.png)
The Illusion of Control: Why Predictive Models Fail in Complex Systems
- Julien Haye
- May 16
- 10 min read
Updated: 2 days ago
Nearly half of risk professionals believe the biggest failure in predictive models is assuming people will act rationally.
That insight emerged from a recent LinkedIn poll asking which model assumptions are most likely to fail in complex systems. While outdated data and unanticipated shocks were noted, it was human behaviour that dominated the responses.
This raises a critical question: How many organisations still rely on tools that assume control, even when reality refuses to cooperate?
A common leadership reflex sits at the heart of this problem: “We have a model for that.” Whether facing market volatility, emerging threats, or strategic uncertainty, the default response is often to reach for historical data and predictive frameworks, many of which are built on a linear view of the world.
Yet recent events have shown how fragile that reliance can be. Risk models failed to anticipate the systemic shocks of the global financial crisis. Forecasts struggled to track COVID-19’s nonlinear spread. Geopolitical models underestimated the escalation of conflict in Eastern Europe and the Middle East. In each case, leaders may have had over-confidence in prediction.
Despite advances in analytics and AI, many decisions remain anchored in a core assumption: that the future will behave like the past. n some extreme cases, boards may spend entire meetings debating the RAG status of the prior quarter, as Horst Simon noted in a recent RiskMasters interview. This backward-looking orientation reflects a deeper challenge. The belief in predictability underpins countless models, frameworks, and investment strategies. But in a fast-moving, interconnected, and unpredictable world, that assumption is increasingly unfit for purpose.
This theme builds on a previous article exploring the difference between Risk vs. Uncertainty, and how mistaking one for the other reinforces blind spots and undermines resilience.
When leaders mistake predictability for preparedness, they build organisations that may appear in control but are unready for the unexpected. Optimised for yesterday’s risks, these systems often collapse under tomorrow’s uncertainty.
This article explores why predictive models fail in complex systems, how cognitive and cultural blind spots reinforce the illusion of control, and what leaders can do to shift from confidence in precision to confidence in adaptability.
Why Predictive Models in Complex Systems Create False Confidence
Statistician George Box famously observed,
"All models are wrong, but some are useful."
This reminder is essential because models, while helpful, are never reality. They are structured approximations, built on selected variables, past data, and assumptions about how the world works.
At their best, models support clarity and structure. They help leaders frame decisions and quantify risk. But the danger comes when models are mistaken for forecasts, or when confidence in their precision overrides critical thinking.
History offers stark reminders of this overconfidence. In 1998, Long-Term Capital Management collapsed when its highly sophisticated models failed to account for rare but interconnected market events. More recently, Silicon Valley Bank relied on interest rate risk models that failed to anticipate how quickly digital-era depositors could trigger a liquidity run. In both cases, the models were not inherently flawed, but their limits were misunderstood or ignored.
Models are useful when they guide awareness, not when they replace judgment. They are maps: helpful when the terrain is familiar, dangerous when it has already changed.
Ready to Lead with Clarity in Uncertain Times?
At Aevitium, we help leaders and risk teams distinguish between what can be controlled and what must be navigated. We move beyond traditional, model-dependent frameworks by embedding leadership foresight, cultural awareness, and strategic adaptability into the heart of risk governance.
Whether you're redefining your risk appetite, strengthening your governance model, or preparing teams to operate with confidence amid uncertainty, we work with you to align people, purpose, and process so your organisation can lead with resilience, not just react.
The Psychology Behind the Illusion
The over reliance on models is a psychological one. When leaders place excessive trust in predictive tools, they are often driven not by data itself, but by deeply embedded cognitive biases that shape how risk is perceived, interpreted, and acted upon.
One of the most pervasive is the illusion of control bias. This is the tendency to overestimate our ability to influence outcomes, particularly in complex or ambiguous situations. When models produce a neat set of projections, they offer overblown reassurance. Leaders feel they are steering the ship, even when the waters beneath are shifting unpredictably.
Closely related is model blindness, where decision-makers default to numerical outputs and structured dashboards, sidelining experience, dissent, and qualitative insight. Numbers give the impression of objectivity, but that objectivity can become blinding. Risk signals from the front line or external stakeholders may be ignored simply because they do not appear on the model.
Then there is confirmation bias: the tendency to seek out and prioritise information that supports existing beliefs. Models can be misused to reinforce a preferred strategy rather than to test or challenge it. When a model output aligns with what leadership already wants to believe, it is more likely to be trusted without scrutiny.
