Is the strangeness of measurement really quantum?

Paper 18 · Pødenphant Lund (2026) · Read on Zenodo

A controllable toy lets me ask which "spooky" measurement effects are special to physics.Some of the famously strange things about measuring a quantum system — the answer depends on the order you ask, the act of looking seems to disturb what you were looking at, a fuzzy state appears to snap into one value the moment you check — might not be special to physics at all. They might just be what any system does when it has to weigh competing possibilities and then commit to one. This is a hypothesis, offered as a lens, not a claim that the brain or a language model is somehow quantum. To test the lens I needed something I could look inside completely while it makes up its mind. A language model is exactly that.

The trick: a system you can see inside

The hard part of the measurement problem is that you usually cannot see the "before." A particle's state is hidden until you measure it, and that hiddenness is part of the puzzle. The same is true in psychology: when someone answers two questions in a row, you see the two answers, never the half-formed leaning that existed before the first answer came out.

A language model is different. Just before it commits to its next word, it holds a spread of candidate answers, each with a weight. With the right access you can read that spread directly. So you can look at the leaning before it resolves, ask a question, and then look again to see exactly how the asking changed it. No particle and no person lets you do that. That is what makes the model a useful stand-in, a controllable model of measurement, even though it is an ordinary classical computer with nothing quantum about it.

The question is deliberately modest. It is not "what is reality made of?" It is: when this fully-visible classical system commits to an answer, which of the famous measurement effects show up, and which do not? Whatever shows up cannot be the quantum part, because there is no quantum part here. Whatever refuses to show up is a candidate for the thing that really is special to physics.

What carried over (and so is probably not special to physics)

Read this way, the model reproduces a surprising amount of the "measurement mystique" from nothing but a classical system conditioning on its own previous answers:

• The act of looking disturbs the answer. Measuring shifts what the model commits to next, and the shift comes mostly from the act of committing itself, not from which answer it lands on.
• Order matters. Ask A then B and you can get a different result than B then A. In the tested pairs, most flipped with order.
• Checking sharpens. Forcing one answer pulls the next one toward a single committed value, the same way a measurement seems to collapse a fuzzy state.
• The "quantum-like" judgment statistics. The model shows the order-effect signatures that the quantum-cognition field uses to model human judgment — including the precise balance physicists call the QQ equality. (These were already known to be reproducible without real quantum physics; here they confirm the model is in the right regime.)
• How you frame the question changes the answer. Ask the same proposition as Yes/No, as True/False, or on a 0–100 scale and you get different commitments. Ask "is a virus alive?" as a yes/no and the model says No; ask for a likelihood and it says 0.89. Forcing a binary squashes "maybe" into a denial; a graded scale keeps the maybe.

The careful conclusion: if a plain classical system produces all of this, then these effects are generic to any bounded system that has to commit a result. The strangeness around them dissolves.

What did not carry over (the genuinely quantum candidate)

The model stopped at a clear line. Two of the deepest quantum signatures simply did not appear.

• Contextuality. There is a careful test (Contextuality-by-Default) that strips away the ordinary order-and-disturbance effects and asks whether anything genuinely quantum remains. The answer was a clean negative in every model and every scenario tested. This is the load-bearing result: the ordinary disturbance is there, but the deep thing is not.
• Phase, or "coherence." Real quantum interference rides on a hidden phase, like overlapping waves. The model's interference is just a kind of classical probability that depends on order and context, with no wave-like phase underneath.

So the line the toy draws is precise. Most of what feels spooky about measurement (disturbance, order-dependence, snapping into one value) is what any committing inference system does. The part that does not come along for free, and so is the best candidate for the genuinely quantum residue, is this phase-carrying, context-deep structure.

A controllable demonstration of Bell's loopholes

Bell's famous theorem says: if the world is realistic, local, and your measurement choices are free, a certain correlation score cannot exceed 2; real experiments exceed it, so one of those assumptions must give. Because everything inside the model is visible, you can switch each assumption on and off and watch the score move, all in one system.

With everything closed, the score sits at about 2, exactly where a well-behaved classical system should. Drop one assumption (let one computation secretly see both questions, or let the setup quietly correlate the answers) and the score climbs to roughly 2.8. The point is not that any single number is surprising; it is that you can see, inside one transparent system, exactly what each shortcut costs. This is a demonstration of the price of each loophole, not a discovery about physical reality.

Why frame it so carefully

It would be easy, and wrong, to read this as "the brain is quantum" or "language models are quantum." I am doing the opposite. The model is an ordinary classical computer; its "origin" is a pile of text, not the laws of physics. The resemblance to quantum measurement is structural, a useful lens for sorting which effects need a quantum explanation and which do not. The system proves nothing about physics. What it does, and does cleanly, is draw the line between the part of the mystery that any deciding machine would show and the part that still belongs to physics alone.

The cite

Read on Zenodo → · Technical version · Dansk version

Related on this site:

• Paper 10 (Race Architecture) — the "everything is a race that resolves into one outcome" picture this paper measures directly.
• Paper 1 (Friction Theory) — the underlying framework: competing routes, and what it costs to commit to one.
• Paper 13 (Operational FT) — how a decision opens up and resolves; the timeline this paper watches inside the model.
• Paper 25 (The Delta) — using a fully-visible model as a clean control for studying the human mind; the same "look inside" move.