You are probably correct when it comes to psi-ontic viewpoints… but that point alone is a great reason to join the psi-epistemic camp! Quantum foundations is hard enough as it is…. 😀
Physics professor at San Jose State. Quantum Foundations. Big fan of space and time, and also many things therein.
617 followers 971 following 179 posts
view profile on Bluesky
You are probably correct when it comes to psi-ontic viewpoints… but that point alone is a great reason to join the psi-epistemic camp! Quantum foundations is hard enough as it is…. 😀
Be-able, rhyming with agreeable. As opposed to observe-able.
I guess he wouldn’t be the first person confused about “hidden variables”. Too many people still think that they are ruled out, despite Bell’s writings and an explicit appearance in Bohmian mechanics. (They’re not ruled out, they just have to violate one of the assumptions of Bell’s Theorem.)
I’m not sure why Whitaker would try to draw some distinction between Einstein’s hoped-for “real parameters” (existing between preparation and measurement) and “hidden variables”. They're the same thing! Bell thought of them in the same way, at least, and called them “local beables”.
* judgement -> reality
Why not make good/evil/neutral refer to judgement of the wavefunction? So, move QBism down to Chaotic Evil, put objective collapse at Chaotic Good, and consistent histories at Chaotic Neutral. And swap Bohm w/ Everett. Neutral good's stance is the QM math needs no interpretation ("idk, seems fine").
Thank you! I haven't heard anything from the original authors yet, but I am curious as to how they might respond. We'll see... (Also, I forgot to tag my co-author @huwprice.bsky.social on the earlier post.)
Hopefully this isn't coming across as an "angry letter", but our analysis here says that it is just postselection. :-) arxiv.org/abs/2508.13431
Our analysis of this paper is now on the arXiv. We didn't weigh in on whether there is hidden entanglement, but we did show it's not needed to explain the results. (With @huwprice.bsky.social ). arxiv.org/abs/2508.13431
Thanks for the kind words! The analysis turned out to be more subtle than I expected, but the classical analog did most of the work for us.
Well, it's a tradeoff. For some safely hidden retrocausality (inside the backwards light cone, so Lorentz covariant), we could actually *restore* both realism and the best parts of locality. No paradoxes or time travel, since it's all at the hidden variable level. What's not to like? 😉
Indeed! In fact, the one thing that everyone in quantum foundations can agree on is that (at least) one of our intuitions must be deeply wrong somehow. Locality, Reality, and Causality are the big three. One of those is going to need a massive revision... (My $$ is on causality, FWIW)
Still embarassing that we're below superdeterminism, tho.
The only way to know for sure if one could “never know the answer” would be to have a better model in the first place. And if an attempt to answer such questions led to this better model, then the questions would still have been “good”, because they would have led to progress in physics! 😉
In the branches of quantum foundations that take such questions seriously (setting aside the antirealist `no reality before measurement’ stance) it’s not “philosophizing”, it’s *modeling*. We build physics models with some hidden quantities and test them to see if the unhidden aspects match QM.
Let me try to explain why it's important to study the foundations of quantum mechanics. (1/n) (Have to do some breathing exercises, b/c to me it's blindingly obvious that "understanding the most important theory in physics" is something physicists should care about, but apparently opinions differ.)
The only difference I see is that atoms were eventually measureable, but what happens between measurements is by definition never directly measureable. Your question here is part of the "ontology problem" (T. Norsen). Just because we don't know the answer doesn't mean it's not a great question!
But does it really depend on the answer? I'd argue that any coherent answer to these questions, if correct, would point in a very useful direction. At minimum it would better inform how to unite QM/QFT with GR. In that context, whether the wavefunction is ontic/epistemic *is* materially important.
Why is it “weird philosophy” to ask why a mathematical tool works to make successful physical predictions? Isn’t answering that question obviously important for making progress in physics? Is certainly has been in the past, in other contexts (thermodynamics, for example).
E&M will have to wait for another day -- but next time you read something about 'advanced' vs 'retarded' solutions, consider how much clearer the analysis would be if the author knew anything about causal models and highlighted their choices of exogenous/input variables.
