In 2022, the Nobel Prize in Physics went to Alain Aspect, John Clauser, and Anton Zeilinger for experiments with entangled photons—experiments whose results are irreducibly probabilistic. No hidden variable, no clever reinterpretation of the formalism, no appeal to observer ignorance can eliminate the role of probability from the quantum description of nature. Yet a powerful tradition in contemporary philosophy insists that all probability is ultimately subjective—a measure of what rational agents believe, not a feature of the world itself.

Bayesianism has earned its influence. As a framework for updating beliefs in light of evidence, it is elegant, general, and remarkably successful across statistics, decision theory, and philosophy of science. But its very success has encouraged an overreach: the claim that physical probability—the kind that governs radioactive decay, quantum measurement, and stochastic processes in statistical mechanics—is nothing more than credence dressed in empirical clothing. This conflation is a philosophical error with deep consequences.

The argument here is not that subjective probability is illegitimate. It is that physical probability possesses structural features that credence alone cannot replicate. Objective chances are embedded in physical laws, invariant across reference classes, and—most strikingly in quantum mechanics—resistant to any interpretation as mere ignorance about underlying states. Understanding why these features matter is essential for anyone who takes both physics and philosophy seriously.

Start with the most basic distinction. When we say that the probability of a tritium atom decaying within 12.3 years is 0.5, we are making a claim about the physical world. This probability is nomologically grounded: it derives from the laws governing weak nuclear interactions. It is not indexed to any particular agent's epistemic state. A physicist in Tokyo, a chemist in São Paulo, and a sentient alien in the Andromeda galaxy—if they each possess the relevant physical theory—will assign the same value. Credence, by contrast, is agent-relative. My confidence that it will rain tomorrow reflects my evidence, my priors, my cognitive situation.

This difference is not merely verbal. Objective chances exhibit reference class invariance in a way that subjective degrees of belief do not. The half-life of tritium is 12.3 years whether we partition the ensemble by location, time of creation, or any other physically irrelevant variable. Credences, however, shift with every change in background information. Bayesians celebrate this responsiveness to evidence as a feature; but the very flexibility that makes credence epistemically virtuous makes it structurally unlike physical probability.

Consider also the role of chance in law-governed explanation. When we explain why approximately half of a large sample of tritium atoms have decayed after 12.3 years, we invoke a lawlike probability. The explanation works because the chance is stable, projectible, and counterfactual-supporting: if we had prepared a different sample under the same conditions, the same statistical regularity would obtain. Subjective credence does not support counterfactuals in this way. My belief that it will rain tomorrow does not explain tomorrow's weather.

David Lewis recognized this asymmetry clearly. In his account of objective chance, he insisted that chances are features of the Humean mosaic—patterns in the actual distribution of events that earn their status through their role in the best systematization of nature. Whether or not one accepts Lewis's Humean metaphysics, the key insight stands: chances do explanatory and nomological work that credences cannot. They figure in physical theories as theoretical posits, not as summaries of what anyone happens to believe.

The subjectivist might respond that all we ever access is our credence about chances, and that talk of objective probability is simply a convenient fiction. But this conflates epistemology with ontology. The fact that we learn about chances through evidence and belief-updating does not entail that chances are beliefs. Electrons have charge regardless of whether anyone measures it; the same structural point applies to the probabilities that govern their behavior.

Takeaway

Physical probabilities are embedded in laws, invariant across reference classes, and support counterfactual explanations—features no subjective credence can possess. Confusing the two collapses a vital distinction between what the world does and what we happen to believe about it.

If objective chance and subjective credence are genuinely different, how do they relate? The most important answer comes from Lewis's Principal Principle: a rational agent's credence in a proposition, conditional on knowing the objective chance of that proposition and possessing no inadmissible information, should equal that objective chance. Formally, if you know the chance of heads is 0.5, and you have no crystal ball or time-travel device, your credence in heads should be 0.5.

This principle is a rationality constraint—a bridge between the metaphysics of chance and the epistemology of belief. It tells us that objective chances are precisely the kind of thing that ought to calibrate credence. But notice the direction of authority: chance constrains credence, not the other way around. The principle presupposes that chance is something credence must defer to, which is incoherent if chance just is credence. You cannot rationally defer to yourself in the way the Principal Principle demands.

