This episode features Harvard physicist Jacob Barandes (the third in a series of conversations with host Curt Jaimungal) explaining his “indivisible stochastic” reformulation of quantum theory. His central claim: quantum theory and ordinary classical probability theory are separated by a single, usually unnoticed assumption—Markovianity (the idea that what happens next depends only on a system’s present state). Drop that assumption, and the formal gap between the two vanishes, yielding a realist picture of quantum theory that sidesteps Bell’s “local realism” verdict and dissolves the measurement problem. The conversation extends from why he rejects rival interpretations, through teaching general relativity, to the foundations of quantum field theory and quantum gravity, ending on legacy.
Local realism and Bell’s theorem
What “local realism” means: objects localized in space possess definite, pre-existing properties (an “ontology” or way they are) that exist independently of measurement, even if there’s subtle interplay between those properties and measuring devices.
The 2022 Nobel press release said the work proved there can be no hidden variables; Barandes argues this is not strictly correct and not what Bell argued. The narrower, accurate claim is that the experiments showed local realism is false.
On “real” vs “definite”: Barandes treats reality/existence as primitive notions impossible to define non-circularly (like consciousness, probability, free will), so they must be taken for granted.
Why he insists on realism: giving up realism entirely is incoherent and self-undermining. If measurement outcomes don’t exist in any sense, science and empirical knowledge lose their grounding. Quoting Adam Becker’s What Is Real?: without realism, “there’s nothing to be local about.” Different interpretations locate outcomes differently (many-worlds, relational/perspectival), but some realism about something is unavoidable.
If the macroscopic world is emergent, you need a substrate for it to emerge from—you can’t have emergence without something underlying it.
The hidden assumption in Bell’s theorem: Bell implicitly assumed Markovianity. Barandes points to Goldstein, Tausk, Norsen, and Zanghì (Scholarpedia article on Bell’s theorem, 2011) and Tim Maudlin’s draft “The Great Rift in Physics.” In Bell’s 1990 paper La Nouvelle Cuisine, the “screening” local beables sit in a spacetime region of finite thickness; non-Markovian laws let you “jump over” those regions while staying inside the light-cone structure—respecting relativistic causality but evading Bell’s conclusion.
Strictly, this isn’t a “loophole” but the theorem doing its job: denying one of its premises (Markovianity). What was missing was a concrete, comprehensive non-Markovian formulation of quantum theory—which Barandes says he stumbled into.
Against instrumentalism and how to judge interpretations
Barandes rejects the popular framing that quantum theory “taught us” to abandon realism and adopt pure operationalism/instrumentalism. That’s a metaphysical choice, not a lesson of physics. Instrumentalism is a fine temporary practical stance, not the end goal—there’s real work for philosophers and physicists to do.
He distinguishes harmless idea generation from wild speculation stacked on speculation, which he dislikes unless it eventually grounds itself in rigorous scrutiny.
His deal-breaking criteria for any candidate world picture:
Empirical adequacy (must reproduce the experimental predictions) — an absolute must.
No ambiguity about in-principle observable matters.
It should show, at least schematically, how the macroscopic everyday world emerges from the picture (his critique, shared with Everett, of Copenhagen: it simply assumes a classical world).
It should avoid an endless list of ad hoc, epicyclical metaphysical hypotheses.
Because quantum theory is so intricate and well-tested, these criteria are extremely restrictive—he worried for years that quantum theory might overdetermine its interpretations (no sensible realist picture exists at all). His indivisible approach is offered as an existence proof that at least one workable realist formulation exists.
Why he rejects the main rival interpretations
Copenhagen / textbook (“shut up and calculate”): empirically adequate by construction, but leads to ambiguities (the Wigner’s friend scenario, where two observers disagree about whether a measurement occurred). Unlike other theories that accept breakdown at singularities, quantum theory met fierce resistance to admitting similar limitations.