Here's my only published take on the fine-tuning, from a while back: arxiv.org/abs/1510.03706 . In a nutshell, fine-tunings aren't the end of the world; they call for explanation, and a number of explanations seem to be available.
As far as defining a speed of (hidden) causal influences, certain models might be amenable to such things: Namely, strictly-forward-causal models, where you’re allowed to hold certain variables fixed while calculating what would happen if you vary others. But not for NFID-violating models.
Lots of pitfalls here; better to use screening-off conditions. It’s important to distinguish measureable&controllable signals (the usual sense of the word) from more general causal influences. The former have a roughly-defined “speed”, but precision is tricky (say, as in tunneling-time expts).
We concluded that Bell-inequality violations are still perfectly compatible with maintaining the "CA=Continuous Action" screening condition in Figure (a), so long as you instead give up the "forward causality" assumption (which we called "NFID=No Future Input Dependence".)
I'm glad, since that was the point of the paper! :-) Notice that we were able to essentially break "Local Causality" into 'locality' and 'causality' assumptions, for more careful separate analysis. The usual assumptions refer to temporal-order in multiple places, making it hard to separate them out.
Very good points -- but there's still a way to do it in GR, dropping all references to lightcones. The below figure is from my Rev Mod Phys piece with Nathan Argaman. Consider the screening region in (a) rather than (c). One can define no-action-at-a-distance to mean that S screens off 1 from 2.
Here’s the original classic on the topic; arxiv.org/abs/1208.4119 . But what really needs doing, still, is classical E&M. Please reach out if/when you start in on that side of things!
Why The Quantum State Only Exists In Our Mind A metaphysical shift that’s happening in the foundations of physics: the wave function is no longer regarded as something real, but just as a description of what we know about the world. youtu.be/SgQdDD3SY84
Very nice! And thanks for leaving a little wiggle room open for that “minority” retrocausal viewpoint. If you’re ever thinking about an article on that side of things, please reach out.... Emily Adlam and @huwprice.bsky.social are also great resources on the topic.
Satisfactory for whom? Not for most people who work in quantum foundations…
Hmm. I’ve seen a few of these and I thought this one was much better than the others. Sure, there are thousands of subtle takes on each of these questions, but my views were at least close to at least one of the options in each case. (With room to elaborate.) Which questions bothered you the most?
Wow -- that was a really-well designed survey! The best one of these I've seen by far.
Uhoh – looks like a few people on Reddit have started playing around with the central ideas in my latest arXiv preprint. Must be time to polish it up and actually submit the damn thing. (But still, always nice to read "it's kind of blown me away"...) www.reddit.com/r/QuantumPhy...
Are you doing a “T1-style” story (all one timeline/universe) or a “T2-style” (many timelines)? Here’s a fun talk I gave once about the differences from a physics perspective. www.youtube.com/watch?v=4748...
I did a little gig doing some “time travel consulting” for the Peabody and Sherman movie, and always felt that was an untapped market I should take better advantage of. If only Nolan had looked me up when he was writing Tenet… :-)
Heading to Alaska soon, where apparently even the landslides are beautiful.
Interesting point! I'd be curious as to what you think of the linked essay, if you read it with that idea in mind.
Maybe there's a reason for that. Maybe, despite GR/block understanding, in the 1920s everyone but Einstein shoehorned QM into a process view, as it was more intuitive. So looking at QM, you’ll see “evidence” of a process view, but that “evidence” was put in by hand for anthropocentric reasons.
My quick take on Bell violations was just posted at the end of this thread. The key is considering "future input dependent" models as described in the Rev Mod Phys paper linked in #10.
Yes, you're right that it's probably very misleading. My main point was aimed at people who still tend to think of measurements as a passive process, just taking in information. In fact they are drastic and massive interventions, using instruments much larger than what is being measured.
As strange as this sounds, if you run through the consequences, you end up with a bunch of results which look precisely like quantum physics. 1) Inherent uncertainty (actual hidden variables), 2) Bell-inequality violations (the hidden variables depend on future settings), 3) Contextuality, etc.
