Plate Nº 59 · recorded November 18, 2026
PhysicsReported finding
Quantum observers can disagree, but only within strict limits
A 2026 theorem shows how two quantum observers can assign different probabilities to one test while sharing certainty, and why a classical record of their measurements restores agreement.
By Marcus Bennett5 min read971 words
In brief
- 2026: The PNAS paper defines “common certainty” and proves when compatible quantum assessments must agree.
- 1/3 vs. 2/3: Alice and Bob can assign different chances to one yes-or-no test while remaining certain of each other’s estimates.
- 1976: Robert Aumann’s classical agreement theorem made common knowledge the benchmark for agreement.
- 1/3: A classical register of measurements and outcomes brings both observers to the same estimate.
- Boundary: Under one underlying state, mutual certainty cannot support opposite claims about a property.

By Marcus Bennett
In a three-part quantum system, one observer can estimate a yes-or-no test at one-third while another estimates it at two-thirds, yet each can remain certain of the other’s assessment.
A report dated November 17, 2026 says this divergence can persist when observers apply quantum probability rules without sharing the same measurement history. María García Díaz and Giannicola Scarpa of Universidad Politécnica de Madrid, with Adam Brandenburger of New York University, report the result in “The quantum agreement theorem,” a 2026 paper in the Proceedings of the National Academy of Sciences.
The study identifies the conditions that allow disagreement and shows how an ordinary record of measurements and outcomes can restore agreement.
What does “agree to disagree” mean here?
Two observers can start with the same description of a quantum system, follow the same probability rules and assign different chances to the same outcome. They can still be certain of each other’s assessments, rather than merely predicting the same result.
The benchmark is Robert Aumann’s 1976 agreement theorem. Under classical probability, people who begin with the same initial probabilities cannot sustain different updated estimates once those estimates become common knowledge, even if each person has private information.
Common knowledge means more than knowing another person’s answer. Each person must know the other’s estimate, know that the other knows it, know that the other knows that they know, and continue through every level.
The new study defines a quantum counterpart, “common certainty.” The researchers prove that this condition forces agreement when the observers’ measurements are compatible with each other and with the property under assessment. Compatibility means the corresponding mathematical operators commute: changing their order makes no difference.
Quantum mechanics adds a complication. A measurement can change the system’s state, and that change can alter the probability the observers want to calculate.
Takeaway: Classical common knowledge compels agreement; the quantum condition permits disagreement only when measurement effects open that gap.
Why can one-third and two-thirds coexist?
The model divides a quantum system into three parts. Alice and Bob each measure one part, then estimate the chance that a separate yes-or-no test involving Alice’s part and an inaccessible third part will return “yes.”
The shared state combines two possibilities that can interfere, a property called quantum coherence. Coherence lets the alternatives affect one another’s calculated probabilities.
- Alice: Her measurement distinguishes the two possibilities and removes their coherence. Whichever outcome she gets, her measurement selects one alternative, leaving her with a one-third chance that the proposed test returns “yes.”
- Bob: His measurement does not distinguish the alternatives and produces the same outcome for both. His calculation therefore retains their coherence, pushing his estimate for a “yes” result to two-thirds.
Each observer updates the shared starting state using only their own outcome. Bob can be certain that Alice assigns one-third whichever outcome she receives, while Alice can be certain of Bob’s estimate. This mutual certainty extends through every level, even though their calculations do not use the same measurement history.
Bob’s calculation does not include the state change caused by Alice’s measurement. The model supplies no extra information about which measurements have already occurred, and it does not describe experimenters who have fully compared their procedures and records.
Alice’s and Bob’s measurements remain compatible with each other: performing them in either order gives the same outcome probabilities. The incompatible pair is Alice’s measurement and the yes-or-no test they assess. By removing the interference that affects the chance of “yes,” Alice’s measurement can change the probability both observers seek.
Takeaway: The one-third and two-thirds estimates remain mutually certain because each calculation conditions on a different measurement history.
How does a classical record restore agreement?
The researchers extend the model with a classical register, an ordinary record of which measurements occurred and which outcomes followed. In the example, the measurement and recording process removes the coherence between the two alternatives in the original system.
Using that resulting state and the recorded outcomes, both observers calculate one-third. Bob need not learn Alice’s particular result: once her measurement enters the state used for his calculation, the interference behind his higher estimate disappears.
More generally, the team proves that assessments based on the recorded state agree whenever they satisfy the common-certainty condition. The researchers write that agreement can emerge through “the physical act of classical recording.”
Quantum disagreement still has a firm boundary. Within this framework, Alice cannot be certain that a property holds while also being certain that Bob is certain it does not, provided every claim refers to the same underlying state.
Takeaway: A shared physical record can align quantum estimates without forcing both observers to know every recorded detail in advance.
What changes for researchers?
For researchers comparing quantum measurements, the theorem identifies a concrete condition for agreement: retain a classical record of the measurements and outcomes used in each assessment. Merely knowing that another observer holds a different estimate does not guarantee that the two probabilities match.
The result also preserves a practical warning. Analysts should not treat mutual certainty as proof that two observers used identical information, because Alice’s and Bob’s calculations in the model deliberately omit different measurement effects.
This is a mathematical result about a model, not a reported laboratory trial. Its scope remains explicit: common certainty applies to separate calculations, and the model does not describe observers who have already compared every procedure and record.
The paper, “The quantum agreement theorem,” appeared in the Proceedings of the National Academy of Sciences in 2026 under DOI 10.1073/pnas.2613021123.
Takeaway: Quantum certainty can permit disagreement, but a shared physical record can close the gap.
via phys.org/rss-feed/ (Source)
More from Marcus Bennett
Show full bio
News editor covering regulation, policy and their impact on general science news at SciBeat.
493 articles
Nearby plates
- Physicists Beat Heisenberg's Uncertainty Limit by Using Hindsight
- After 25 Years, Physicists Crack the Entangled Measurement Puzzle for W States
- 51 Qubits and an Error Bar: A New Way to Check Quantum Machines
- Near-Identical Photons Bring Quantum Communication Closer
- Physicists Propose a Supercurrent Way to Read Qubits Without Wrecking Them