Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Physics Today

Bayesian Probability and One Bad Apple

Bruno Lunelli

Published Apr 1, 2003• 0 citations
Publisher

Abstract

The brilliant, attention-capturing sentences at the beginning of Michael Berry’s “Singular Limits” (Physics Today, May 2002, page 10) appear untenable unless one considers conditional (Bayesian) probability. In fact, biting an apple and finding no maggot may indicate either the worst or the best experience of the apple eater. Respectively, the eater may have swallowed the entire maggot with some bite or no maggot at all. The outcome depends on a preexisting condition: the presence, or absence, of a single apple inhabitant.Real things may be even more complex: A particularly unfortunate eater may have gotten an apple with multiple maggots. The situation described also appears to be a suitable illustration of the collapse of probability by observation.© 2003 American Institute of Physics.

DOI
10.1063/1.1580035

Related papers

The role of probability in statistical physics

Romano Scozzafava

In this expository paper we discuss the role of probability in statistical physics: it should be that of a general tool guiding induction and must not be restricted only to the empirical interpretation coming from its evaluation by an observed frequency. It is true that in many cases the value of a probability can be sensibly expressed by a suitable frequency, but this does not entails that the latter can be taken as definition of probability. Moreover, in each situation in which one thinks that probabilities are physically real things and believes that something which is only “very probable” can be regarded as being “practically certain”, some “small” probabilities are actually ignored: this attitude of mind may render illegitimate also the probabilistic interpretation of the relevant physical laws. A “law” which is based on a probabilistic interpretation cannot aim at explaining why a fact that we forecast will occur, but rather at explaining why we forecast that this fact will occur. Finally, we emphasize the role of conditional probability, and we show that looking at it from the right perspective allows to avoids also classic paradoxical aspects of quantum mechanics, without the need to resort to the interpretation based on the dualism “particle-wave”.

