What do quantum physicists mean when they say "observer"?

December 29, 2007 6:01 PM

In quantum physics, what's an observer?

Okay, so the term "observer" comes up a lot in explanations of quantum physics. When I hear about Schrödinger's cat or the double-slit experiment, I hear "observer" and I think of some guy in a lab coat and bow tie leaning in and POOF! There's a cat! And of course that gets me thinking things like "What if you videotaped the live/dead cat and watched it later?" and "Hey, why can't the cat observe itself?"

When online discussions head in that direction, generally someone steps in and says "No, it doesn't work that way, when we say observer we don't mean necessarily a person playing peek-a-boo." But when they try to explain what they DO mean by observer, well, I'm sad to say I haven't quite grasped it.

So can anyone out there explain it the way Isaac Asimov would have? (Or did, I suppose.) Or can you point me to someone who can?

posted by L. Fitzgerald Sjoberg to science & nature (21 comments total)

I am not a physicist. But I'd imagine that, since observation of something requires interaction with it, that "observation" really just means the interaction between your observational instrument (photons or whatever) and the subject of observation. Probably wildly incorrect, but maybe someone else will come along and elucidate.
posted by Chris4d at 6:16 PM on December 29, 2007

You are over complicating it. An observer is anyone in the act of observing. The assumption in any quantum physics situation is that there are only the observer and everything else (which is being observed). The cat experiment is misleading because it is a gross oversimplification of a complex idea. It is so simple that, if you over think it, it does not work.
posted by slavlin at 6:16 PM on December 29, 2007

An observer is someone who measures something in the system. You don't know what state the system is in with certainty until you measure it.
posted by jeffamaphone at 6:18 PM on December 29, 2007

An observer would be any entity capable of causing a resolution of the quantum state of the situation.Even a mirror could qualify.
posted by Dipsomaniac at 6:18 PM on December 29, 2007

As far as I can tell, there is no good consensus on what constitutes an observer. There's endless handwaving, though.

My own take on it is that QM is about what is known, as opposed to being about what is, which means that an observer is any entity capable of knowing things. Of course, that simply pushes the handwaving out by one level - what, exactly, does it mean to know something?

I will be watching this thread keenly for signs of people who can actually explain this in a lucid and reasonable fashion.
posted by flabdablet at 6:20 PM on December 29, 2007

It's a tricky question, one of those turtles all the way down infinite regression things. If you as an observer collapse a waveform to make something 'real', who collapsed *your* waveform? Who observes the observers, etc etc?

That 'observer' stuff was all just part of one interpretation of what is going on in the quantum realm. There are other interpretations that don't require an observer to collapse a waveform, such as John Cramer's transactional interpretation of quantum mechanics. The problem is that most physics books still teach the Copenhagen Interpretation as gospel, when it was only ever a best guess at what was going on.

If you want to get away from observers and know more about the transactional interpretation, I'd go right to the source.
posted by foobario at 6:26 PM on December 29, 2007

Here's the first hit for google search of quantum+observer. According to this, an observer can be a measurement device.
posted by DarkForest at 6:30 PM on December 29, 2007

An observer is anything that can cause a wave function to collapse. It emphatically does not imply intelligence.
posted by Steven C. Den Beste at 6:46 PM on December 29, 2007 [1 favorite]

You're actually touching on one of the main unresolved problems of quantum physics, namely what is the role of the "observer" in wave-function collapse. It's certainly true that the idea of an "observer" in the standard establishment view of QM (the "Copenhagen interpretation") is not well-defined. Rather, the Copenhagen interpretation is a set of rules that can be thought of as empirically describing what happens to quantum systems when we humans observe them. The exact role of the "observer" in the act of observation has been the subject of a lot of quasi-philosophical speculation over the past 90 years, some of them sillier than others, and some of them put forth by physicists who really should have known better. For an example of this, see "Wigner's friend"; Wigner was a brilliant mathematical physicist and helped to rigorize the field of quantum mechanics in the 50s, but this particular bit of speculation was, in retrospect, kind of silly.

The most promising idea to resolve the question of what, exactly, an "observer" is is the idea of decoherence. Under this idea, a quantum mechanical system starts behaving classically (with definite states, no superposition phenomena, and so forth) when it interacts with another system with sufficiently more degrees of freedom than it has. When that happens, the "quantum information" that makes the smaller system behave in a quantum fashion "leaks out" into the larger system, and the larger system ends up always seeing the smaller system as having a definite state. There's a lot of complicating mathematics that's involved in trying to quantify exactly when the transition takes place, what size the systems have to be relative to each other, how strongly they can interact, and so forth, but it's really a quantification of an idea that's been around for a while, and it makes intuitive sense to me and a to a fair number of other physicists.
posted by Johnny Assay at 6:50 PM on December 29, 2007 [2 favorites]

The Wikipedia article on Schroedinger's cat is good, and shows that even quantum physicists don't agree on this sort of thing.