These biases have shaped real-world outcomes. During the 2008 financial crisis, risk models across institutions significantly underestimated the likelihood of systemic failure. Leadership teams, reassured by seemingly robust value-at-risk calculations, held onto assumptions of diversification and liquidity far longer than conditions warranted. The same pattern re-emerged during early phases of the COVID-19 pandemic, when business continuity plans, and pandemic response models often lagged the reality of exponential spread and behavioural shifts. In both cases, leaders were slow to adapt because their confidence in existing models muted signals that should have prompted action.
The illusion of control persists because it is comforting. It protects decision-makers from ambiguity. But in doing so, it also obscures the very risk signals that matter most in moments of change.
Complexity Is Not Chaos: It’s Adaptive
One of the most important distinctions in modern risk thinking is the difference between complicated and complex systems. The terms are often used interchangeably, but they describe fundamentally different realities and the way leaders approach each has major consequences for risk modelling.
A complicated system may have many parts, but it is ultimately knowable. With enough expertise, time, and data, it can be mapped, understood, and controlled. An aircraft engine, a financial reporting system, or a logistics routing algorithm are all complicated. They require skill and precision to manage, but they follow clear rules and can be fixed when broken. In these systems, the relationship between cause and effect is stable and predictable.
A complex system, by contrast, is adaptive, nonlinear, and emergent. It cannot be fully understood by examining its parts in isolation. Change in one area can trigger unexpected effects elsewhere, often amplified by feedback loops and time lags. Ecosystems, global financial markets, supply chains, public health systems, and organisational cultures all fall into this category. They evolve over time, shaped by interdependencies, shifting behaviours, and context-specific responses. As futurist and strategist Roger Spitz highlighted in our recent RiskMasters interview, complex systems require a completely different orientation to leadership, one grounded in adaptability and resilience, not control.
In complex systems, forecasting becomes fragile. Linear projection falls short because the variables do not stay fixed. For example:
A small disruption in one part of a global supply chain can cascade into a multi-month shortage due to just-in-time dependencies and lack of redundancy.
A central bank rate change intended to cool inflation can trigger behavioural responses in markets that amplify volatility instead.
An organisation rolling out a new risk framework may find its implementation success depends less on technical design and more on how people adapt to perceived shifts in power or accountability.
What makes complex systems especially challenging is that they cannot be solved. They can only be engaged with. The goal is not to find the perfect model, but to design systems and strategies that are flexible, responsive, and capable of learning over time.
Complexity is not chaos. It is patterned but unpredictable. Leaders who treat it like a complicated problem to be engineered into submission are likely to find that their well-intentioned interventions have unintended consequences and that their reliance on linear models leaves them vulnerable in precisely the moments when agility matters most.
📘 The Risk Within provides a roadmap for embedding psychological safety into risk management. It identifies critical touch points across the risk lifecycle and offers clear actions to align leadership, culture, and governance. It is designed to help risk functions integrate more deeply into the business and strengthen decision-making at every level.
Rethinking Models from First Principles
Before exploring how leaders can build preparedness, it helps to return to basics. Models are simplified representations of reality. We create them to reduce cognitive and operational complexity, not to predict every outcome.
Yet most models are built on fragile assumptions: that systems are stable, that relationships between variables are linear, and that people will behave predictably. These assumptions rarely hold in complex environments, where feedback loops, behavioural shifts, and nonlinear dynamics make control elusive.
When we strip those assumptions away, the role of a model changes. It is no longer a tool for forecasting what will happen. Instead, it becomes a way to frame possibilities, challenge thinking, and prepare for what might unfold.
This shift, from prediction to orientation, sets the stage for more adaptive, resilient approaches to risk.
From Prediction to Preparedness
If models cannot reliably tell us what will happen, then the better question for leaders becomes: What could happen, and how will we respond?
This mindset shift, from prediction to preparedness, changes how organisations approach risk. It replaces reliance on fixed forecasts with a commitment to resilience, adaptability, and forward-looking strategy.
Preparedness starts with scenario planning. Rather than fixating on a single outcome, scenarios help leaders explore multiple plausible futures. They allow organisations to stress-test assumptions and build strategic range into their thinking.