Transport Theory and Statistical Physics•Jan 1, 2000•5 citations

Measured responses to quantum Bayesianism

Michael Nauenberg

Some time ago, John Bell and I wrote a tongue-in-cheek article on the interpretation of quantum mechanics;11. J. S. Bell, M. Nauenberg, in Preludes in Theoretical Physics, A. de Shalit, H. Feshbach, L. van Hove, eds., North Holland, Amsterdam (1966), p. 270;reprinted in J. S. Bell, Speakable and Unspeakable in Quantum Mechanics, Cambridge U. Press, New York (1987), p. 24. in it we remarked that only a minority of physicists had any interest in the topic and that the typical physicist thought he would understand it “if ever he can spare twenty minutes to think about it.” According to David Mermin, however, that situation has changed dramatically, and “new interpretations [of quantum mechanics] appear every year.”Actually, most textbooks on the subject continue to be based on the well-established Born interpretation of quantum mechanics, which postulates that the absolute square of the Schrödinger wavefunction, projected onto a particular quantum state, gives the probability for the occurrence of that state after an appropriate measurement has been made. Mermin indicated that the problem with the Born interpretation involves a confusion, “prevalent among physicists,” regarding the nature of probability. He presented his preference for a “personalist Bayesian view,” which leads separate “agents,” like Eugene Wigner and his friend, to make different probability assignments based on their beliefs. But as Richard Feynman remarked, “Nature does not know what you are looking at, and she behaves the way she is going to behave whether you bother to take down the data or not.”22. R. P. Feynman, R. B. Leighton, M. Sands, The Feynman Lectures on Physics, vol. 3, Addison-Wesley, Redwood City, CA (1989), p. 3–7.Indeed, as applied to experiments, probability simply corresponds to the relative frequency for the occurrence of a given final state after identical states are prepared initially and measurements are made by appropriate detectors—devices that, through an irreversible macroscopic process, provide a permanent record of the outcome. Hence, the results of an experiment are independent of any observers, whose only role is to set up appropriate measurement devices, calculate the probabilities from quantum theory, and then write papers based on the recorded results.As Max Born also emphasized, in classical mechanics the situation is analogous, when one takes into account that in practice initial conditions in an experiment are known only imprecisely.33. M. Born, Dan. Mat. Fys. Medd. 30 (2), 1 (1955). Hence the observed evolution of the system is also determined by a probability function that satisfies the Liouville equation. But in that case, after a measurement is performed, no one talks about the “collapse” of this classical function, whereas in quantum mechanics it is common to use that controversial expression for the Schrödinger wavefunction.Mermin discussed a “shifty split” between the classical and the quantum domain, a split that supposedly “haunts quantum mechanics.” He claimed that “regardless of what is split from what, all versions of the shifty split are vague and ambiguous.” However, that is not always the case, as a simple example shows. Consider an initial quantum wavepacket for a particle under an attractive central force that varies inversely with the square of the radial distance—the classic problem of planetary motion from the viewpoint of quantum mechanics, for example, or an excited electron in a Rydberg atom. For a finite time, the wavepacket will follow a classical Keplerian elliptic orbit, but like the Liouville distribution for the analogous classical problem, from the outset it will disperse due to the spread in initial position and momentum.For the classical distribution to be in close correspondence with the quantum wavepacket, its initial spread is assumed to satisfy the uncertainty relation. Eventually, however, the head of the wavepacket, like in the classical distribution, will catch up with its tail. Afterwards, further spreading of the wavepacket leads to interference effects between the two components, and the classical–quantum correspondence ceases to be valid. The time when such interference effects first appear is of order nτ/3, where n is the mean principal quantum number of the wavepacket and τ is the Kepler period.44. M. Nauenberg, Phys. Rev. A 40, 1133 (1989). https://doi.org/10.1103/PhysRevA.40.1133 Experimentally, that transition has been observed to agree with theory in Rydberg atoms55. M. Nauenberg, C. Stroud, J. Yeazell, Sci. Am., June 1994, p. 44. https://doi.org/10.1038/scientificamerican0694-44 for n = 100, τ = 152 picoseconds, and nothing shifty about this classical–quantum correspondence is found. For planetary orbits, however, n is of order 1070, and therefore the answer to Albert Einstein’s often repeated remark “Is the moon there when nobody looks?” (see David Mermin, Physics Today, April 1985, page 38) is yes.REFERENCESSection:ChooseTop of pageREFERENCES <<1. J. S. Bell, M. Nauenberg, in Preludes in Theoretical Physics, A. de Shalit, H. Feshbach, L. van Hove, eds., North Holland, Amsterdam (1966), p. 270; Google Scholarreprinted in J. S. Bell, Speakable and Unspeakable in Quantum Mechanics, Cambridge U. Press, New York (1987), p. 24. Google Scholar2. R. P. Feynman, R. B. Leighton, M. Sands, The Feynman Lectures on Physics, vol. 3, Addison-Wesley, Redwood City, CA (1989), p. 3–7. Google Scholar3. M. Born, Dan. Mat. Fys. Medd. 30 (2), 1 (1955). Google Scholar4. M. Nauenberg, Phys. Rev. A 40, 1133 (1989). https://doi.org/10.1103/PhysRevA.40.1133, Google ScholarCrossref, CAS5. M. Nauenberg, C. Stroud, J. Yeazell, Sci. Am., June 1994, p. 44. https://doi.org/10.1038/scientificamerican0694-44, Google ScholarCrossref© 2012 American Institute of Physics.