Partially-controllable final boundary constraints are very strange things, however. Imagine that a closely-spaced grid, where particles were always constrained to hit one of the grid points. By moving the grid, I would effectively have some control over the particles’ precise initial conditions.
That’s not to imply that nothing important happens at an interaction. Even in classical physics, when small things interact with big things, the big things act like boundary constraints on the small things. So I think of measurements as final boundary constraints, with control of the measured basis.
And when we learn something new, we update our knowledge. This looks like a discontinuous “collapse”, but it’s not anything real: just a (Bayesian) update, like when a curtain is revealed and we see the prize isn’t there. No probability wave is jumping from curtain to curtain, of course.
Measurements are just interactions, of course. There can’t be any fundamental difference. The only real difference is a subjective one: we learn something from a measurement. A practical difference is that we can carefully control the measurement basis, as opposed to some random interaction.
Thanks to @mercurynews.com for covering the $55.9M in NIH cuts to Bay Area research funding. I appreciate that they didn't just focus on the elite institutions (a colleague and I are quoted) & want to share some key takeaways 🧪 1/ www.mercurynews.com/2025/05/05/b...
QM is a "peculiar mixture describing in part realities of Nature, in part incomplete human information about Nature—all scrambled up by Heisenberg and Bohr into an omelette that nobody has seen how to unscramble". - E.T. Jaynes. I'm for unscrambling. Better: not scrambling in the first place!
You inspired me to write a bit of a rant about working in quantum foundations... :-)
11) Finally, it is undisputable that something counter-intuitive is happening, at least at some level. So always remember that some of your intuitions must be wrong, no matter how much you instinctively trust them. (Spoiler alert: it’s our intuitions about time and causality.) *end*
10) The assumptions behind Bell’s Theorem are crucially important, especially the details. Once you’ve ruled out the impossible, whatever is left, no matter how improbable, is telling us where to look. arxiv.org/abs/1906.04313
9) The best way to *not* ignore GR is to build a GR-compatible ontology. So: no real Hilbert spaces or Fock spaces; everything ontological should be a function that could live on a spacetime manifold (maybe with a few extra dimensions, if required, but no more).
8b) And PLEASE, PLEASE don't stick with this archaic 19th century viewpoint where time and space are fundamentally distinct. We should have moved on from that by now.
8) Don’t ignore relativity, don’t ignore GR. Sure, a lot of classical models were first developed in the non-relativistic limit and then had relativity added in. But there’s no guarantee that this path might work in general, and it’s pretty doubtful that it could ever work here.
7) To a quantum system, an actual measurement is like a sledgehammer probing a speck of dust. Don’t assume that what actually happens in the absence of a measurement is related to what you would have found during a (counterfactual) strong measurement.
6) At first, don’t try to recover the formalism of QM; try to recover the quantum *phenomena*. Keep in mind that what experiments measure is almost certainly not a complete description of what is actually happening. Take the uncertainty principle literally.
5) Start with the hardest-to-explain basics, not single-particle models. Start with 2- and 3-qubit entanglement experiments, the Hong-Ou-Mandel effect, the photoelectric effect, and other highly-constraining basic phenomena.
4) Looking to the QM formalism to find that deeper structure is probably useless. You can’t look to the form of the heat equation to see that an atomic picture + stat mech can explain it. You need to start with the deeper level and work back up from there.
3) QM is a self-contained, tightly-constructed puzzle-box that doesn’t quite all fit together. So don’t just pull one piece out of the QM-puzzle-box; rip it all apart from the start. Every piece is suspect. We need a completely different structure, built from scratch.
2) There is never any reason to think that we’ve finally hit conceptual fundamentality, in any theory. QM wavefunctions and QFT fields are no different. There’s almost certainly some deeper level of understanding and/or description that is needed.
Unsolicited advice for making progress in quantum foundations: 1) If there were a gentle tweak or simple idea that would make clear sense of quantum theory, beyond the for-all-practical-purposes-level, it would certainly have been found by now.
Sure, that's a valid quantum measurement, if not a complete one. But any state of knowledge would "collapse" here. Throw a rock, blindfolded, towards a pane of glass; no shattering noise also gains you information. This doesn't mean any physical "collapse" has happened. (We still need an ontology!)
