Risks and uncertainties in strategy

In 2007, several of the world's largest banks employed thousands of risk analysts and ran complex quantitative models. These models could price individual mortgage - backed securities with astonishing precision. They could calculate default probabilities, assign credit ratings, and generate value - at - risk estimates that satisfied regulators and reassured shareholders. However, by 2008, the global financial system was on the verge of collapse. These models did not fail to achieve their intended goals; rather, they were applied to a problem they were fundamentally unable to handle.

Individual mortgages can be calculated, and their default rates can be estimated based on historical data. However, the intricate interactions among millions of bundled, split, and re - bundled securities produce completely different results. When housing prices fell, the chain reaction spread rapidly through interconnections that no model could depict, because these connections are themselves emergent properties of the system's complexity. Banks treated this real uncertainty as a risk problem, which was a fundamental cognitive misalignment.

This is not a story about mathematical errors but about the misuse of concepts. We often use the term "risk," but in fact, we are referring to something completely different. Project risk assessment, geopolitical risk analysis, discussions about artificial intelligence risk: in most cases, the term "risk" has become a synonym for "something bad might happen." This term masks a distinction that is crucial in practice.

Vaughn Tan (a professor at the National University of Singapore Business School, who has long studied how organizations deal with uncertainty and is the author of the book Uncertainty Mindset) has spent years dissecting how organizations deal with the unknown. He described a chain reaction that makes this confusion have serious consequences: how we name something determines what we think it is, and what we think it is determines how we act, which ultimately determines the outcome. When we label a situation as "risk," we are actually implying that the future state is knowable and its probability is calculable. This implication activates a specific set of tools: quantitative models, cost - benefit analysis, expected value calculations, and insurance mechanisms. If there is indeed uncertainty in the situation, these tools will create false confidence, which is worse than admitting ignorance.

Therefore, confusing risk and uncertainty is not a problem of inaccurate language expression but a problem of misguided action. Using words correctly is a prerequisite for correct response. This article traces the difference between the two, explores why we always confuse them, and analyzes the practical consequences of this confusion.

Difference

In 1921, American economist Frank Knight (one of the important founders of the Chicago School of Economics, whose distinction between risk and uncertainty remains a fundamental concept in the fields of economics and management to this day) published the book Risk, Uncertainty, and Profit, which clarified two distinct relationships people have with the unknown.

Risk describes a situation where the possible future states and their probabilities are known. For example, a fair die has six faces, and the probability of each face appearing is one - sixth. Insurance companies can calculate the likelihood of a 45 - year - old non - smoker dying in the next ten years because actuarial tables are based on a large historical data set. In these cases, formal rational decision - making is applicable. Cost - benefit analysis is effective. Expected value calculations make sense. The unknown factors can be precisely measured in a certain sense.

Uncertainty refers to a situation where the possible future states cannot be fully enumerated, or their probabilities cannot be reliably determined. For example, the launch of a truly novel product, the long - term impact of a new technology, and the direction of geopolitical conflicts: these all involve unknown factors that are difficult to quantify. Formal rational tools cannot be effectively applied and may even be misleading.

Knight distinguished three types of probabilities, which help to clarify this boundary. A priori probability is purely deductive reasoning: for example, the six faces of a die or the 52 cards in a deck of playing cards. Statistical probability is inferred from historical data: for example, mortality tables, insurance claim records, and batting averages. Estimated probability depends on subjective judgment: for example, an entrepreneur's intuition about market demand or an analyst's assessment of political stability.

The key is that only a priori probability can fully support the operation of a formal rational system. Although statistical probability is close to a priori probability, it implies an assumption that the future is similar enough to the past so that historical frequencies still hold. Estimated probability goes further and depends on judgment rather than data. However, in practice, these three types of probabilities are often confused. Organizations apply quantitative frameworks that are only applicable to a priori probability to estimated probability. This systematic confusion is the root of the problem.