Physics Today•Nov 30, 2012•0 citations

Measured responses to quantum Bayesianism

Art Hobson

Because it’s important that physicists discuss the problems of quantum foundations, I was delighted to see David Mermin’s commentary on “fixing the shifty split.” But I must disagree with my friend’s view.There are multiple ways of interpreting probabilities.11. A. Hobson, Concepts in Statistical Mechanics, Gordon and Breach, New York (1971), p. 27. In quantum physics, the Bayesian “degree of belief” interpretation can be appropriate for mixed states, but only the frequentist or “ensemble” interpretation applies to the pure states that Mermin discusses.A recent paper by Matthew Pusey and coauthors22. M. F. Pusey, J. Barrett, T. Rudolph, http://arxiv.org/abs/1111.3328v1. concludes that if two different pure quantum states are assigned to a single physical situation, one of the states is objectively wrong. The proof’s main assumption is that any quantum system has some set of real physical properties, labeled λ. Pusey and coauthors show that two different pure states cannot represent the same λ. So, although their paper implies but does not assume that quantum states are physically real, it does assume that each specific physical situation has behind it some kind of physical reality. Chris Fuchs, quoted favorably by Mermin, might dispute the notion that some kind of physical reality actually exists, but most physicists are probably sufficiently realist to grant that notion. Without such a notion, science lies somewhere between solipsism and superstition.Whenever I hear subjective interpretations of quantum physics, I wonder about such questions as the one Mermin quotes from Albert Einstein: Can a mouse collapse a wavefunction? Mermin’s Bayesian response, that a mouse cannot but a physics student can, doesn’t reassure me. Were wavefunctions not collapsing before there were physicists?The “Wigner’s friend” paradox, as described in Mermin’s commentary, doesn’t need a subjective interpretation to find a solution. In fact, Eugene Wigner abandoned his interpretation around 1970 once he became aware of the ideas underlying quantum decoherence.33. M. Schlosshauer, Decoherence and the Quantum-to-Classical Transition, Springer, Berlin (2007), p. 364. Suppose Wigner’s friend makes a measurement, the outcome of which is known to the friend but not to Wigner. Regardless of whether the outcome is known, and in fact regardless of the presence of the friend, the measurement device must decohere the quantum system that’s being measured, putting the system into a mixed state of the either/or form, with no interferences, just like a coin that’s been flipped. A definite outcome has thus occurred, and Wigner and his friend will both agree on this. Neither the friend nor Wigner affects the outcome. She has simply “looked” and now knows the outcome, while Wigner doesn’t know. That is not paradoxical and does not require a revision of our understanding of pure quantum states.There’s no reason to give up on a realist interpretation of quantum physics, much less a realist interpretation of nature itself.REFERENCESSection:ChooseTop of pageREFERENCES <<1. A. Hobson, Concepts in Statistical Mechanics, Gordon and Breach, New York (1971), p. 27. Google Scholar2. M. F. Pusey, J. Barrett, T. Rudolph, http://arxiv.org/abs/1111.3328v1. Google Scholar3. M. Schlosshauer, Decoherence and the Quantum-to-Classical Transition, Springer, Berlin (2007), p. 364. Google Scholar© 2012 American Institute of Physics.

Physics Today•Nov 30, 2012•1 citation

Measured responses to quantum Bayesianism

Ching -Hung Woo

David Mermin’s exposition on QBism was excellent. However, it left the reconciliation between subjective and objective probability up in the air. There must be circumstances in which the subjective probabilities of different scientifically trained agents coincide, and hence the subjective and objective approaches also coincide. Quantum mechanics reminds us, if such a reminder is needed, that the human experiences that produce coinciding subjective probabilities, and hence interesting science, form a small subset of all human experiences.© 2012 American Institute of Physics.

Physics Today•Nov 30, 2012•1 citation

Quantum Knowledge, Quantum Belief, Quantum Reality: Notes of a QBist Fellow Traveler

Howard Barnum

I consider the "Quantum Bayesian" view of quantum theory as expounded in a 2006 paper of Caves, Fuchs, and Schack. I argue that one can accept a generally personalist, decision-theoretic view of probability, including probability as manifested in quantum physics, while nevertheless accepting that in some situations, including some in quantum physics, probabilities may in a useful sense be thought of as objectively correct. This includes situations in which the ascription of a quantum state should be thought of as objectively correct. I argue that this does not cause any prima facie objectionable sort of action at a distance, though it may involve adopting the attitude that certain dispositional properties of things are not "localized" at those things. Whether this insouciant view of nonlocality and objectivity can survive more detailed analysis is a matter for further investigation.

arXiv:1003.4555•Mar 23, 2010•7 citations

Probabilities are single-case, or nothing

D M Appleby

Physicists have, hitherto, mostly adopted a frequentist conception of probability, according to which probability statements apply only to ensembles. It is argued that we should, instead, adopt an epistemic, or Bayesian conception, in which probabilities are conceived as logical constructs rather than physical realities, and in which probability statements do apply directly to individual events. The question is closely related to the disagreement between the orthodox school of statistical thought and the Bayesian school. It has important technical implications (it makes a difference, what statistical methodology one adopts). It may also have important implications for the interpretation of the quantum state. 1 1.