The idea is that at the regular scale, when we observe things, it seems to be a passive operation. The information reaches our eyes or ears without us doing anything. At the quantum level, you need to take a more active role. The classic example is resolving the vector or position of an electron. How do you do that? If you rig up a quantum radar detector to get its vector, you're bouncing something (photons, whatever) off the electron, and the interaction between the photon and the electron is significant enough to knock the electron off course. If you want to know its position, you can build an extremely tiny brick wall to stop it, but then you've really changed it's vector. So the electron's position and vector are said to be described by a probability cloud or wave function that represents the unknowabilityness. The problem is essentially that at the quantum level, everything is so small that nothing is insignificant by comparison. In order to get a reading, you must interact, and that interaction unavoidably alters the thing you're observing.
posted by adamrice at 6:52 PM on December 29, 2007 [1 favorite]

Feynman's formulation basically removes the observer paradox and I heartily recommend it.

If you go for the Copenhagen Interpretation, which to my mind is ugly, then the observer is essentially any measurement taken of a closed quantum system which transmits information about the encapsulated system to the 'outside'. The problem with this is that you can always reformulate the problem to encapsulate the observer within a new, larger, quantum system.

For example, instead of considering the box containing Schrodinger's cat, and a scientist lifting the lid, you consider a room containing both of them.

Once you do this, you have to start making up crap about 'conscious observers' and very quickly get into the realm of metaphysics.
posted by unSane at 6:59 PM on December 29, 2007

I can't give a very compelling answer, but maybe I can sketch out some vague sense of the issue. Perhaps it will help to look at a measurement and not an observer. They are essentially the same concept, but "observer" sounds too anthropocentric for me. Now tied to the idea of measurement is the process of "collapsing the wavefunction". Whenever you measure an observable (like energy, momentum, spin, etc), you collapse the wavefunction of the system you're measuring. So before the measurement you could have a mixed state of say spin up and spin down, when you measure the spin you get a single result back. If you measure the spin again immediately after the first measurement, you get the same result back. The act of making these measurements changes the system, which I suspect you have already read about.

Now this is where things get hand-wavey. In some sense, a measurement is when we can stop "playing quantum mechanics" and think about the world (at least somewhat) classically.

Say I want some spin-up particles. I could just pass them through a magnetic field, which will interact and "measure" their spin, and deflect them differently if they're spin up or down. This works experimentally, and it's easy to deal with because it's classical-seeming, even if the idea of spin is not. But if I wanted to be really, really accurate, I should really say that I'll construct a magnet with a certain waveform like a crystal lattice, and do some field theory to show how the wavefunction of the magnets will interact with the particles (or wave packet if you want to get picky) to give me spin-up particles in a certain area. It's doable, but it's unpleasant! Hypothesizing a measurement is a way to treat part of a problem classically. This is not always the right thing to do (otherwise QM would be worthless), but it's very useful sometimes.

Some physicist will probably have to correct me on some parts of this, but I'm hoping this wasn't too far off. On preview, it looks like Johnny Assay has some more informed thoughts than my ramblings.
posted by kiltedtaco at 7:04 PM on December 29, 2007

when i was a young physicist (well student of really), we learned nothing about observers only observations or more precisely, measurements.

standard quantum mechanics has little to say about how you make a measurement, it is merely a theory about fundamental relationships between those measurements (by which we mean measurements of physics quantities, or really measurements of quantities ultimately dependent on your 'phase space' variables, most commonly position and momentum though not always...)

you can of course add the observation device (including the scientist in charge of it) into your physical problem, you could also add the whole universe into the physical problem: physics problems always idealize the situation by ignoring aspects of the world which you believe are likely to not affect the situation you wish to study and quantum mechanics is no exception. but quantum mechanical predictions (at least the standard theory) don't, in principle, depend upon including the physics of the 'observer' into the experiment you are performing.

i can't say i've ever really thought about it but I believe some schools of thought tried to explain quantum mechanics in terms of some unexplained fundamental interconnectedness between all matter i.e. the matter in the obsevation device is somehow 'one' with matter being observed. unfortunately its hard to delve into the foundations of quantum mechanics without first learning quantum mechanics: often popular books can't avoid the controversies surrounding the foundation (or lack thereof) of QM which I think is unavoidably baffling if you don't have a real (i.e. not popular) understanding coming into it.
posted by geos at 8:45 PM on December 29, 2007

Something worth bearing in mind is that the success of QM demonstrates clearly that the real world, at least when observed at the finest levels of detail, is not precisely predictable: no two experimental setups are 100% identical - not even pairs of experiments involving exactly the same equipment and exactly the same initial conditions, the experiments being separated only by time.

The world just does what it does, and the best we can do is (a) watch it doing that (b) construct some mathematics that will tell us just how likely it is that some prediction we make about what we are about to see will be correct. That mathematical construct is the wave function.

It's important to embrace the idea that we don't know what's going on between the time we measure our initial observables and the time the wave function collapses (i.e. we measure our final observables). We don't even know if it's meaningful to say that anything is going on. All we can work with is what we can observe - and to make that view coherent, it doesn't matter whether that observation is made by a human being, or by some piece of equipment.

