Alhassanain(p) Network for Heritage and Islamic Thought

Quantum Theory: a Pragmatist Approach

Richard Healey,

Philosophy Department,

University of Arizona,

Tucson, AZ 85721-0027, USA

rhealey@email.arizona.edu

Table of Contents

ABSTRACT 3

1. Introduction 4

2. The objectivity of quantum probabilities 7

2.1 Quantum probabilities are objective 8

2.2 Quantum probabilities do not represent physical reality 10

3. How quantum theory limits description of physical reality 13

4. The relational nature of quantum states 22

4.1  Rovelli’s Relationism 22

4.2 Quantum Bayesian Relationism 22

4.3  Reference-frame Relationism 23

4.4 Agent-situation Relationism and Wave-Collapse 24

4.5 Why quantum probabilities are not Lewisian chances 26

5. The objectivity of physical description in quantum theory 31

5.1 Why violations of Bell Inequalities involve no physical non-locality 31

5.2 Objectivity, Inter-subjectivity and Wigner’s friend 34

6. Conclusion 38

References 41

Notes 45

ABSTRACT

While its applications have made quantum theory arguably the most successful theory in physics, its interpretation continues to be the subject of lively debate within the community of physicists and philosophers concerned with conceptual foundations. This situation poses a problem for a pragmatist for whom meaning derives from use. While disputes about how touse quantum theory have arisen from time to time, they have typically been quickly resolved, and consensus reached, within the relevant scientific sub-community. Yet rival accounts of themeaning of quantum theory continue to proliferate[1]. In this article I offer a diagnosis of this situation and outline a pragmatist solution to the problem it poses, leaving further details for subsequent articles.

1. Introduction

What is it to interpret quantum theory? Addressing this question, van Fraassen ([1991]) characterized the interpretative task as an attempt to say: ‘What is really going on, according to this theory?’ and ‘How could the world possibly be how this theory says it is?’ This ties interpretation directly to representation: it assumes that the theory offers representations and/or descriptions of the physical world. In (Healey [1989], p.6) I expressed sympathy for such a tie as follows: ‘I should like to add [...] that a satisfactory interpretation of quantum mechanics should make it clear what the world would be like if quantum mechanics were true.’ But I continued by noting that it would be inappropriate to criticize a proposed interpretation solely on the grounds that it does not meet this constraint. A theory may further the goals of physics without itself offering novel representations or descriptions of physical reality. If quantum theory is such a theory, then we need an account of how and why it is able to achieve its enormous success. To provide such an account is to offer an interpretation of quantum theory. That is what I set out to do here.

The claim that quantum theory does not itself offer novel depictions of reality may strike some readers as obviously false. What could be more novel than representing the state of a system by a mathematical object such as a wave-function, state vector or density operator, especially when this may represent it as in a superposition, or as entangled with other systems? This is surely quantum theory’s distinctive way of describing physical systems, whether or not the description is complete. But there has long been a rival view according to which quantum states convey knowledge or information concerning a system or ensemble without describing its physical condition. I shall elaborate a version of this view that assigns a two-fold role to the quantum state. It plays its primary role in the algorithm provided by the Born Rule for generating quantum probabilities. The quantum state’s secondary role cannot be so simply described, but here is the general idea. Any application of quantum theory involves claims describing a physical situation[2]. For example, in an application of the theory to predict or explain results of a contemporary two-slit interference experiment involving detection of individual particles, some claims will describe the apparatus, while others will describe the results of the experiment. But while claims concerning where individual particles are detected contributing to the interference pattern are considered permissible (and even essential), claims about which slit each particle went through are typically alleged to be “meaningless”[3]. The secondary role of the quantum state is to offer guidance on the legitimacy and limitations of descriptive claims about a physical situation. The key idea here is that even assuming unitary evolution of the quantum state of system and environment, delocalization of system state coherence into the environment will typically (though not always) render descriptive claims about experimental results and the condition of apparatus and other macroscopic objects beyond  reproach.

