not because it's optimal, but because that's what people learned to use. Or the asteroid strike 66 million years ago which led to the extinction of dinosaurs, creating the opportunity for mammals to dominate - a completely contingent event that shaped the entire course of evolution.
But what if this distinction, while seemingly clarifying, actually obscures more than it reveals? What if separating phenomena into "emergent" versus "contingent" categories prevents us from understanding how change actually happens through time?
The Traditional View
Across different academic disciplines, emergence and contingency are treated as distinct and complementary ways of understanding change. Evolutionary biologists point to emergence in how bilateral symmetry develops reliably in vertebrates, while attributing the rise of mammals to contingent evolutionary outcomes. Sociologists treat status hierarchies as emergent properties of human groups, while viewing particular forms of social organization as contingent outcomes of historical processes. Economists describe market prices as emerging from aggregate behavior, while treating the dominance of particular companies as contingent on historical accidents.
This conceptual division runs deeper than mere academic categorization. It shapes practical problems across domains. Urban theorists treat density and land prices as emergent properties of population growth, while viewing a city's particular form as contingent on historical decisions about zoning and infrastructure. Innovation scholars see certain patterns of technological improvement as emerging from market competition, while treating the dominance of particular technological standards or industrial locations as contingent on policy choices and historical accidents.
The separation of phenomena into "emergent" versus "contingent" creates two distinct modes of analysis and intervention. For emergent systems, analysis focuses on understanding underlying rules and mechanisms - like studying how urban density patterns arise from economic activity. For contingent phenomena, the emphasis falls on historical analysis and path dependency - like examining how early infrastructure decisions shape a city's later development possibilities.
This separation proves useful in certain contexts. When dealing with truly emergent phenomena like market pricing or agglomeration economies, focusing on system dynamics helps identify leverage points for intervention. Similarly, recognizing contingency in areas like industrial development helps understand how specific choices create lasting consequences.
However, I think this binary framing becomes problematic in three ways:
First, it creates artificial boundaries around complex phenomena. Consider semiconductor manufacturing: Is the concentration of production in certain regions emergent from scale economies and technological learning curves, or contingent on specific policy decisions and investments? Reality involves both, yet our analytical frameworks push us to study these separately.
Second, it shapes our sense of what's possible. Labeling something as "emergent" often implies inevitability, potentially blinding us to intervention opportunities. On the other hand, viewing something as purely "contingent" might cause us to underestimate systematic constraints on change. Current debates about nationalizing semiconductor production illustrate this tension - those who see existing patterns as inevitable and those who see them as purely policy-driven - both - miss crucial aspects of the challenge.
Third, and most importantly, this distinction breaks down at crucial moments of change. Major transitions typically involve both emergent dynamics and contingent choices interacting in complex ways.
Dissolving the Distinctions
Both emergence and contingency deal with how complex systems resist reduction to simple causal chains. Both wrestle with unpredictability and the limits of determinism. Both ultimately rest on the "cascade problem" in causation - the fact that every state of affairs emerges from an effectively infinite causal web extending backward in time.
When we label something "emergent," we are artificially bounding a system to study how certain properties arise from defined lower-level interactions. When we label something "contingent," we are artificially bounding historical sequences to study how certain events shaped outcomes. (I say 'label' because these are first and foremost speech acts). In both cases, these boundaries are arbitrary - there's no objective way to determine where an emergent system's boundaries lie or which historical factors count as contingent.
Consider how this plays out in China's rapid urbanization: Do we bound the system at the level of economic development (emergence of urban clusters following universal patterns of industrialization) or at the level of specific policy decisions about hukou reform and local government financing? When examining the rise of the Chinese megacities, the boundary we draw determines whether we perceive "inevitable" urbanization patterns or "contingent" policy choices.
Similarly, in semiconductor trade controls, where do we draw the system boundary? We can examine the emergent properties of global supply chains and technological interdependence, or we can focus on contingent decisions about export controls and industrial policy. Is TSMC's dominance in advanced chip manufacturing an emergent property of scale economies and network effects, or a contingent result of specific policy choices and historical timing? The answer depends entirely the boundaries we draw.
Also, both concepts rely on counterfactual thinking that is philosophically problematic. Emergence requires us to imagine systems without their emergent properties to understand how those properties arise. Contingency requires us to imagine alternative historical paths to understand how actual paths matter. These counterfactuals are mental and linguistic constructs - the actual universe simply proceeds according to its nature.
The 'Everything is Nothing' Problem
However, dissolving the emergence-contingency distinction is definitely problematic.