This view is still widely held: in Barandes’s 2024 incoming-PhD survey, about half chose orthodox/Copenhagen. Anton Zeilinger explicitly embraced Copenhagen in his 2019 Harvard Lee Historical Lecture—treating wavefunction collapse as mere information update, with no measurement problem. Barandes respectfully disagrees.
Bohmian / de Broglie–Bohm pilot-wave theory: very hard to generalize beyond fixed numbers of non-relativistic particles to relativistic, fermionic, interacting field theories like the Standard Model. David Wallace’s “why is the sky blue?” (Rayleigh scattering) argument shows it struggles with relativistic problems, signaling empirical inadequacy—though this could conceivably be fixed.
Everettian / many-worlds: his core objection is that you cannot get genuine probability out of a deterministic branching picture. Decision-theoretic derivations (Deutsch 1999, Wallace 2012) and branch-counting/coarse-graining approaches (Simon Saunders) are, on close inspection, either circular or laden with speculative assumptions.
You can’t simply impose a probability measure by axiom, because in modern Everettian accounts branches are emergent and approximate, not fundamental—just as you can’t axiomatically declare emergent chairs must all be yellow.
Many-worlds also requires too many additional, hard-to-justify assumptions (e.g., even the intuition that a branch with a zero coefficient “doesn’t exist” is questionable, since via Strocchi/Heslot’s reframing of the wavefunction as classical harmonic oscillators, a zero coefficient is like a non-oscillating oscillator that still exists). Each added assumption lowers his credence.
The origin and meaning of indivisible stochastic processes
The backstory: As a college student at Fermilab one summer, Barandes self-taught linear algebra; disappointed by his exam grade, he did an extra-credit project on stochastic processes—his first exposure, before learning quantum mechanics. (He frames this as Vonnegut’s “good news / bad news,” echoing how Heisenberg’s failed exam question on a topic related to uncertainty may have helped inspire the uncertainty principle.)
Roughly eight–nine years later, doing Markov-chain Monte Carlo simulations of black holes in his PhD work, he revisited his old photocopied notes with the eyes of a trained physicist and was struck by the formal parallels between stochastic processes and quantum theory: both use probabilities encoded in vectors (probability vectors vs. state vectors), time evolution by square matrices (stochastic matrices vs. unitary operators), and observables as functions/operators (random variables vs. self-adjoint operators).
The discovery moment: While preparing to teach quantum mechanics to undergraduates who knew probability but not linear algebra or complex numbers, he tried to massage both formalisms to look similar, hoping to reduce the gap students must “jump” to something simple. Instead, the gap disappeared—he had one formalism for both. It took him a while to realize he had implicitly dropped the Markov assumption, letting the process depend on past details not mediated by the present.
Checking the literature, he found almost nothing on modeling quantum theory with non-Markovian processes. Earlier stochastic reconstructions (Fritz Bopp in the 1940s—mentioned in Everett’s writings; Imre Fényes in the 1950s; Edward Nelson 1960s–80s) had all assumed Markovianity.
What “indivisible” specifically means: the laws let you predict the system probabilistically from a starting conditioning time to chosen later times, but not from an arbitrary intermediate time forward—the laws fail to “divide” in time. This is a particular, unstructured form of non-Markovianity (the term comes from quantum-information literature ~2006, applied to classical-looking stochastic processes by Milz and Modi, ~2020–2021—almost simultaneous with Barandes’s independent arrival).
Indivisible process vs. a non-Markovian “realizer” (term suggested by philosopher Alex Meehan, UW-Madison): a single fully-specified non-Markovian process assigning probabilities to every trajectory/detail is one “realizer.” An indivisible process specifies only the rudimentary laws and leaves the empirically insignificant behind-the-scenes details unfixed—so it represents an equivalence class of many possible realizers. You can’t extract a unique set of behind-the-scenes frequencies; the laws simply don’t determine them.
Practical payoff: Fully non-Markovian models normally seem intractable—you’d need infinite lawful information. Indivisible processes give a way to model genuinely non-Markovian, memory-dependent systems with only a limited set of laws, without truncating to a Markov approximation. Since the framework is equivalent to quantum theory, it clearly makes rich predictions.