Many papers get bogged down analyzing non-local weak values, but I think we really only need the *local* (spacetime-localized) weak values. Say, the weak values of each of the 3 Pauli operators for a spin-1/2 particle (even if it’s entangled!) That's what I propose here: arxiv.org/abs/2412.05456 .
Note that I'm talking about a weak *value* ontology, not weak measurement. You don’t have to actually do the weak measurement to calculate the weak value, for any eventual strong measurement outcome. The original weak value paper is here: journals.aps.org/prl/abstract...
Of course, if you instead measure the photon on a distant screen, building up an interference pattern over many shots, now the single-shot weak values are localized in both slits, as an extended field. That seems right to me, too. That's one reason I'm drawn to a weak-value-based ontology.
I'm actually looking at weak values inside quantum circuits these days, for a simple+modular framework. But the double slit is instructive, too. If you measure which slit the photon goes through, the weak values of the field are indeed entirely localized that one slit. That seems right to me.
I don’t necessarily think that weak values are a complete lowest-level description, but they might be the best ontology that QM has to offer, in its current form. Strong measurements require a substantial interaction which are going to change those weak values, even as they inform our knowledge.
But weak measurements introduce an interesting possibility on this front, one that Yakir Aharonov has long advocated, and one that I’m coming around to myself. What if the “weak values” (the averaged results of an infinitely-weak measurement) are themselves the objective ontology?
Before making sense of measurements (and the measurement problem), one first needs to take a stand on Norsen's "ontology problem": what is the stuff that QM is describing? What is our subjective state of knowledge, what is objectively real, and how is the real stuff changed by a measurement?
Please tell us about the demo!
If you're interested, here's an "unlisted" youtube video of me giving a rare talk on superdeterministic models, and how they relate to retrocausal models. The sound gets better after the first few minutes. www.youtube.com/watch?v=q356...
A very good, wide-ranging interview with Emily Adlam -- most of it on quantum foundations, along with a strong case for retrocausal/all-at-once approaches. youtu.be/6I2OhmVWLMs?...
It’s worth looking back at Heisenberg’s original paper before he fully switched over to Bohr’s language about unknowns not even properly existing. That paper was all about the impossibility of learning everything by using tools/interactions that we didn’t know everything about in the first place.
And it’s not that they’re unlearnable -- it’s that they are unlearnable *at the time*, by an agent who doesn’t know the future. Those two restrictions are clearly related, somehow. I don’t know the future, and I don’t know *some* properties of the present. Eventually I can piece together the past.
I’m not sure it’s coherent to talk about learning or knowing things without talking about agents. And there are certainly restrictions we put on a generic agent that we don’t find problematic -- such as assuming that agent can’t know the future until it happens. This is not that different.
I’m biased, but realist psi-epistemic models are worth looking at, starting with Rob Spekkens’ “Toy Theory”. The possible ways to then allow Bell-violations are categorized in this 2020 Rev. Mod. Phys. piece, along with an analysis of some “all-at-once” models. arxiv.org/abs/1906.043...
Yes! I'll post the link when it has been uploaded, but may not be for a while.
Alas, we only have only toy models that work for special cases, like maximally-entangled states and GHZ states. But we're finally making good progress towards a general framework, including partially-entangled states, here: arxiv.org/abs/2412.05456 . I'm giving a talk on this Friday at Chapman.
Well, take the Bohr atom, in terms of deBroglie waves. That's really a sort of "cyclic constraint", right? Having the wave come back in the same phase after one cycle quantizes the angular momentum (and energy). But it's only retrocausal if you can choose the basis in which it ends up quantized.
I wouldn’t want loops in the causal model diagram, but I’m fine with some variables being jointly caused by cyclic constraints. In fact, that could be interpreted essentially like the “old quantum theory” (Bohr-Sommerfeld), which itself is a very interesting source of retrocausal model ideas.
Yes, better. This one at least made it very clear that you're not *directly* measuring the negative duration, but using another measurement that implies a negative duration.