•Jan 1, 2004•18 citations
The brilliant, attention-capturing sentences at the beginning of Michael Berry’s “Singular Limits” (Physics Today, May 2002, page 10) appear untenable unless one considers conditional (Bayesian) probability. In fact, biting an apple and finding no maggot may indicate either the worst or the best experience of the apple eater. Respectively, the eater may have swallowed the entire maggot with some bite or no maggot at all. The outcome depends on a preexisting condition: the presence, or absence, of a single apple inhabitant.Real things may be even more complex: A particularly unfortunate eater may have gotten an apple with multiple maggots. The situation described also appears to be a suitable illustration of the collapse of probability by observation.© 2003 American Institute of Physics.
In this expository paper we discuss the role of probability in statistical physics: it should be that of a general tool guiding induction and must not be restricted only to the empirical interpretation coming from its evaluation by an observed frequency. It is true that in many cases the value of a probability can be sensibly expressed by a suitable frequency, but this does not entails that the latter can be taken as definition of probability. Moreover, in each situation in which one thinks that probabilities are physically real things and believes that something which is only “very probable” can be regarded as being “practically certain”, some “small” probabilities are actually ignored: this attitude of mind may render illegitimate also the probabilistic interpretation of the relevant physical laws. A “law” which is based on a probabilistic interpretation cannot aim at explaining why a fact that we forecast will occur, but rather at explaining why we forecast that this fact will occur. Finally, we emphasize the role of conditional probability, and we show that looking at it from the right perspective allows to avoids also classic paradoxical aspects of quantum mechanics, without the need to resort to the interpretation based on the dualism “particle-wave”.
Some time ago, John Bell and I wrote a tongue-in-cheek article on the interpretation of quantum mechanics;11. J. S. Bell, M. Nauenberg, in Preludes in Theoretical Physics, A. de Shalit, H. Feshbach, L. van Hove, eds., North Holland, Amsterdam (1966), p. 270;reprinted in J. S. Bell, Speakable and Unspeakable in Quantum Mechanics, Cambridge U. Press, New York (1987), p. 24. in it we remarked that only a minority of physicists had any interest in the topic and that the typical physicist thought he would understand it “if ever he can spare twenty minutes to think about it.” According to David Mermin, however, that situation has changed dramatically, and “new interpretations [of quantum mechanics] appear every year.”Actually, most textbooks on the subject continue to be based on the well-established Born interpretation of quantum mechanics, which postulates that the absolute square of the Schrödinger wavefunction, projected onto a particular quantum state, gives the probability for the occurrence of that state after an appropriate measurement has been made. Mermin indicated that the problem with the Born interpretation involves a confusion, “prevalent among physicists,” regarding the nature of probability. He presented his preference for a “personalist Bayesian view,” which leads separate “agents,” like Eugene Wigner and his friend, to make different probability assignments based on their beliefs. But as Richard Feynman remarked, “Nature does not know what you are looking at, and she behaves the way she is going to behave whether you bother to take down the data or not.”22. R. P. Feynman, R. B. Leighton, M. Sands, The Feynman Lectures on Physics, vol. 3, Addison-Wesley, Redwood City, CA (1989), p. 3–7.Indeed, as applied to experiments, probability simply corresponds to the relative frequency for the occurrence of a given final state after identical states are prepared initially and measurements are made by appropriate detectors—devices that, through an irreversible macroscopic process, provide a permanent record of the outcome. Hence, the results of an experiment are independent of any observers, whose only role is to set up appropriate measurement devices, calculate the probabilities from quantum theory, and then write papers based on the recorded results.As Max Born also emphasized, in classical mechanics the situation is analogous, when one takes into account that in