Barandes hopes the tool finds use beyond physics—finance, biostatistics, neuroscience, machine learning—anywhere memory effects matter. He stresses it only entered the literature ~2020–2021 and remains an almost blank page.
In the indivisible picture: probability, kinematics, dynamics
He separates three things often conflated as “quantum”: probability theory, kinematics (how states/configurations are represented), and dynamics (rules for change).
In his approach, probability is ordinary classical probability (welcomed by statisticians who feared quantum theory needs a non-classical probability), and kinematics is essentially classical—particles have definite arrangements, fields have definite intensity patterns. Only the dynamics are new: non-Markovian, indivisible, probabilistic laws rather than Markovian deterministic differential equations.
He criticizes Sidney Coleman’s famous lecture “Quantum Mechanics in Your Face” (1994) for conflating these three and for its dismissive, condescending tone toward philosophers—an attitude Barandes thinks discourages the cross-disciplinary work physics needs. Coleman’s probability arguments (relying on infinite repetitions of experiments) and GHZ-based claims are no longer considered definitive.
Teaching general relativity and black holes
When teaching graduate GR, Barandes devotes half a class to open Q&A on black holes, no lecturing—letting students dwell on what first drew many of them to physics.
A common confusion: given extreme time dilation near a black hole, how does anything ever fall in? Spacetime diagrams in the usual coordinates make infalling trajectories appear never to reach the horizon.
The resolution is about coordinate choice: GR has no preferred coordinate system. The coordinates good for solving the Einstein field equation are often badly behaved at the horizon. Other equally valid coordinates that agree with a faraway observer’s wristwatch—such as Gullstrand–Painlevé coordinates (also called “global rain” coordinates, based on raindrop-like free-fall trajectories)—reveal what actually happens to infalling probes. He finds the image of raindrops quietly falling onto a black hole poetically beautiful.
Coordinate-independence vs. general covariance: “General” reflects Einstein’s aim to handle arbitrary (non-rectilinear, accelerated, free-falling) coordinate systems, not just Minkowski ones. Covariant (weaker than invariant) means the laws and mathematical objects retain their structural integrity and consistent meaning under coordinate changes—things don’t look identical in every frame, but the theory’s form is preserved.
Lessons from Nima Arkani-Hamed
Barandes did the first half of his PhD with Nima Arkani-Hamed (particle phenomenology) before switching to quantum gravity under Frederik Denef when Nima left for the Institute for Advanced Study.
From Nima he absorbed effective field theory: treating quantum field theories not as exact fundamental descriptions but as progressively better approximations valid in different regimes—“our theories serve us rather than us serving our theories.” He also gained an appreciation of the particle–field relationship via representation theory.
A pivotal moment came in Nima’s course “Quantum Mechanics in Spacetime” (which famously assigned no homework and gave everyone A’s), motivated by the cosmological challenge: how to make sense of quantum theory for the whole universe when there are no external observers (a problem Everett raised in 1957).
Nima argued (inspired partly by Coleman’s lecture) that you could derive unitary evolution and the Born rule from just the eigenvalue–eigenstate link plus decoherence and the right measuring devices. Barandes spent a long time trying to reconstruct this and concluded it can’t be done—you can’t get the probabilistic postulates out for free. (Nima later didn’t recall holding those exact views.) This frustration helped push Barandes toward quantum foundations and toward realizing he wanted to be a philosopher of physics.
Philosophy of physics as a method
Barandes draws a contrast between physical philosophy (what our best theories say about traditional metaphysical questions) and philosophical physics (using analytic/philosophical tools to make progress on physics problems)—distinct from theoretical, mathematical, experimental, applied, or computational physics.
The method is rigorous scrutiny: clear definitions, clearly stated premises, careful argument, and hunting for implicit assumptions and unnoticed connections—“looking under rocks.”
