Information Is Not Meaning

Thesis

Information theory measures distinguishable possibilities and correlations; meaning concerns what a distinction signifies or does for a system. The two can cooperate in language, biology, and machines without becoming the same concept.

The word has two lives

Supporting/contextual references: [meaning-shannon-1948]

In ordinary speech, information is something that tells us what is going on. In Claude Shannon’s theory, information has a narrower task: quantify uncertainty associated with possible messages. A fair binary choice carries one bit; a choice among four equally likely alternatives carries two. The measure is exact because it does not ask whether the message is true, useful, or beautiful. It counts distinctions under a specified probability model.

That restraint is easy to lose. A string produced by a random-number generator may have high Shannon entropy while communicating nothing to a reader. A repeated sentence may carry little new information for someone who already knows it, yet convey urgent meaning in a crisis. The same physical pattern can have different roles because information and meaning answer different questions.

A pattern needs a role

Supporting/contextual references: [meaning-morris-1938] [meaning-millikan-1984]

Consider 01000001. Under ASCII it is the letter A; under another convention it is the number 65; in an unconfigured circuit it is a sequence of voltage levels. Nothing in the bare pattern announces which interpretation is correct. Meaning arises from a relation among a physical carrier, a code or practice, and a system able to use the distinction. The pattern matters because something has been organized to respond to it.

This dependence on a role does not make meaning arbitrary. A traffic light works because infrastructure, people, and regulations stabilize a convention. A gene sequence has biological significance because cellular machinery has been shaped to bind, copy, and translate molecular arrangements. An interpreter can be mistaken, but interpretation is constrained by regular consequences. Meaning is not an invisible fluid added to a bit; it is an achievement of organized relations.

Correlation is not reference

Supporting/contextual references: [meaning-dretske-1981] [meaning-morris-1938]

A footprint carries information about a foot in the causal sense: given suitable background knowledge, its shape and depth support an inference. But the footprint does not refer to a foot in the same way the word foot does. A smoke plume can indicate fire without intending to indicate anything. A written warning is different because a community has trained readers to treat marks as symbols in a practice of communication.

Philosophers often distinguish natural signs from conventional or intentional signs. The distinction is not absolute; language is embodied in sounds, ink, and neural states, while natural traces can be recruited into investigation. Still, it prevents a common mistake. The existence of a correlation does not imply that the correlated system has a message. Meaning involves norms of use, possible misunderstanding, and a perspective from which a distinction counts for something.

Brains, bodies, and environments

Supporting/contextual references: [meaning-millikan-1984] [meaning-clark-2016]

Meaning is not confined to private minds. A bacterium can move up a chemical gradient because receptor states regulate swimming; a bee can use a dance to guide a colony; a person can read a map. These examples differ in complexity, but each connects physical differences to action within an organized system. The meaning of a signal is partly what it enables the system to do, given its history and needs.

Neuroscience can investigate how perception, memory, prediction, and action make symbols usable. Cognitive science can ask how compositional language and shared attention arise. None of this requires a mysterious semantic particle. Nor does it reduce meaning to raw compression. A nervous system may compress regularities, but what a pattern means depends on how it participates in a world of bodies, goals, institutions, and consequences.

Machines and the temptation to overclaim

Supporting/contextual references: [meaning-dennett-1987] [meaning-dretske-1981]

A computer manipulates symbols according to formal rules. It can store a sentence, translate a code, or control a robot, and its physical states can carry information for users and designers. Whether a machine understands is a further question. Correct symbol manipulation may support sophisticated behavior without settling whether there is experience, reference, or a point of view inside the system.

The distinction cuts both ways. Saying that a machine lacks human meaning because it is made of silicon assumes a particular theory of substrate. Saying that every computation understands because it has states assumes a particular theory of functional organization. Better questions ask what capacities are present: Can the system use representations flexibly? Can it learn norms, correct errors, and integrate its own history? Technical information measures cannot answer those questions alone.