Subjectivists have attempted to domesticate the Principal Principle by reinterpreting chance functionally: objective chance is whatever plays the credence-guiding role. But this maneuver is circular if no independent characterization of chance is offered. What makes a particular value the one credence should track? The answer, invariably, appeals to physical features—frequencies in law-governed processes, symmetries in physical state spaces, dynamical properties of systems—that are not themselves credences.

The Principal Principle also illuminates why the subjective-objective distinction matters practically. In statistical mechanics, we assign probabilities to microstates of a gas. Are these credences reflecting our ignorance of the exact microstate, or are they objective features of the dynamics? The answer affects how we understand thermodynamic irreversibility, the status of the Second Law, and the nature of equilibrium. If probabilities in statistical mechanics are purely subjective, the arrow of time becomes a feature of our epistemic limitations rather than of the physical world—a conclusion most physicists find deeply implausible.

The Principal Principle, then, is not a concession to subjectivism. It is a demonstration that the two kinds of probability are conceptually linked but ontologically distinct. Rational agents should align credence with chance; but the normative force of that 'should' depends on chance being something independent of credence—a feature of nature to which belief must answer.

Takeaway

The Principal Principle shows that objective chance constrains rational credence, not the reverse. This normative asymmetry is only intelligible if chance is something genuinely independent of belief—something in the world that our degrees of confidence must answer to.

The strongest case against subjectivism about physical probability comes from quantum mechanics. In classical statistical mechanics, one can at least entertain the possibility that probabilities reflect ignorance of a determinate underlying state. The gas has a precise microstate; we just don't know which one. Quantum mechanics systematically blocks this move.

Bell's theorem and its experimental confirmations—those Nobel-winning experiments—establish that no theory of local hidden variables can reproduce quantum statistics. The correlations between entangled particles violate Bell inequalities, meaning the probabilistic predictions of quantum theory cannot be explained by supposing that particles carry pre-existing definite values we are merely ignorant of. The probabilities are not epistemic gaps; they are features of the physics itself.

This has a precise philosophical consequence. If quantum probabilities were subjective credences reflecting ignorance of an underlying determinate reality, then a more complete theory—one specifying those hidden variables—should in principle be possible. Bell's theorem shows that no such local completion exists. The Kochen-Specker theorem goes further, demonstrating that quantum observables cannot all possess simultaneously definite values independent of measurement context. The structure of quantum probability is non-classical: it does not conform to a single Boolean algebra, and it resists embedding into any classical probability space that could be interpreted as ignorance over pre-existing facts.

QBism—Quantum Bayesianism—represents the most sophisticated attempt to maintain a subjectivist reading of quantum probability. QBists argue that quantum states encode an agent's beliefs about future experiences, not features of mind-independent reality. But this comes at a steep cost. It renders the extraordinary empirical success of quantum theory—its agent-independent predictive accuracy across every domain from particle physics to cosmology—a profound coincidence. If quantum probabilities are merely personal, why do all competent physicists converge on the same values? The QBist answer—that the formalism is a normative tool for belief coordination—uncomfortably echoes the very objectivity it seeks to deny.

The lesson from quantum mechanics is not simply that some probabilities are objective. It is that nature's probabilistic structure is more constrained and more alien than anything in our subjective doxastic repertoire. Quantum probability spaces have a geometry—captured by Hilbert space structure—that has no counterpart in classical belief revision. This mathematical distinctiveness is a powerful indicator that we are dealing with a feature of reality, not a feature of minds.

Takeaway

Quantum probabilities cannot be understood as ignorance of hidden definite states. Bell's theorem, the Kochen-Specker theorem, and the non-classical structure of quantum probability spaces all point to the same conclusion: some probabilities are woven into the fabric of nature itself.

The subjectivist program in probability has genuine virtues—it clarifies the logic of belief revision and grounds statistical inference in rational norms. But extending it to all probability, including the chances posited by fundamental physics, is an overreach that obscures more than it illuminates.

Physical probabilities are law-governed, reference-class invariant, counterfactual-supporting, and—in the quantum domain—irreducible to ignorance about underlying determinate states. The Principal Principle shows that the relationship between chance and credence is one of normative deference, not identity. These are not subtle philosophical quibbles; they shape how we understand the arrow of time, the meaning of quantum theory, and the structure of physical explanation.

To take probability in physics seriously is to accept that the world itself is, at its deepest levels, stochastic in ways that no amount of information or rational reflection can dissolve. That is not a limitation of our knowledge. It is a discovery about reality.