ON THE LIMITS OF SCIENTIFIC OBJECTIVITY

People don’t agree on much—religion? politics? music? food?—and disagreement on such matters divides nations and communities right down to the level of individuals. Science is remarkable for its ability to bring people together, transcending differences of opinion. It has flourished by seeking out those things on which all conscientious inquirers can agree to be observed facts. We have a word for these, namely, ‘data’: what all conscientious inquirers can agree on because at some level these are simply given to us in observation—the mice died, plants in this generation were mostly taller than their parents, the litmus turned blue while the flame turned yellow, the mercury rose in the tube, the needle pointed to three, the detector clicked, and so on.

Philosophical challenges have been mounted against the view that observation gives us direct access to objective truth about the world, from Thomas Kuhn’s and Norwood Russell Hanson’s arguments that all observation is theory-laden to Wilfrid Sellars’s critique of the ‘myth of the given’.  But science is successful—indeed possible—only insofar as sincere inquirers are able to reach agreement on what they have observed. Such agreement is always defeasible in the light of further information as science progresses. But it is enough to make observation reports of objective physical facts the foundation on which scientific knowledge rests, as Venice rests on piles driven into the mud of the lagoon (to appropriate Karl Popper’s simile).

Recent arguments by Časlav Brukner and other physicists, however, have been taken to show that data of the type that provide the evidence for what may be our most successful physical theory are not objective physical facts at all, since they depend on, or are relative to, the observer who collects them. The theory is quantum mechanics, and the arguments are based on extensions of a thought experiment suggested long ago by the physicist Eugene Wigner and now known as the paradox of Wigner’s friend. Wigner described a hypothetical scenario where an observer remains outside an isolated laboratory, inside which his friend makes a quantum measurement. Wigner took quantum theory to imply here that while this measurement has an outcome for his friend, it has no outcome for the outside observer unless and until he himself observes the subsequent state of the laboratory. The new arguments are meant to show that two apparently reasonable assumptions lead to an antinomy: the assumption that quantum theory is universally applicable and the assumption that every quantum measurement has a single objective physical outcome.

The antinomy is related to, but distinct from, the traditional quantum measurement problem that the physicist Anthony Leggett once described this way:

[M]ost interpretations of quantum mechanics at the microscopic level do not allow definite outcomes to be realized, whereas at the level of our human consciousness it seems a matter of direct experience that such outcomes occur.

The antinomy afflicts a popular class of interpretations that seek to dissolve this measurement problem by denying that a state assigned to a system in quantum theory completely describes its momentary physical properties. For example, Schrödinger’s cat may actually be either alive or dead even when its quantum state doesn’t say which it is. Such interpretations still face a new problem when they take quantum theory to apply to a whole isolated laboratory, including any observer making a quantum measurement inside it.

I sketch an argument based on one thought experiment to set up the antinomy and show why this constitutes an epistemic, rather than a logical, paradox. Then I apply an analytic framework developed by the philosophers David Kaplan and John MacFarlane to show formally how this paradox might be resolved within philosophical semantics. This framework permits a sentence expressing the outcome of a quantum measurement to be true when assessed in one context but not when assessed in another. It can be used to resolve the paradox, but only if it is clear what constitutes a context of assessment here.

Quantum theory itself enables the necessary clarification when applied to model the kind of environmental interaction required for a measurement to have an outcome. An outcome sentence that expresses the outcome of a quantum measurement in the thought experiment may be assessed for truth or falsity only in a context where this environmental interaction can be modelled in the right way by quantum theory; outside of such a context its truth value simply cannot be assessed.

Since an outcome sentence that is evaluated as true in one context will never be assessed as false in another, this is not a case of what MacFarlane called truth relativism. And it is misleading to say that a true outcome sentence here states an observer-dependent fact, since whether it states a fact depends not on ‘the observer’ (whether conscious or not) but on the physical, environmental context of the laboratory in which the measurement is made. This context may change, especially when the laboratory is itself subject to an external measurement. So any relativity of a fact about a measurement outcome in the thought experiment is relative not just to a physical object such as the observer or their laboratory, but to that object’s contemporaneous physical state. And there are no relative facts here if that would require an outcome sentence to be true in one context but false in another.

A fact about the outcome of a quantum measurement in the thought experiment is perspectival: whether it is a fact depends on the physical context in which an outcome sentence potentially expressing it is assessed. The outcome of such a quantum measurement is a fact only from a perspective in which that sentence can be evaluated as true. But how could a physical fact be perspectival? Can such a merely perspectival fact have the kind of objectivity that scientific objectivity requires of observational data for them to provide evidence for quantum theory? In the heart of the article I answer these questions by considering why we need the notion of a true statement—a statement of fact—in general, and what kind of objectivity science requires of such facts. This requires a quick plunge into the disputed waters of the philosophy of truth.

In preparation I consider partly analogous cases in which it is difficult to assess sentences as true or as false in some physical contexts. Ludwig Wittgenstein asked how a sentence like ‘It is five o’clock on the sun’ could possibly be assessed here on earth. Such sentences are indeed assessable as true or false here on earth, but only because we have a theory of space and time capable of providing a wider context that includes both the sun and the earth at once. They would not have been assessable if space and time had been different in the way Kurt Gödel showed to be possible according to the general theory of relativity.

Quantum theory predicts probabilities for the outcomes of joint measurements occurring in more than one isolated laboratory, but only when applied in a wider context that includes them all at the same time. So it permits joint assessment of outcome sentences in the thought experiment only if there is such a wider context that then includes all those laboratories. That is why assessments of outcome sentences are limited to certain physical contexts in the thought experiment, and so why facts about these outcomes are perspectival. No verificationism is assumed: it is theory, not observation, that imposes limits that lie beyond what is observationally accessible according to that theory.

A context-independent notion of truth cannot be applied to outcome sentences in the scenario of the thought experiment. Now Huw Price, along with other philosophers, has argued that truth is not a substantial property or relation of correspondence to things in the world. But he denies that truth is merely a linguistic convenience because it also serves an important social–normative function. The norm of truth spurs reasoned debate within a community, and so helps to coordinate the thoughts and consequent behaviour of its members, to the benefit of all. If a community is segmented into physically isolated groups (as in the thought experiment), a notion of truth can serve this function in each group but not in the community as a whole. The appropriate notion of truth for such a splintered community is therefore contextual rather than universal. It is precisely the contextual notion that may be applied when assessing outcome sentences in the scenario of the thought experiment. In that scenario, facts are naturally perspectival.

To rest with a perspectival notion of fact is to abandon a philosophical ideal of objectivity that Thomas Nagel called the view from nowhere and Bernard Williams called the absolute conception of the world. I call this transcendent objectivity. It is unattainable in the thought experiment. But science requires only a more modest immanent objectivity—that a claim about the physical world (including an outcome sentence) receive the same truth value (true or false) if assessed in any actual context. Outcome sentences in the thought experiment are not even immanently objective. However, there are physical reasons why the scenario of the thought experiment will never be realized in the actual world and these ensure that we all actually share a single context of assessment. In this context, not just outcome sentences but most familiar claims about the physical world may be assessed for truth or falsity. The true ones state facts that are immanently objective, including any data we may need to support objective scientific knowledge.