I call this interpretation pragmatist for several reasons. First it takes the uses and applications of the theory to have explanatory priority over its representational capacities. In his trenchant critique of contemporary formulations of quantum mechanics, Bell ([1987], p.125) unfavorably compared the theory so formulated to classical mechanics: ‘Of course, it is true that also in classical mechanics any isolation of a system from the world as a whole involves approximation. But at least one can envisage an accurate theory, of the universe, to which the restricted account is an approximation.’ I doubt that onecan envisage a detailed and accurate representation of the universe within classical mechanics. Obtaining and using a complete and accurate mathematical model of the universe within classical mechanics would vastly exceed the combined observational and cognitive capacities of humanity or any other physically realizable community of agents, while only in use would such a mathematical model represent anything. But it does not constitute a criticism of a formulation of quantum theory that within it one cannot envisage a complete and accurate representation of the universe, since no successful use or application of quantum theory to cosmology or anywhere else requires that one be able to do so.

The second pragmatist motif concerns the interpretation of quantum probabilities, which are taken to be neither subjective nor straightforwardly objective, but function as a source of authoritative advice to an agent on what to expect and so how to act in specific physical situations. Probabilities derived from the Born Rule do not describe statistical frequencies, even  in ideal infinite ensembles: nor do they describe objective chances of individual events. But to accept quantum theory is to commit oneself to apportioning one’s partial beliefs in accordance with the probabilities generated by the Born Rule as applied to a quantum state appropriate to one’s physical circumstances. I begin to spell this out in more detail in section 2 below.

While not itself issuing descriptive claims about physical reality, quantum theory does advise an agent on the scope and limitations of descriptive claims it may make in a given situation.[4]The advice does not consist in declaring some such claims simply meaningless and so impermissible while others are meaningful and legitimate. Instead the theory places limitations on the inferential power of claims pertaining to the physical situation in which the agent finds itself, or which it represents itself as occupying. Now it is characteristic of pragmatist approaches to meaning to take the content of a descriptive claim to derive ultimately from its inferential relations to other claims and commitments rather than from how it corresponds to the reality it purports to represent.[5]Accepting quantum theory means following its advice to limit the inferential power associated with descriptive claims that may be appropriate in a specific physical situation. So the theory modifies the content of those claims.

Bell ([1987], p.41) introduced the term ‘beable’ (in contrast to quantum theory’s ‘observable’) to apply to things ‘which can be described in “classical terms”, because they are there. The beables must include the settings of switches and knobs on experimental equipment, the current in coils, and the readings of instruments.’ (p.53) He emphasized that by ‘classical terms’ he (following Bohr) ‘refers simply to the familiar language of everyday affairs, including laboratory procedures, in which objective properties -beables - are assigned to objects.’ (p.41)  This at least suggests that a claim about current (for example) derives its content in part from the primitive semantic fact that ‘current’ refers to (the value of the) current, taken to be intelligible independently of one’s disposition to countenance (defeasible) inferences involving this claim (such as the inference that the current consists in the motion of tiny electrically charged particles through a metal, or that it has a source if the current is not zero). More importantly, it assumes that acceptance of quantum theory can in no way modify the content of a claim about beables. But a pragmatist may question that assumption. In section 3 I will offer an alternative account of judgments an agent using quantum theory may make about its physical situation that allows for modification of the content even of claims about its macroscopic environment, including the readings of instruments.

2. The objectivity of quantum probabilities

Any attempt to understand quantum theory must address the significance of probabilities derived from the Born Rule, which I write as follows

probρ (A εΔ )=Tr( ρP A [Δ])            (Born Rule)

whereA is a dynamical variable (an “observable”) pertaining to a systems , ρ represents a quantum state of that system by a density operator on a Hilbert space Hs , Δ is a Borel set of real numbers (soA εΔ states that the value ofA lies in Δ), andP A[Δ] is the value for Δ of the projection-valued measure defined by the unique self-adjoint operator on Hscorresponding toA. Born probabilities yielded by systems’ quantum states are the key to successful applications of quantum theory to explain and predict natural phenomena involving them. If one denies that the quantum state describes or represents the physical properties or relations of any system or ensemble of systems, then its main job is simply to yield these probabilities. But what kind of probabilities are these, and what, exactly, are they probabilities of?