His paradigm example: Einstein interrogated the seemingly mundane “scenery” of inertial reference frames—not building a new atomic or electromagnetic model—and relativity fell out. Barandes’s own connection between stochastic processes and quantum theory is the same kind of work.
Science requires some unjustifiable metaphysical starting points—e.g., that experimental devices exist and that induction works (Hume’s circularity problem; he cites John Norton’s The Material Theory of Induction).
He stresses (echoing his earlier conversation with Emily Adlam) the track record of philosophy-informed foundational work yielding concrete physics: EPR/entanglement, decoherence, quantum advantage, Bell’s theorem, the no-cloning and no-signaling theorems, and downstream quantum teleportation, cryptography, and information.
Foundations of quantum field theory and quantum gravity
Why little “interpretation of string theory” work exists: string theory is newer, still speculative and unconfirmed, and mathematically very intricate—a large time investment that deters many philosophers of physics. Philosophers prefer interpreting empirically established theories so their work endures. (He recommends Nick Huggett for philosophy of string theory.)
What string theory taught him about QFT (not string-contingent): the relationship between particles and fields. A particle’s intrinsic properties—mass, charge, spin—reveal the structure of its corresponding field. For example, the photon’s zero mass and spin-1 character are tied to the gauge invariance of the electromagnetic field. (Weinberg’s QFT books show this too, but Barandes found it more intuitive through string theory.)
Open problems in QFT foundations:
The measurement problem persists but is more hidden: practical QFT computes scattering amplitudes (the S-matrix) from far-past to far-future particle states using the Born rule, usually with a single measurement and no explicit collapse. Students are surprised the Schrödinger equation largely disappears.
Reconciling rigorous algebraic QFT (C*-algebras) with the non-rigorous but empirically successful Lagrangian/effective QFT of the Standard Model—we have no rigorous algebraic version of the Standard Model.
The ontology question: are fields the real entities, or particles, or something else? Both particle-based and field-based ontologies have been argued untenable; some leading theorists admit we don’t fully understand what QFT is.
QFT on curved spacetime, the Unruh effect (accelerating observers seeing particles), and unitary inequivalence of representations. He recommends Clifton and Halvorson’s review “Are Rindler Quanta Real?” as both an introduction to the C*-algebraic formulation and a treatment of this problem. (He suggests Michael Miller, Doreen Fraser, David Baker, Noel Swanson, Hans Halvorson as interview subjects.)
Particles popping in and out / fluctuating spacetime:
Non-relativistic quantum theory conserves particle number; in the indivisible approach these are like probabilistic snapshots of fixed particle arrangements.
Relativistically, particle number is generally not conserved (subject to conservation laws like charge—if charged particles appear, antiparticles must too). Modeled either with Fock spaces (second quantization), where configurations include any number of particles, or with fields, where there’s a persistent substrate that gets excited in changing patterns (emergent particles, no literal winking in/out). He mentions third/higher quantization as a speculative direction.
Quantum gravity is openly speculative—no fully realized theory exists. Functional-integral (path-integral/partition-function) techniques sometimes reach results the Hilbert-space approach can’t, hinting Hilbert spaces may not be fundamental to gravity—especially since Hilbert spaces suit an external time parameter, which GR lacks (he notes the Page–Wootters formalism as a workaround).
Summing complex amplitudes over whole spacetimes has murky metaphysical status. In his indivisible view there are no literal superpositions—superposition math just encodes indivisibility—so quantum gravity might involve one spacetime (or one substrate from which spacetime emerges), with amplitude-sums no more metaphysically literal than a path integral for ordinary particles. He stresses he doesn’t yet have a firm enough grip to offer a concrete picture.
Legacy
Asked what a future history book should say about him, Barandes leads with character rather than discoveries: that he was kind, genuinely good to the people he cared about and to everyone, that he tried to make the world better, and that he created spaces where people felt safe, nurtured, and loved.
On scholarship, he hopes his projects “pan out”—he’s not claiming a theory of everything, but hopes to address important foundational problems, or at least inspire others toward new directions. He also hopes any account would mention his family, whom he loves.