Knight's practical view is about profit. Risk can be eliminated through insurance. If the probability of loss is known, insurance can be priced, and risk is no longer a factor in economic decision - making. In contrast, uncertainty cannot be insured because no one can price something that cannot be measured. Knight believed that an entrepreneur's profit is precisely the reward for taking on real uncertainty: profit comes from the inherent unpredictability of things, from the unknowability of the results of human activities, and from the fact that probability calculations are neither possible nor meaningful in many cases.

This is not an abstract philosophical distinction. It has a direct impact on how organizations build decision - making mechanisms, allocate resources, and prepare for the future.

Why Do We Make Mistakes?

If this distinction is so important, why are we always so persistent in confusing it? The answer lies in three levels.

The Words Themselves

Vaughn Tan conducted a detailed analysis of the language dimension. This problem has two aspects. First is overuse: the term "risk" is used to refer to too many different things, so that it has almost lost its meaning as a diagnostic term. In one of Tan's seminars, ten participants proposed more than eight different informal definitions of "risk": potential problems, known problems and mitigation measures, probability of negative outcomes, cost of inaction, downside risk exposure, assessed threats, items on the risk register, etc. Each definition implies a different relationship with the unknown, but the term "risk" itself masks these differences.

The second aspect is appropriation: the term "uncertainty" has been appropriated by some fields, which use it to refer to a concept closer to calculable risk. In the fields of artificial intelligence and machine learning, "uncertainty quantification" usually refers to the confidence interval of model predictions. In mainstream economics, "uncertainty" usually refers to the volatility that can be modeled with a probability distribution. In both cases, the term has been appropriated to describe situations that Knight would classify as risk rather than uncertainty. This appropriation deprives the term of its diagnostic power where it is most needed.

As Tan wrote: "Confusing terminology about the unknown hinders organizations from establishing a good relationship with the unknown."

Hidden Emotions

Uncertainty instinctively makes people feel uncomfortable. Risk thinking provides psychological comfort because it implies that the situation is, in principle, controllable and predictable. Assigning a probability to a threat, even if the probability is unreliable, is much better than admitting that the situation cannot be quantified at all.

This emotional aspect is rarely openly discussed in organizational life. Fear, anxiety, and unease caused by the unknown are not on the meeting agenda. However, they are powerful drivers of behavior. Organizations develop what Tan calls "antibodies" against truly facing uncertainty: a reflexive reaction that turns every uncertain situation into a risk management exercise because risk management is more emotionally acceptable than admitting uncertainty.

The result is a form of organizational self - comfort. The risk framework cannot cure the root problem but only alleviate the symptoms. Boards that receive risk dashboards feel informed. Teams that complete risk assessments feel prepared. These activities are real, the efforts made are sincere, and the sense of comfort they bring is immediate. However, the cost will be apparent later: when reality does not match the model and the organization finds that it has wasted all its preparation energy on the wrong preparations.

Institutional Infrastructure

Cost - benefit analysis, expected value calculations, risk management departments, insurance frameworks, regulatory compliance mechanisms: the entire institutional architecture of modern organizations is built on risk thinking. Once a situation is classified as "risk," this architecture will automatically start. Analysts conduct quantitative assessments, committees review the risk register, and boards receive risk dashboards. This mechanism is impressive and is indeed very effective within its applicable scope.

The problem is that this mechanism does not have an "off" button and cannot automatically stop operating when it is not applicable. When the situation is truly full of uncertainty, there is no institutional mechanism to indicate: "Our standard tools are not applicable here, and we need to adopt a different approach." Instead, these tools are still used because they are readily available, and using them seems to be a form of due diligence. The institutional framework itself creates a demand that does not match the actual nature of the problem.