Aren’t hidden variables enough for that? :-) I don’t see the need to mess around with spacetime itself just to explain entanglement. If you think entanglement is everywhere, and not just epistemic, just imagine what a jumble that would make of the ordinary spacetime we hold to be well-tested.
Here’s a relevant slide from a talk I’m giving at a Foundations conference at Chapman next week.
Maybe. But accepting that gives up on relativistically covariant models, at the level of the hidden variables. Bohmian models aren’t; there are evident “planes of simultaneity” in the hidden phenomena. Retrocausal models, solved all-at-once, can let you can foliate time however you want.
No unusual dynamical *equations* are needed at all, just a controllable constraint on some future variables. Say you could adjust a grid on the ground, and a thrown projectile was somehow constrained to land on one of the grid vertexes. Ordinary trajectory dynamics would then be retrocausal.
Latest version of my work on entanglement with @kenwharton.bsky.social. We propose that entanglement is a kind of selection artefact. This was a talk at @griffith.edu.au, and the slides are accessible here: bit.ly/4h9dMeo #philsky #PhilSci #quantum #entanglement
I find it interesting, too, but it doesn't seem like there's any overlap with retrocausal approaches to quantum foundations. Still, Jacob and I are having a zoom conversation in a few days, so maybe I'll learn otherwise!
I’m convinced that the only way to *not* get confused on this topic is to take Huw Price’s “view from nowhen”. The block-universe view is crucial. Then add Judea Pearl’s causation as counterfactual analysis of possible universes, as opposed to any “flow of time”. www.amazon.com/dp/0195117980
Here’s my take: What you can know and what you can control are linked. A measureable variable can be set to some value. The unmeasured/unconstrained parts of what you’re interacting with -- the bits you don’t know -- are the only free variables on which retrocausation could then possibly act.
Also, the same argument can be made against *any* hidden variable model. If you could see the Bohmian trajectories, you could signal faster than light. Or if you could witness the GRW ontology. You need an uncertainty principle in those cases, too. This is no different.
One can reverse the logic here, too. Premise: the laws of the universe must be self-consistent. Which laws may then possibly exist? The idea is that self-consistency is a source of explanation, not something that needs to be explained. See Emily Adlam: arxiv.org/abs/2110.15898
Yes, there must be some sort of fundamental limit on what one can know – say, an “uncertainty principle” -- to prevent signaling to the past in a retrocausal model. If only we had such a thing! ;-) True, it’s not interpreted in this way now, but go read Heisenberg’s original paper.
The second law of thermodynamics in the context of a time-symmetric block universe. So, basically, timeless love. :-)
And yet another take on the issue can be found in my 25-year old novel, in the context of a new religion, "Symmology", I think I called it. But that's a much longer read... :-) www.amazon.com/Divine-Inter...
Yes, very interesting question, and there's not much written on it. Here's the best paper on the topic. You can see that I took a different take on it in "Aloha", but something like Schulman's result is much more plausible. arxiv.org/abs/cond-mat...
If anyone is interested, it looks like someone has posted a version of it here. A pretty short read, although you may be tempted to read it again backwards. ;-) www.zayix.com/GameLibrary/...
A little fun news for me... Apparently the last science fiction story I ever wrote, 20 years ago, is now being discussed in at least one Philosophy course! The story is called "Aloha", originally published in the magazine Analog.
Wait -- what? You're discussing "Aloha" in a philosophy course?!
Tonight I got to see the first-ever spacecraft image of Mars, from Mariner 4's flyby. Apparently, before the raw image was processed, people at JPL printed out paper strips of numerical pixel values and -- coloring by hand! -- used pastels to distinguish different brightness levels.
So: retrocausation is still causation in the interventionist sense, just in an unusual temporal order. And it's definitely directed. If you are thinking in terms of symmetrical relations, you're thinking about correlation, not causation.
But that evidence is lacking in the quantum world, down at the hidden variable level. Here, there’s no empirical restriction on a future setting being correlated with a past hidden variable. That’s the path to retrocausal accounts of quantum theory. It’s still "causation", by Pearl’s lights.