Why quantum information does not close the gap

Supporting/contextual references: [meaning-wheeler-zurek-1983] [meaning-shannon-1948]

Quantum theory gives information a richer mathematical setting. Qubits, entanglement, and no-cloning constrain how states can be prepared, transformed, and learned. Quantum information can be physically useful in cryptography and computation, and entanglement is a measured structure of correlations. None of those facts turns a wave function into a sentence or an interaction into an act of communication without a protocol.

A quantum state may encode an agent’s expectations, describe an operational preparation, or represent a structure in a deeper theory, depending on interpretation. Measurement produces outcomes that can become records, but the meaning of those outcomes depends on the apparatus and question. “Quantum” intensifies the physics of distinction; it does not supply semantics. Consciousness and language remain separate explanatory problems.

Open research directions

Supporting/contextual references: [meaning-millikan-1984] [meaning-clark-2016] [meaning-dennett-1987]

Live work connects information theory with biology, cognitive science, and philosophy of language. Researchers ask how semantic content can be naturalized without reducing it to arbitrary observer labels, how neural systems ground reference in action, and whether information-theoretic measures can distinguish a useful representation from a merely compressive code. These questions need experiments and formal criteria, not slogans about everything being information.

Artificial intelligence makes the issue practical. What evidence would show that a model has learned a representation with stable reference rather than a statistical association? How should embodiment, social training, and self-correction enter the analysis? There is no settled test that converts successful behavior into a verdict about understanding. Keeping the uncertainty explicit is more informative than granting or denying meaning by intuition.

A distinction worth protecting

Supporting/contextual references: [meaning-shannon-1948] [meaning-dretske-1981] [meaning-millikan-1984]

Information is powerful partly because it travels. Engineers use it to design channels; physicists use it to describe state and entropy; biologists use it to study inheritance and regulation. Meaning can be carried by those structures, but portability of mathematics does not erase differences in purpose. A measure of uncertainty is not a theory of reference, just as a map’s file size is not its geography.

The careful conclusion is positive rather than dismissive. Meaning is real when physical distinctions participate in practices of interpretation, control, memory, and action. It is not a ghostly ingredient, and it is not exhausted by a count of bits. Asking what a signal means means asking who or what can use it, under which norms, with what consequences. That question begins where Shannon’s measure leaves off.

The practical boundary

Supporting/contextual references: [meaning-morris-1938] [meaning-dennett-1987] [meaning-clark-2016]

The distinction is not merely verbal. A thermostat can carry information about temperature and regulate a room without understanding the word temperature. Human meaning depends on learned practices, shared norms, and the capacity to treat a sign as about something. Those capacities have physical conditions, but they are not identical to the semantic achievement.

A useful test is to ask what would change if the pattern changed. A circuit that stops a motor has a control role; a community that revises a word alters coordinated behavior; a random string can change its Shannon quantity without changing anything for a system that does not use it. A sign can matter to one system and not another, so meaning is relational without being private whim. Substrate constrains what can be stored and transmitted, while practice constrains correct use.

Information gives us a vocabulary for distinctions; meaning tells us why some distinctions guide action, answer questions, or become reasons. A written sentence has a measurable carrier and a role within a language community. Damage can destroy access to meaning, while a perfect copy can remain meaningless to a system without the practice needed to use it. Meaning is not less physical for being relational; it is more specific.

Sources & references

Supporting/contextual references, not claim-level proof.

1. Claude E. Shannon — A Mathematical Theory of CommunicationBell System Technical Journal 27(3), 379–423; 27(4), 623–656, 1948.
2. Charles W. Morris — Foundations of the Theory of SignsUniversity of Chicago Press, 1938.
3. Fred Dretske — Knowledge and the Flow of InformationMIT Press, 1981.
4. Ruth Garrett Millikan — Language, Thought, and Other Biological CategoriesMIT Press, 1984.
5. Daniel C. Dennett — The Intentional StanceMIT Press, 1987.
6. John Archibald Wheeler and Wojciech H. Zurek (eds.) — Quantum Theory and MeasurementPrinceton University Press, 1983.
7. Andy Clark — Surfing UncertaintyOxford University Press, 2016.