Consequences

When risk tools are applied to uncertain situations, the result is not a minor error but a systematic misguidance. The 2008 financial crisis mentioned above confirms this core pattern: precision itself is the danger because it creates confidence where humility is needed. Banks were not short of sophisticated tools. They ran Monte Carlo simulations (a mathematical method that estimates the probability distribution of complex systems through a large number of random samplings and is widely used in risk assessment in the financial field), stress tests, and correlation analyses. These tools provided seemingly precise answers to a problem that could not be given a precise answer at all.

The collapse of Silicon Valley Bank in 2023 shows that this confusion persists even when the consequences are apparent. Almost everyone attributed this event to "poor risk management." The bank concentrated its assets in long - term bonds without hedging against interest rate fluctuations. From a traditional perspective, this seems to be a simple case of risk management failure. However, the root cause of Silicon Valley Bank's collapse was not this. Depositors withdrew $42 billion in one day. The speed of the bank run was unprecedented, and the large network of depositors composed of technology companies and venture capitalists exacerbated this trend as they acted in synchronization. Interest rate risk exposure is a risk problem. However, such a large - scale and rapid deposit flight is an uncertainty problem. The standard diagnostic language classifies both as "risk," thus ignoring a more important question: Did the bank's model really target the correct threat category?

The consequences of mislabeling are far more than just inaccuracy. Organizations will optimize in completely the wrong direction. Organizations that treat uncertainty as risk will build complex quantitative models, creating an illusion of control. They focus on perfecting predictions rather than cultivating the ability to respond effectively in unpredictable situations. They mistake precision for accuracy and control for being well - prepared.

The Daniel Ellsberg Paradox (Daniel Ellsberg, an American economist and strategic analyst, famous for leaking the "Pentagon Papers" in 1971, but his most important academic contribution in the field of economics is the "Ellsberg Paradox" proposed in 1961, which reveals human irrational behavior when facing ambiguous probabilities) was confirmed through empirical research in 1961, which helps to explain why this pattern is so persistent. Ellsberg pointed out that people dislike ambiguity even more than they dislike risk. When choosing between a bet with a known probability and a bet with an unknown probability, people tend to choose the option with a known probability, even if the expected values of the two are the same. This aversion to ambiguity is not based on conscious thinking. This means that organizations will tend to interpret a situation as "risk" because risk is cognitively easier to control than uncertainty.

Beyond the Binary Opposition: Multiple Forms of the Unknown

The distinction between risk and uncertainty is only a starting point, not an end. Although this binary division reflects some realities, the realm of the unknown is far more complex and diverse than what this binary classification scheme implies.

Why "Unknown"?

Tan deliberately introduced the term "unknowing" to replace "uncertainty." This is more for practical than aesthetic reasons. As shown in the above language analysis, the term "uncertainty" has been overused, so using it will exacerbate the confusion it is supposed to solve. The term "unknowing" is indeed a bit clumsy. However, its value lies precisely in this clumsiness: it is not easily used casually, forcing the speaker to clearly indicate which type of unknowing is being discussed.

Four Sources of Ignorance

Tan pointed out that ignorance may stem from four different aspects.

First, unclear causal relationships. The relationship between actions and results is not clear. For example, when a central bank lowers interest rates, its systematic impact spreads through some unclear channels, ultimately leading to consequences that no causal model can reliably predict. The actions are well - thought - out; however, due to the opaque causal chain, the results are truly uncertain.

Second, unknown action space. It is even difficult to determine which actions are feasible. System complexity, social dynamics, or incomplete information may blur all possible action plans. An organization facing a new crisis may not know what it can do, let alone what it should do. The available options themselves are part of the unknown.

Third, unknown result space. The set of possible results itself is not clear. Either the existing results have not been discovered, or the relevant results do not yet exist. Innovation essentially exists in this field: a truly novel product or technology will create results that cannot be enumerated in advance.

Fourth, unknown preferences. We do not know which results are desirable. This is directly related to sense - making: when we do not know what we value, we make subjective judgments, and no calculation can replace this judgment. Preference uncertainty is the field where sense - making becomes a creative response to the unknown.