       If one clear conclusion has been established by foundational work, it is that not every probability derivable by applying the Born rule to a system with quantum state ρ can be taken as a quantitative measure of ignorance or uncertainty of the real-numbered value of a dynamical variable on that system. Born probabilities are not analogous to probabilities in classical statistical mechanics in that they cannot be jointly represented on any classical phase space: quantum observables are not random variables on a common probability space.[6]However, as I expressed it the Born rule specifies, for any state ρ, a probability for each sentence of the formS : The value ofA lies in Δ. The traditional way to resolve the resulting tension is to take each instantiation of the Born rule to observableA to be (perhaps implicitly) conditional onmeasurement ofA , and to assume or postulate that only observables represented by commuting operators can be measured together.[7]Whether this resolution is satisfactory has been the topic of a heated debate.[8]I will address aspects of this in the next section, which offers an account of what the Born probabilities are probabilities of. But whatever they concern, what kind of probabilities are these?

       It is common to classify an interpretation of probability as either objective or subjective. Accounts of probability in terms of frequency, propensity or single-case chance count as objective, while the personalist Bayesian interpretation counts as subjective. But then how should one classify classical (Laplacean), logical and “objective” Bayesian notions of probability? It will be best to leave the tricky issue of objectivity aside for a while, so I begin instead by classifying accounts of probability on the basis of their answers to the question ‘Does a probability judgment function as a description of anything in the natural world?’ von Mises’s ([1922]; [2003], p.194) answer to this question was clear:

Probability calculus is part of theoretical physics in the same way as classical mechanics or optics, it is an entirely self-contained theory of certain phenomena

So was Popper’s ([1967], pp.32-3)

In proposing the propensity interpretation I propose to look upon probability statements as statements about some measure of a property (a physical property, comparable to symmetry or asymmetry) of the whole experimental arrangement; a measure, more precisely, of a virtual frequency

These are expressions of what I will call anatural property account of probability.

       Accounts of probability as a natural property typically take this to be a property of something in the physical world independent of the epistemic state of anyone making judgments about it. When de Finetti wrote in the preface to his ([1974]) ‘PROBABILITY DOES NOT EXIST’, this was what he meant to deny.  But he wrote elsewhere ([1968], p.48) that probability

means degree of belief (as actually held by someone, on the ground of his whole knowledge, experience, information) regarding the truth of a sentence, or eventE (a fully specified ‘single’ event or sentence, whose truth or falsity is, for whatever reason, unknown to the person).

and it is at least plausible to suppose that an actual degree of belief is a natural property of the person holding it. If so, even the arch subjectivist de Finetti here adopts a natural property account of probability! Of course, he would insist that different persons may, and often do, hold different beliefs, which makes probability personalist - varying from person to person - and to that extent subjective.

       On other “subjectivist” views, an agent’s degrees of belief count as probabilities only in so far as its overall epistemic state meets a normative constraint ofcoherence ,[9]since otherwise these partial beliefs will not satisfy an analog of Kolmogorov’s ([1933]) axioms defining probability mathematically as a finitely additive, unit-normed, non-negative function on a field of sets. Ramsey ([1926]), for one, took probability theory as a branch of logic, the logic of partial belief and inconclusive argument.[10]So viewed, probability theory offers an agent prescriptions for adjusting the corpus of its beliefs so that its total epistemic state meets minimal internal standards of rationality - standards that are nevertheless met by the total epistemic state of few if any actual agents. A probability judgment made by an agent then counts as an expression of its partial degree of belief and a commitment to hold its epistemic state to this minimal standard of rationality. Such a probability judgment does not function as a description of the agent’s own belief state, and is certainly not a description of a natural property of anything else in the physical world.

       On the present approach, quantum probabilities given by the Born rule do not describe any natural property of the system or systems to which they pertain, or of any other physical system or situation: nor is it their function to describe any actual agent’s state of belief, knowledge or information. Their function is to offer advice to any actual or hypothetical agent on the extent of its commitment to claims expressible by sentences of the formS : The value ofA ons lies in Δ - roughly, what degree of belief or credence to attach to such a claim.

