The mugging scenario that exposes a fatal flaw in expected value math
Imagine a mugger who has no cash to take from you, but offers a deal: give him $10 now, and in exchange, he will pay you $10 quadrillion tomorrow using supernatural powers. Even if you think there is only a 1-in-a-trillion chance he is telling the truth, expected value math says you should pay, because the potential reward ($10,000) outweighs $10. Philosopher Nick Bostrom created this thought experiment to show how extreme, low-probability claims can hijack rational decision formulas.
The Mugger Without a Weapon
In a conventional street encounter, a mugger uses the threat of physical force to compel a victim to hand over their wallet. The victim weighs the immediate, tangible danger against the loss of their cash and rationally decides that compliance is the safest course of action. In 2009, philosopher Nick Bostrom introduced a thought experiment that inverted this dynamic entirely: what if a mugger carried no weapon and offered no physical threat, but instead wielded the formal mathematics of rational decision theory?
In Bostrom's scenario, the mugger approaches Blaise Pascal and makes a seemingly absurd proposition. The mugger admits he has no weapon, but promises that if Pascal hands over his wallet, the mugger will use supernatural or extra-dimensional powers to deliver an astronomical reward—such as millions of days of immense happiness or vast wealth. When Pascal predictably expresses extreme skepticism, the mugger does not argue about the plausibility of the claim. Instead, he simply invites Pascal to name whatever microscopic probability he feels is appropriate, then scales the promised reward up until the mathematical expected value makes handing over the wallet the mathematically mandatory choice.
The Heritage of Pascal's Wager
The thought experiment deliberately echoes the famous seventeenth-century argument known as Pascal's Wager. Blaise Pascal argued that a rational person should live as though God exists, because even if the probability of God's existence is considered small, the potential reward of infinite salvation—or the penalty of infinite damnation—vastly outweighs any finite sacrifice made during an earthly life. In expected value terms, multiplying any positive probability by an infinite payoff produces an infinite expected utility, making belief the mathematically dominant strategy.
Bostrom's mugger modernizes this conundrum by stripping away the necessity of literal infinity. In standard decision theory, infinities create well-known mathematical paradoxes and division-by-zero errors. Bostrom demonstrated that decision theory remains vulnerable even when restricted entirely to finite numbers. Because the mugger can name any arbitrarily large finite sum—be it quadrillions, decillions, or numbers far exceeding the count of atoms in the observable universe—he can easily outpace any tiny, non-zero probability the target might assign to the claim.
The Flaw in Expected Utility Maximization
The core of the problem lies in standard Expected Utility Theory, which directs agents to evaluate every possible choice by multiplying the probability of each outcome by the utility (or value) of that outcome, and then summing the results. For most ordinary daily choices, this approach is exceptionally reliable. It helps insurance companies price risk, guides investors in allocating capital, and provides a formal basis for automated decision-making in computer science and artificial intelligence.
However, the mugging reveals a structural vulnerability when the theory encounters extreme tail events. If an agent must assign a tiny, positive, non-zero probability to every logically possible statement, an exploitative actor can hijack the agent's behavior simply by inventing hypothetical scenarios with sufficiently enormous stakes. This leaves pure expected-utility maximizers open to manipulation by ungrounded, unfalsifiable promises of cosmic proportions, compelling them to sacrifice real resources in the present for negligible chances of unfathomable gains.
Capping Value and Truncating Probabilities
Philosophers and mathematicians have proposed several structural modifications to prevent rational agents from falling victim to this trick. One straightforward approach is to place a hard upper bound on utility functions. If there is a maximum limit to how much utility any single event can generate—or if the subjective value of additional goods exhibits steep diminishing returns—the mugger can no longer scale the promised payoff to infinity. Under a bounded utility function, multiplying an extremely small probability by a strictly capped reward fails to overcome the tangible cost of handing over the wallet.
Another proposed fix is probability truncation, where an agent rounds any probability below a certain microscopic threshold down to absolute zero. If a claim is deemed so outlandish that its likelihood is one in a trillion, the decision-maker simply treats it as impossible and ignores it entirely. However, probability truncation introduces its own theoretical pitfalls, such as violating standard axioms of probability and leaving agents vulnerable to the 'lottery paradox,' where millions of individually negligible probabilities add up to a significant combined probability that the model mistakenly ignores.
Occam's Razor and Complexity Penalties
A more sophisticated resolution involves how rational thinkers assign prior probabilities to extraordinary claims using principles of algorithmic information theory and Occam's razor. Under frameworks like Solomonoff induction, the prior probability of a hypothesis decreases exponentially with the descriptive complexity required to state it. To genuinely explain how a seemingly ordinary stranger possesses the power to manipulate universal outcomes and deliver astronomical payoffs requires an immensely complex model of the physical world.
Under this view, as the mugger increases the magnitude and complexity of the supernatural promise, the prior probability of the claim does not remain static; it plummets at a rate that outpaces or matches the growth of the proposed reward. If the penalty for complexity grows faster than the payoff scale, the resulting expected value shrinks toward zero rather than expanding, neutralizing the mugger's mathematical leverage.
The Symmetry Problem and Modern Relevance
A further philosophical counter-argument focuses on the inherent symmetry of unprovable claims. If an agent considers the possibility that a mugger has magical powers and will reward compliance, the agent must equally consider the possibility of an invisible, counter-balancing entity who will punish them with equal severity for succumbing to the mugger's request. Because there is no empirical evidence favoring the mugger's claim over the mirror-image counter-claim, the astronomical payoffs cancel each other out, leaving the agent free to evaluate only the immediate, observable facts.
Pascal's mugging remains a critical topic in contemporary philosophy, decision theory, and artificial intelligence safety. As researchers design advanced autonomous systems and long-termist policy frameworks that must weigh low-probability, high-consequence global risks, understanding how to handle extreme values without being paralyzed by wild hypotheticals is an essential challenge in modern rationality.
Key takeaways
•Pascal's mugging is a thought experiment by Nick Bostrom showing how standard expected utility theory can be exploited by claims with microscopic probabilities and astronomical payoffs.
•Unlike the original Pascal's Wager, which relies on the concept of infinite rewards, Pascal's mugging demonstrates that standard decision formulas fail even within strictly finite numbers.
•Proposed solutions include bounding utility functions, discounting extraordinarily complex claims via algorithmic probability, and recognizing that unprovable extreme claims are cancelled out by symmetrical counter-claims.
•The thought experiment is widely studied in artificial intelligence safety and rational decision theory to prevent autonomous agents from being paralyzed by extreme tail risks.