practice initial conditions in an experiment are known only imprecisely.33. M. Born, Dan. Mat. Fys. Medd. 30 (2), 1 (1955). Hence the observed evolution of the system is also determined by a probability function that satisfies the Liouville equation. But in that case, after a measurement is performed, no one talks about the “collapse” of this classical function, whereas in quantum mechanics it is common to use that controversial expression for the Schrödinger wavefunction.Mermin discussed a “shifty split” between the classical and the quantum domain, a split that supposedly “haunts quantum mechanics.” He claimed that “regardless of what is split from what, all versions of the shifty split are vague and ambiguous.” However, that is not always the case, as a simple example shows. Consider an initial quantum wavepacket for a particle under an attractive central force that varies inversely with the square of the radial distance—the classic problem of planetary motion from the viewpoint of quantum mechanics, for example, or an excited electron in a Rydberg atom. For a finite time, the wavepacket will follow a classical Keplerian elliptic orbit, but like the Liouville distribution for the analogous classical problem, from the outset it will disperse due to the spread in initial position and momentum.For the classical distribution to be in close correspondence with the quantum wavepacket, its initial spread is assumed to satisfy the uncertainty relation. Eventually, however, the head of the wavepacket, like in the classical distribution, will catch up with its tail. Afterwards, further spreading of the wavepacket leads to interference effects between the two components, and the classical–quantum correspondence ceases to be valid. The time when such interference effects first appear is of order nτ/3, where n is the mean principal quantum number of the wavepacket and τ is the Kepler period.44. M. Nauenberg, Phys. Rev. A 40, 1133 (1989). https://doi.org/10.1103/PhysRevA.40.1133 Experimentally, that transition has been observed to agree with theory in Rydberg atoms55. M. Nauenberg, C. Stroud, J. Yeazell, Sci. Am., June 1994, p. 44. https://doi.org/10.1038/scientificamerican0694-44 for n = 100, τ = 152 picoseconds, and nothing shifty about this classical–quantum correspondence is found. For planetary orbits, however, n is of order 1070, and therefore the answer to Albert Einstein’s often repeated remark “Is the moon there when nobody looks?” (see David Mermin, Physics Today, April 1985, page 38) is yes.REFERENCESSection:ChooseTop of pageREFERENCES <<1. J. S. Bell, M. Nauenberg, in Preludes in Theoretical Physics, A. de Shalit, H. Feshbach, L. van Hove, eds., North Holland, Amsterdam (1966), p. 270; Google Scholarreprinted in J. S. Bell, Speakable and Unspeakable in Quantum Mechanics, Cambridge U. Press, New York (1987), p. 24. Google Scholar2. R. P. Feynman, R. B. Leighton, M. Sands, The Feynman Lectures on Physics, vol. 3, Addison-Wesley, Redwood City, CA (1989), p. 3–7. Google Scholar3. M. Born, Dan. Mat. Fys. Medd. 30 (2), 1 (1955). Google Scholar4. M. Nauenberg, Phys. Rev. A 40, 1133 (1989). https://doi.org/10.1103/PhysRevA.40.1133, Google ScholarCrossref, CAS5. M. Nauenberg, C. Stroud, J. Yeazell, Sci. Am., June 1994, p. 44. https://doi.org/10.1038/scientificamerican0694-44, Google ScholarCrossref© 2012 American Institute of Physics.
Because it’s important that physicists discuss the problems of quantum foundations, I was delighted to see David Mermin’s commentary on “fixing the shifty split.” But I must disagree with my friend’s view.There are multiple ways of interpreting probabilities.11. A. Hobson, Concepts in Statistical Mechanics, Gordon and Breach, New York (1971), p. 27. In quantum physics, the Bayesian “degree of belief” interpretation can be appropriate for mixed states, but only the frequentist or “ensemble” interpretation applies to the pure states that Mermin discusses.A recent paper by Matthew Pusey and coauthors22. M. F. Pusey, J. Barrett, T. Rudolph, http://arxiv.org/abs/1111.3328v1. concludes that if two different pure quantum states are assigned to a single physical situation, one of the states is objectively wrong. The proof’s main assumption is that any quantum system has some set of real physical properties, labeled λ. Pusey and coauthors show that two different pure states cannot represent the same λ. So, although their paper implies but does not assume that quantum states are physically real, it does assume