2.1 Quantum probabilities are objective

Consider as an example the experiment of Tonomuraet al. ([1989]) in which the positions of electrons are detected in an electron-biprism version of the two-slit experiment by the discrete, localized flashes they produce on a sensitive screen as each makes its contribution to the classic interference pattern. Before running the experiment, an experimenter does not know whether the following statementS 34 is true: The position of the 34th electron to contribute to the interference pattern is on the left hand side of the screen. Afterwards he or she can check its truth-value by watching the video that recorded each individual flash as it occurred. Assuming the experiment has been correctly set up so that the whole apparatus up to and including the screen has exact left-right symmetry, the Born probability ofS 34 will be ½. Beforehand, an experimenter who makes this assumption and accepts quantum theory should therefore believeS 34 to the same degree that she disbelieves it. This partial belief will dispose the experimenter to behave in ways she would not have behaved if she had taken the Born probability ofS 34 to be .99: she may accept a bet onS 34 she would have declined, or she may simply decide it is not necessary to readjust the apparatus to get a more symmetric interference pattern.

       This is how quantum probabilities serve as what Bishop Butler famously called the very guide of life. It is quantum theory’s great achievement to have made available such a wonderfully reliable guide of such extraordinarily wide applicability. But note that to be guided by quantum theory in this way one needs to know more than just the Born rule - one needs to know what system or type of system to apply it to, and what quantum state to assign to that system. This kind of “know how” is a prerequisite for the successful application of any physical theory, classical as well as quantum. The success of quantum theory is due in large part to the hard-won acquisition of this kind of knowledge by physicists, which has often required great originality and ingenuity. Once acquired, much of it can be conveyed to others and taught to students, although applying Born probabilities to novel situations remains a skill that cannot be mastered by rote learning. But the important point is that knowing what quantum state to assign to a system in order to apply the Born rule constitutes objective knowledge, tacit or otherwise.[11]

       While there are disputes about what quantum state to assign in a particular situation, these are typically resolved by the same kind of debates within the relevant scientific community as those surrounding the establishment of a novel biochemical structure. Exceptions to this generalization will be discussed in section 4.3. They are important because they illustrate respects in which quantum states, unlike classical states, are relational. But not everything that is relational is subjective, as debtors and widows are only too aware. What quantum state to ascribe to a system is not at the whim of each agent’s subjective beliefs, and nor are the Born probabilities consequent on this ascription. There are at least three reasons why these probabilities are objective.

(1) There is widely shared agreement on them within the scientific community

(2) A norm is operative within that community requiring resolution of any residual disagreements

(3) This norm is not arbitrary but derives directly from the scientific aims of prediction, control and explanation of natural phenomena.

(Details of the derivation must here remain a matter for further investigation.)

Each of these is a reason both for why the Born rule is an essential part of the objective content of quantum theory and for why quantum state ascriptions and the consequent Born probabilities are themselves objective.

2.2 Quantum probabilities do not represent physical reality

  After formulating a minimalist account of truth, Wright ([1992]) presents considerations that may incline one to deploy a richer notion of truth as correspondence to objective reality in some domain. By applying these considerations to ascriptions of quantum probabilities we can further articulate the sense in which these are objective even though they do not describe any natural property of or associated with quantum systems.

         The first consideration stems from the Cognitive Command constraint, of which this is an abbreviated version of Wright’s first ‘extremely rough’ formulation:

A discourse exhibits Cognitive Command if and only if it isa priori that differences of opinion arising within it can be satisfactorily explained only in terms of “divergent input”, “unsuitable conditions”, or “malfunction”. (pp.92-3)

He gives this later formulation to qualify and simplify his first formulation:

It isa priori that differences of opinion formulated within the discourse, unless excusable as a result of vagueness in a disputed statement, or in the standards of acceptability, or variation in personal evidence thresholds, so to speak, will involve something which may properly be regarded as a cognitive shortcoming. (p.144)

What is the idea of the Cognitive Command constraint, and what motivates it? Here is what Wright says:

The formulation offered is an attempt to crystallise a very basic idea we have about objectivity: that, where we deal in a purely cognitive way with objective matters, the opinions which we form are in no sense optional or variable as a function of permissible idiosyncracy, but arecommanded of us - that there will be a robust sense in which a particular point of viewought to be held, and a failure to hold a particular point of view can be understood only as a rational/cognitive failure. (p.146)

This nicely captures the sense in which I claim quantum probabilities are objective. Indeed, one can locate a twofold source of the command in this case. Quantum theory itself commands that quantum probabilities conform to the Born rule: the community’s collective evidence-based judgment commands use of a particular quantum state in the Born rule. Neither command is arbitrary: the authority in each case rests ultimately on experimental and observational results  and collective judgment of their evidential bearing. But Wright continues