It also includes pre-mortems and red teaming. A pre-mortem asks, “If this plan failed, why did it fail?” before action is taken. Red teaming challenges dominant views by assigning individuals or groups to adopt a contrarian stance, helping reveal blind spots and surface alternative risks. Both techniques create space for structured dissent and reflection.
Adaptive decision-making plays a critical role. Instead of depending on static risk assessments or rigid planning cycles, resilient organisations build systems that allow for real-time sensing, feedback, and course correction. Decision-making becomes iterative, and governance supports change rather than resisting it.
These practices rely on more than structure. They require a culture of psychological safety, where teams feel secure enough to question assumptions, admit uncertainty, and surface weak signals. In this context, dissent is not disruption; it’s protection.
Leadership in a World Without Guarantees
The role of leadership is evolving. In a world where complexity is constant and disruption unavoidable, the most effective leaders are those who design systems and cultures that can adapt when variables shift.
To lead well in this environment, three capabilities are essential:
1. Ambiguity tolerance: Resilient leaders grow comfortable with uncertainty. Instead of rushing to impose clarity, they stay with complexity long enough to ask better questions and avoid oversimplification.
2. Learning loops: They treat every decision as a hypothesis and every outcome as data. Through feedback, review, and course correction, they encourage learning at every level of the organisation.
3. Decentralised decision authority: Frontline teams often spot weak signals first. Leaders who empower distributed decision-making enable faster, more informed responses, especially when speed matters most.
These traits reflect a broader leadership shift. Rather than focusing on control, effective leaders operate within their circle of influence shaping culture, governance, and learning environments that support adaptation. The goal is to become better equipped to navigate it.
Final Thought: Letting Go of the Illusion of Control
“Control is not the goal. Resilience is.”
This simple reframing captures the core message of this article. In a complex and unpredictable world, the most dangerous risk is the illusion that we can eliminate volatility through precision and planning alone.
Leaders must resist the seduction of false certainty. Forecasts, models, and dashboards can offer valuable insights, but when they become substitutes for critical thinking or proxies for control, they introduce fragility into the system. The overconfidence they inspire often leaves organisations least prepared when conditions change.
Resilience demands something different. It calls for humility to accept what we do not know, curiosity to explore what might happen next, and strategic foresight to build structures that can flex, learn, and adapt over time.
But resilience is a collective effort. A truly adaptive organisation requires a shared commitment to continuous learningacross teams, across functions, and across leadership levels. This means creating space for challenge, rewarding insight over certainty, and treating reflection as part of performance.
As a closing reflection, consider this:
Where in your organisation are you clinging to models, forecasts, or assumptions that no longer reflect the world as it is?
And what would it take to let go of the illusion that control is the same as preparedness?
Letting go of that illusion is not a weakness. It is the first step toward building systems, and leadership cultures, that are ready for whatever comes next.
About the Author: Julien Haye
Managing Director of Aevitium LTD and former Chief Risk Officer with over 26 years of experience in global financial services and non-profit organisations. Known for his pragmatic, people-first approach, Julien specialises in transforming risk and compliance into strategic enablers. He is the author of The Risk Within: Cultivating Psychological Safety for Strategic Decision-Making and hosts the RiskMasters podcast, where he shares insights from risk leaders and change makers.
FAQs
What are predictive models in complex systems?
Predictive models in complex systems are tools used to forecast future outcomes based on historical data and assumed relationships. However, in systems that are adaptive and nonlinear, such as financial markets or supply chains, these models often fail to account for unexpected interactions and emergent behaviour.
Why do predictive models often fail in complex systems?
Predictive models often fail in complex systems because they rely on assumptions of stability, linearity, and rational behaviour. Complex systems are dynamic and interconnected, meaning small changes can lead to unpredictable outcomes that models struggle to capture.
How can organisations manage risk in complex systems?
Organisations can manage risk in complex systems by shifting from prediction to preparedness. This includes using scenario planning, pre-mortems, and adaptive decision-making frameworks that promote flexibility and real-time learning.
What is the difference between complicated and complex systems in risk management?
Complicated systems are knowable and predictable with enough expertise, such as an engine or a tax system. Complex systems, like financial markets or organisational culture, are adaptive and unpredictable, making them harder to model accurately.
How can leaders avoid over-reliance on predictive models?
Leaders can avoid overreliance on predictive models by fostering a culture of curiosity, decentralised decision-making, and psychological safety. This allows teams to question assumptions, surface blind spots, and adapt to change as it happens.
Comments