The Physics of Not Knowing

Thesis

Physics distinguishes ignorance caused by missing information from uncertainty encoded in a theory’s possible outcomes, while showing that both become meaningful only through physical procedures. The disciplined question is not whether nature is “known” in the abstract, but which distinctions can be prepared, measured, transmitted, and predicted.

Not knowing has several shapes

Supporting/contextual references: [pnn-heisenberg-1927] [pnn-vonneumann-1955] [pnn-shannon-1948]

Suppose a coin is under a cup. You do not know whether it shows heads or tails, but the coin may already have a face upward. Now imagine a quantum preparation whose measurement outcomes cannot be represented as merely unknown pre-existing values for every possible measurement. The two situations both invite the phrase uncertainty, yet they do different explanatory work. One concerns an observer’s missing record; the other concerns the structure of allowed states and predictions.

A third case is practical ignorance. A weather model can be uncertain because its initial data are incomplete, its resolution is finite, or its equations amplify small errors. A fourth is statistical description: a probability distribution may summarize a population even when every member has a definite state. Careful science does not turn these into one fog. It asks what was prepared, what was measured, what the probability means, and whether a better interaction could in principle resolve the distinction.

From ignorance to probability

Supporting/contextual references: [pnn-shannon-1948] [pnn-vonneumann-1955]

Classical probability often begins with a sample space and a state of information. If a fair die is rolled behind a screen, assigning one-sixth to each face is rational before the result is seen. The die’s motion is physical, but the probability assignment represents what an observer can expect from the available setup. In statistical mechanics, probability distributions similarly summarize macroscopic knowledge over many compatible microstates, while the gas itself has particles and dynamics.

This epistemic role is not a weakness. It lets probability guide decisions, tests, and predictions without pretending that a number is a material ingredient. It also exposes a responsibility: probabilities require a reference class, preparation procedure, or model. “The chance is fifty percent” is incomplete until we say fifty percent under what ensemble, information state, or repeated protocol. The physics of not knowing starts by making the hidden conditioning visible.

The quantum limit

Supporting/contextual references: [pnn-heisenberg-1927] [pnn-vonneumann-1955] [pnn-bell-1964]

Quantum mechanics makes uncertainty structural in ways that have no simple classical translation. A state can be prepared so that a measurement of one observable has a sharp result, while a complementary measurement has a spread described by the uncertainty relations. The point is not merely that an experimentalist lacks a tiny hidden value. Some pairs of observables are represented by noncommuting operators, and the theory does not provide a joint assignment with the same status as a classical list of properties.

The uncertainty principle is often illustrated by position and momentum, but its meaning is broader. Preparation, dynamics, and measurement constrain which distinctions can be jointly sharp. Quantum information theory gives these constraints operational form: no-cloning limits the copying of unknown states, and incompatible measurements make some information expensive or impossible to acquire together. These are testable features of a formalism, not invitations to say that consciousness manufactures ambiguity.

What a measurement can reveal

Supporting/contextual references: [pnn-heisenberg-1927] [pnn-bell-1964] [pnn-fuchs-2014]

A measurement is an intervention that couples a system to an apparatus and produces a record. Its outcome is not a view of every property the system possesses. It is a result generated under a specified setting, with a specified resolution, and often with a disturbance that changes which later questions can be asked. Asking what a photon “really had” before every possible measurement can exceed what the experimental arrangement defines.

This does not make measurement arbitrary. Calibration links pointer positions to quantities; repeated preparations yield frequencies; control experiments expose noise and bias. A result can be surprising without being unconstrained. The distinction is between underdetermination and anything-goes subjectivism. A theory may say that more than one outcome is possible while still predicting exact distributions, correlations, and bounds on what any apparatus can do.

Information has a thermodynamic price

Supporting/contextual references: [pnn-landauer-1961] [pnn-bennett-1982] [pnn-shannon-1948]

Knowing is not a purely mental act once a system records what was learned. A detector changes state, a memory stores a distinction, and communication copies a pattern into another physical medium. Landauer’s analysis showed that logically irreversible erasure has a thermodynamic cost in an idealized setting: resetting an unknown bit to a standard state requires heat dissipation of at least kT ln 2 when the usual assumptions apply. The formula concerns a physical operation, not an occult value attached to knowledge.