If we're analysing China's urbanization patterns and if we can't separate the emergent forces of agglomeration economies from contingent policy choices about land use and infrastructure, how do we develop testable hypotheses about urban development? Viewing everything as part of the same temporal process risks losing our ability to isolate variables and understand causal relationships.
Or if we can't distinguish between market-driven emergence and policy-driven contingency in semiconductor supply chains, how do we evaluate the effectiveness of industrial policy? When outcomes emerge from complex interactions of systemic and specific factors, learning from policy successes and failures becomes increasingly difficult.
The attempt to transcend the emergence-contingency distinction risks losing crucial analytical tools for both research and policy-making. Without some way to separate systematic forces from specific interventions, it becomes almost impossible to a) design controlled studies, b) evaluate the impact of specific policies, c) compare outcomes across different contexts, d) develop evidence-based policy recommendations.
These aren't merely academic concerns. They speak to fundamental questions about how we develop knowledge of complex systems and use that knowledge to inform practical interventions. Dissolving the distinctions cannot fully solve the measurement problem (isolating variables to measure casual effect), or the counterfactual problem (much of social science relies on asking what could have happened under different conditions).
From Prediction to Practice
But let me try and make a case for why we should move beyond prediction and control frameworks. Current policy analyses employ sophisticated models accounting for both market dynamics and policy interventions. They recognize multiple causal factors: scale economies, network effects, learning curves, policy incentives, geopolitical choices, etc. Yet these analyses still ultimately try to separate "natural" economic forces from "interventionist" policy choices - as if these were meaningfully distinct processes rather than deeply intertwined aspects of industrial development.
Even nuanced frameworks that acknowledge both elements still implicitly treat them as separate forces to be weighed against each other: how much is TSMC's dominance due to emergent industrial clustering versus strategic policy choices? How much did government funding versus market demand drive early semiconductor development? These questions assume we can meaningfully separate and measure these factors. But Taiwan's semiconductor success story reveals something more complex: decades of evolution where policy choices created new market possibilities which enabled new policy choices which shaped new market developments - a continuous process that resists decomposition into "inevitable / emergent" and "intervention / contingent" elements. Early policy choices shaped what market forces could emerge, while emerging market patterns shaped what contingent policies became possible.
Complex adaptive systems exhibit "the adjacent possible" - the set of potential states accessible from the current state. This concept helps us understand how apparently distinct processes interweave through time. What appears as emergence often reflects the systematic exploration of adjacent possibles, while what we call contingency often represents selection among these possibilities. The distinction breaks down because both processes operate simultaneously and inseparably in the temporal evolution of complex systems.
This reframing changes the goal. The goal isn't better prediction but better practice - developing ways of being, thinking and acting that match the temporal complexity of the systems we're trying to influence.
This pattern appears across policy domains. In urban development, successful cities don't simply emerge from agglomeration economies, nor are they purely planned - they develop through continuous interaction between policy choices and organic growth. In education policy, successful reforms work neither by simply letting learning patterns emerge nor through top-down mandates, but by understanding how policy interventions can shape and be shaped by existing educational practices. In environmental policy, effective interventions neither just regulate against market forces nor rely purely on market solutions, but create conditions where economic and environmental improvements can develop together.
Practically, it would mean embracing the following shifts:
1. Focus on Sequences, Not States: Rather than engineering specific outcomes, policy should enable productive sequences of development. Successful urban policies demonstrate this approach, understanding how each intervention creates possibilities for future development.
2. Shape Option Space Rather Than Outcomes: Instead of trying to directly create desired results, policy can focus on shaping what developments become possible.
3. Work Across Multiple Timescales: Effective policy operates simultaneously on different temporal scales: short-term incentives, medium-term capability building, and long-term option creation.
4. Leverage Existing Processes: Rather than creating change from scratch, policy can identify and amplify promising developments already underway.
Each of these need different analytical and measurement tools and frameworks, and several of them from outside the field of complexity science. I plan to explore some of these in future posts.
Traditional approaches seek either to engineer specific outcomes or to "let the market decide." Now, the goal shifts to enable productive trajectories of development. This applies whether we're talking about industrial capabilities, urban communities, educational achievements, or environmental sustainability.
Policy success comes not from choosing between emergence and intervention, but from understanding how to engage with the continuous unfolding of complex systems, better navigation of evolving possibility spaces.
The emergence-contingency distinction has served as a useful fiction in both analysis and policy making. It has helped develop important analytical tools, frameworks for intervention, and ways of understanding complex systems. But it has perhaps hit its ceiling of usefulness in a world where we increasingly need to engage with complexity rather than simplify it away.
Ultimately, the emergence-contingency distinction may be most valuable as a reminder of how our conceptual categories limit our practical capabilities.