that each specific physical situation has behind it some kind of physical reality. Chris Fuchs, quoted favorably by Mermin, might dispute the notion that some kind of physical reality actually exists, but most physicists are probably sufficiently realist to grant that notion. Without such a notion, science lies somewhere between solipsism and superstition.Whenever I hear subjective interpretations of quantum physics, I wonder about such questions as the one Mermin quotes from Albert Einstein: Can a mouse collapse a wavefunction? Mermin’s Bayesian response, that a mouse cannot but a physics student can, doesn’t reassure me. Were wavefunctions not collapsing before there were physicists?The “Wigner’s friend” paradox, as described in Mermin’s commentary, doesn’t need a subjective interpretation to find a solution. In fact, Eugene Wigner abandoned his interpretation around 1970 once he became aware of the ideas underlying quantum decoherence.33. M. Schlosshauer, Decoherence and the Quantum-to-Classical Transition, Springer, Berlin (2007), p. 364. Suppose Wigner’s friend makes a measurement, the outcome of which is known to the friend but not to Wigner. Regardless of whether the outcome is known, and in fact regardless of the presence of the friend, the measurement device must decohere the quantum system that’s being measured, putting the system into a mixed state of the either/or form, with no interferences, just like a coin that’s been flipped. A definite outcome has thus occurred, and Wigner and his friend will both agree on this. Neither the friend nor Wigner affects the outcome. She has simply “looked” and now knows the outcome, while Wigner doesn’t know. That is not paradoxical and does not require a revision of our understanding of pure quantum states.There’s no reason to give up on a realist interpretation of quantum physics, much less a realist interpretation of nature itself.REFERENCESSection:ChooseTop of pageREFERENCES <<1. A. Hobson, Concepts in Statistical Mechanics, Gordon and Breach, New York (1971), p. 27. Google Scholar2. M. F. Pusey, J. Barrett, T. Rudolph, http://arxiv.org/abs/1111.3328v1. Google Scholar3. M. Schlosshauer, Decoherence and the Quantum-to-Classical Transition, Springer, Berlin (2007), p. 364. Google Scholar© 2012 American Institute of Physics.
David Mermin’s exposition on QBism was excellent. However, it left the reconciliation between subjective and objective probability up in the air. There must be circumstances in which the subjective probabilities of different scientifically trained agents coincide, and hence the subjective and objective approaches also coincide. Quantum mechanics reminds us, if such a reminder is needed, that the human experiences that produce coinciding subjective probabilities, and hence interesting science, form a small subset of all human experiences.© 2012 American Institute of Physics.
I consider the "Quantum Bayesian" view of quantum theory as expounded in a 2006 paper of Caves, Fuchs, and Schack. I argue that one can accept a generally personalist, decision-theoretic view of probability, including probability as manifested in quantum physics, while nevertheless accepting that in some situations, including some in quantum physics, probabilities may in a useful sense be thought of as objectively correct. This includes situations in which the ascription of a quantum state should be thought of as objectively correct. I argue that this does not cause any prima facie objectionable sort of action at a distance, though it may involve adopting the attitude that certain dispositional properties of things are not "localized" at those things. Whether this insouciant view of nonlocality and objectivity can survive more detailed analysis is a matter for further investigation.
Physicists have, hitherto, mostly adopted a frequentist conception of probability, according to which probability statements apply only to ensembles. It is argued that we should, instead, adopt an epistemic, or Bayesian conception, in which probabilities are conceived as logical constructs rather than physical realities, and in which probability statements do apply directly to individual events. The question is closely related to the disagreement between the orthodox school of statistical thought and the Bayesian school. It has important technical implications (it makes a difference, what statistical methodology one adopts). It may also have important implications for the interpretation of the quantum state. 1 1.