The cost is clarifying because it separates acquisition, storage, copying, and erasure. Reversible computation can in principle avoid the specific cost of logical erasure by preserving input distinctions, though real devices incur other losses. A memory can also become useless without being literally erased, if noise destroys its correlations. Not knowing is therefore partly a physical condition: the relevant record may never have formed, may have dispersed, or may remain inaccessible to the agent who needs it.

The observer’s uncertainty

Supporting/contextual references: [pnn-fuchs-2014] [pnn-vonneumann-1955] [pnn-zurek-2009]

In some interpretations, a quantum state is best understood as an agent’s organized expectations rather than a literal inventory of a system’s intrinsic properties. This view makes not knowing central, but not in the sense that the world is invented by a private mind. An agent’s expectations are constrained by preparation, prior records, the Born rule, and future experience. Another agent can assign differently because their evidence differs; communication can then change both accounts.

Other interpretations treat the state as part of an objective physical description, perhaps supplemented by collapse, branching, or hidden variables. They disagree about what uncertainty means while agreeing on the empirical predictions of ordinary quantum experiments. This is why a measured distribution rarely settles an interpretation by itself. The data tell us which structures a theory must reproduce; the philosophical choice concerns what role those structures play in reality.

Open research directions

Supporting/contextual references: [pnn-zurek-2009] [pnn-bennett-1982] [pnn-fuchs-2014]

Several questions remain live. Are there experimentally accessible deviations from quantum theory that would reveal an objective collapse mechanism? Can quantum gravity tell us whether uncertainty is altered when spacetime itself fluctuates? How should probabilities be interpreted in cosmology when there is no external ensemble of universes to sample? These are distinct problems, and none is solved by repeating that “everything is uncertain.” Their answers require precise models and discriminating observations.

A related frontier concerns quantum technologies. Error correction turns unreliable microscopic states into robust logical information, but thresholds depend on noise models and architecture. Researchers still ask how much contextual information can be extracted from complex quantum systems, and which thermodynamic costs accompany error correction and feedback. The practical aim is not to eliminate uncertainty. It is to characterize it well enough to route around, correct, or exploit it without confusing control with omniscience.

A disciplined humility

Supporting/contextual references: [pnn-heisenberg-1927] [pnn-landauer-1961] [pnn-zurek-2009]

The useful response to a limit is to identify its source. A missing instrument calls for measurement; incompatible observables call for a new experimental question; underdetermination calls for rival models. Naming the limit turns uncertainty from atmosphere into a research object. It also protects the distinction between epistemology and ontology: inability to jointly measure two quantities does not by itself say that reality lacks both.

That distinction changes what we do next. If a record is missing, build a better detector. If the observables are incompatible, redesign the question. If a theory is underdetermined, compare predictions rather than treating mystery as evidence. The point is not to possess reality all at once, but to learn which limits arise from a state, an apparatus, a model, or the world’s causal structure.

Physics earns humility by making limits precise: a distribution can be measured, a spread derived, an erasure cost calculated, and a horizon located. That precision leaves no warrant for mind-induced facts or preferred spiritual stories, but it does leave room for genuine unknowns. Wonder is most durable when the boundary is stated clearly enough that a future experiment could move it.

Sources & references

Supporting/contextual references, not claim-level proof.

1. Werner Heisenberg — Über den anschaulichen Inhalt der quantentheoretischen Kinematik und MechanikZeitschrift für Physik 43 (1927), 172–198.Publisher link
2. John von Neumann — Mathematical Foundations of Quantum MechanicsPrinceton University Press, 1955.
3. Claude E. Shannon — A Mathematical Theory of CommunicationBell System Technical Journal 27 (1948), 379–423, 623–656.Publisher link
4. Rolf Landauer — Irreversibility and Heat Generation in the Computing ProcessIBM Journal of Research and Development 5 (1961), 183–191.Publisher link
5. Charles H. Bennett — The Thermodynamics of Computation—A ReviewInternational Journal of Theoretical Physics 21 (1982), 905–940.Publisher link
6. John S. Bell — On the Einstein Podolsky Rosen ParadoxPhysics 1 (1964), 195–200.
7. Christopher A. Fuchs, N. David Mermin, and Rüdiger Schack — An Introduction to QBism with an Application to the Locality of Quantum MechanicsAmerican Journal of Physics 82 (2014), 749–754.Publisher link
8. Wojciech H. Zurek — Quantum DarwinismNature Physics 5 (2009), 181–188.Publisher link