It is tempting to say that this just is, primitively, what is involved in thinking of a subject matter as purely objective, and of our mode of interaction with it as purely cognitive; and that the Cognitive Command constraint, as formulated, is merely what results when the basic idea is qualified to accommodate various germane kinds of vagueness. [...] But [...] the truth is that the constraint does not reflect a wholly primitive characteristic of the notions of objectivity and cognitive engagement but derives its appeal, at least in part, from a truism to do with the idea ofrepresentation . For to think of oneself as functioning in purely cognitive mode, as it were, is, when the products of that function are beliefs, to think of oneself as functioning in representational mode (ibid .)

Reflection on the epistemic function of probability ascriptions should prompt one to question this “truism”. For while one is clearly functioning in cognitive mode when assessing probabilities, the products of that function arepartial beliefs, and while each of these does indeed have some kind of representational content, only in the case of derivative, higher-order applications does this concernprobabilities - in the fundamental situation, the content of each partial belief represents a possible state of the world free of any natural “probability properties”.[12]The constitutive function of quantum probability statements is not to represent certain probabilistic aspects of the physical world, but to guide agents in forming appropriate partial beliefs about non-probabilistic aspects of the physical world. The intuition Wright takes his Cognitive Command constraint to express survives undercutting of any possible justification by appeal to alleged truisms about representation.

       Wright introduces a second consideration favoring deployment of a richer notion of truth as correspondence to objective reality in some domain.

Let thewidth of cosmological role of the subject matter of a discourse be measured by the extent to which citing the kinds of states of affairs with which it deals is potentially contributive to the explanation of thingsother than , orother than via , our being in attitudinal states which take such states of affairs as object [...]. The crucial question is [...] what else there is, other than our beliefs, of which the citation of such states of affairs can feature in […] explanations. (pp.196-7)

Ascriptions of quantum probability have very narrow cosmological role. While they may and do play a role in a huge variety of applications of quantum theory in prediction as well as explanation, in each case the contribution of quantum probabilities is indeedvia an agent’s being in attitudinal states which take quantum probabilities as object. Someone may object that it is a basic role of quantum probabilities to explain frequencies observed, say, in experimental tests of Bell inequalities. But of course a claim about frequencies follows from a claim about probabilitiesonly with a certain probability : so a judgment that an observed frequency is explained by a quantum probability itself proceedsvia an agent’s being in attitudinal states which take quantum probabilities as object.

       The narrow cosmological role of quantum probability statements provides further support for the conclusion that these do not represent natural properties. But it does nothing to undermine theobjectivity of quantum probabilities.

       Despite his avowed subjectivism about probability, David Lewis ([1980]) undertook to offer a subjectivist’s guide to a kind of objective probability he called chance.

Along with subjective credence we should believe also in objective chance. The practice and the analysis of science require both concepts. Neither can replace the other. Among the propositions that deserve our credence we find, for instance, the proposition that (as a matter of contingent fact about our world) any tritium atom that now exists has a certain chance of decaying within a year. ([1986], p.83)

I will explain in section 4 why quantum probabilities should not be taken to have all the features Lewis attributed to chance. But Lewis was right to believe that objective probabilities figure in science, and that these include quantum probabilities. This makes it particularly interesting that he took a single principle to capture all we know about chance, namely the Principal Principle, whose initial statement was as follows:

LetC be any initially reasonable credence function. Lett be any time. Letx be any real number in the unit interval. LetX be the proposition that the chance, at timet , ofA ’s holding equalsx . LetE be any proposition compatible withX that is admissible at timet . ThenC (A/XE )=x . ([1986], p.87)

If the only thing we know about objective probabilities is that they command an agent to adjust its credences (partial beliefs) so they equal the corresponding objective probabilities, then it is not surprising that they carry so little explanatory weight. The “thinness” of Lewis’s account of probability as it occurs within physics reinforces the application of Wright’s point - that narrowness of cosmological role is convincing evidence against a representational view of quantum probability as a natural property of (something in) the world. But this in no way undermines Lewis’s claim to be offering an account ofobjective probability.