The existence of any of these sources is sufficient to place a situation in the category of real uncertainty rather than calculable risk.

Complementary Frameworks

Some other frameworks depict adjacent fields. Each framework supplements specific content that Knight's binary classification and Tan's classification themselves cannot cover.

The Cynefin framework, developed by Dave Snowden (a Welsh knowledge management scholar who worked at IBM and later founded the Cognitive Edge research institution, focusing on the application of complexity theory in organizational decision - making) and Mary Boone (an American management consultant who has long studied leadership and complex systems), is a diagnostic tool with five domains: simple (later renamed clear), complicated, complex, chaotic, and disorder. Risk exists in the simple and complicated domains, where causal relationships are either obvious or can be discovered through expert analysis. Uncertainty exists in the complex and chaotic domains, where causal relationships can only be seen after the fact or cannot be distinguished at all. The Cynefin framework adds an operational question that Knight did not raise to the distinction between risk and uncertainty: how to determine which domain you are in? The core warning of this framework is that domain mismatch is the main strategic error. Treating a complex situation as a complicated situation means applying expert analysis when experimentation is needed.

Sense - making. Karl Weick (an American organizational psychologist, an emeritus professor at the University of Michigan, one of the most influential scholars in the field of organizational behavior, whose "sense - making" theory has profoundly influenced people's understanding of how organizations make judgments in ambiguous environments) proposed ambiguity as a category independent of risk and uncertainty. Under uncertainty, probabilities are unknown; under ambiguity, meanings are unclear. This addition is crucial because dealing with ambiguity requires a fundamentally different approach from risk management or dealing with uncertainty. Sense - making is a continuous process that projects reasonable narratives onto ambiguous situations: reasonableness is more important than precision, and action is as important as analysis because action generates data for interpretation. Knight and Tan focus on what we do not know, while Weick focuses on what we cannot yet interpret.

Antifragility. Nassim Nicholas Taleb (a Lebanese - American scholar, a former financial trader, and now a professor at New York University, famous for his best - selling books The Black Swan and Antifragile and for criticizing mainstream financial models for underestimating extreme events, well - known among Chinese readers) thoroughly expanded Knight's distinction. He believes that the problem is that we pretend to be able to measure what we actually cannot. Standard probability models systematically underestimate the likelihood and impact of extreme events (what he calls "black swan" events) because they assume a thin - tailed distribution, while the actual situation is a thick - tailed distribution (a thin - tailed distribution means that extreme events are extremely rare, and a thick - tailed distribution means that extreme events occur more frequently than predicted by conventional models). Taleb's contribution is to provide a set of design vocabulary for dealing with uncertainty: a fragile system is damaged by fluctuations, a robust system can resist fluctuations, and an antifragile system can benefit from them. His practical principle is the "negative approach": avoid making huge fragile commitments, build options, and allow for trial and error. Small failures can avoid catastrophic consequences.

Connections

Sense - making. Sense - making plays a role in the fourth type of unknown in Tan's theory: the unknown about values. It is a generative response to preference uncertainty. When we do not know which results are desirable, we make subjective value judgments, and no calculation can replace this judgment.

Fictitious expectations. Jens Beckert (a German sociologist, the director of the Max Planck Institute for the Study of Societies, focusing on economic sociology, whose book Imagined Futures explores the core role of expectations and uncertainty in the operation of capitalism) starts his theory of capitalist dynamics from the divergence point between risk and uncertainty. When real uncertainty dominates, rational calculation fails. Fictitious expectations are the social mechanism for economic actors to coordinate under uncertainty: collective imagination of the future replaces probability calculation. As Beckert wrote, "The failure of the rational actor theory is not because actors are unwilling to maximize their own utility, but because it cannot deal with the consequences of real uncertainty."

Scenario planning. Scenario planning is explicitly anti - prediction and aims to deal with uncertainty rather than risk. Pierre Wack (a French strategic planner, a core figure in the long - term planning department of