Putting this another way: we know that if we set up experimental conditions X, we will get result Y or Z; using the wave function, we can generate really good predictions about what percentage of the time we will get from X to Y vs. X to Z. As to what's "really" going on between X and Y or between X and Z: we don't, and probably can't, know - we simply have no way of making a fog-free map of that territory. If we did, that would necessarily involve some observable or measurable O between X and Y or between X and Z, which makes the experiment different.

You often hear talk about a photon "going through" both slits at once in the classic double-slit experiment, in order to interfere successfully with itself and give rise to the characteristic pattern of light and dark fringes on the photographic plate on the far side. I can't see a way to justify that language. We have a photon emission event on the upstream side of the slits: that's an observable. A little later, we have a photon absorption event downstream of the slits: that's an observable too. The point I'm trying to make is that any talk of "what the photon is doing" betwen emission and absorption is, by its nature, a handwaving Just So story. We have no way to observe the photon in transit, and we really ought to remain silent about its existence during that time and speak only of its wave function.

Given this view, the wave function collapses when an event occurs that has consequences we don't need to invoke wave functions for in order to make good predictions, and the only place left for handwaving is in unpacking the meaning of "good".
posted by flabdablet at 11:16 PM on December 29, 2007

This question does not have a clear, settled answer. As Johnny Assay says, it is an unsettled problem of physics and depends on which interpretation of quantum mechanics you subscribe to.

Here are some examples of what different camps claim constitute an observer:

• An observer is a being with "consciousness." What exactly consciousness is subject to debate, but humans have it and microscopes don't. I think Penrose is an advocate of this idea in some form.
• A collapse occurs when a quantum system reaches a "thermodynamically irreversible state."
• A collapse occurs when a graviton is emitted.
• Collapse occurs instantaneously, but is triggered by waves traveling backwards in time, which creates the illusion of observation causing collapse.
• Collapse doesn't occur. All quantum possibilities occur. Conscious observers can only be aware of one distinct history.

posted by justkevin at 9:10 AM on December 30, 2007

Several good answers here already, particularly Johnny Assay's and JustKevin's (though the list of interpretations is, perhaps inevitably, incomplete).

The linguistic point that bears emphasis is that "observation" is often used where "measurement" ought to be. The reason "measurement" is a better word is that a quantum system can, and almost always does, undergo the irreversible change involved in measurement without a human around to observe it. In a sense, what has been done is to expand the word "observer" to include a certain class of machines, e.g. devices that detect photons. But this is somewhat misleading. The two-slit experiment, for example, does not seem quite as strange when it is understood that the detector apparatus introduced in between the slits and the photographic plate constitutes an "observer".

Not to say that this explains why or how measurement should have this effect, but it at least removes a confusion about the relevance of consciousness to quantum theory.
posted by voltairemodern at 9:50 PM on December 30, 2007

If you go for the Copenhagen Interpretation, which to my mind is ugly, then the observer is essentially any measurement taken of a closed quantum system which transmits information about the encapsulated system to the 'outside'. The problem with this is that you can always reformulate the problem to encapsulate the observer within a new, larger, quantum system.

For example, instead of considering the box containing Schrodinger's cat, and a scientist lifting the lid, you consider a room containing both of them.


Not a physicist, but have studied undergrad QM. Id be interested to hear if the following is a correct interpretation or not.

To me, there doesn't seem to be a paradox. If physicists lift off the lid and make an observation, they (through particle/photon interaction) collapse the wave function. The variable in question is no longer a probabilistic distribution, but is concretely known/measured/observed.

However, if the entire lab is a closed system, with the physicists themselves inside, then no external measurement has been made from outside our lab system. The being outside the lab system has no access to the physicists' measurement - to the external observer, there is still a probabilistic distribution. Even though there may have been particle interaction inside the system when the physicists peeked in the box, this is only seen as the interaction between two wave functions, which result in more wave functions (no collapse).

Thus, from the perspective of the physicists inside the lab, there was a measurement/observation of the system and the wave function has collapsed, but from the perspective of the being outside the laboratory, there has not been any measurement of the system, and the wave function(s) remain probabilistic.
posted by jpdoane at 10:32 PM on December 30, 2007

Kudos to johnny assay for the link to decoherence. Seems like a mature and sensible approach to the observer question. Glad that I wasn't completely bonkers with my first guess, too. But I really like justkevin's last one for trippiness: Collapse doesn't occur. All quantum possibilities occur. Conscious observers can only be aware of one distinct history.

Just thought I'd throw out some praise, since OP didn't mark a best answer ;)

\me wishing that stephen hawking had a mefi account...
posted by Chris4d at 11:11 PM on January 7, 2008

Chris4d: ""observation" really just means the interaction between your observational instrument (photons or whatever) and the subject of observation."

Dipsomaniac: "An observer would be any entity capable of causing a resolution of the quantum state of the situation.Even a mirror could qualify."

DarkForest: "Here's the first hit for google search of quantum+observer. According to this, an observer can be a measurement device."

kiltedtaco: "Whenever you measure an observable (like energy, momentum, spin, etc), you collapse the wavefunction of the system you're measuring... The act of making these measurements changes the system"


All of the statements I quoted above are empirically false. It is not the measurement device or the act of measuring that collapses the wavefunction, but rather only whether a human being has access to the measurement data. This is demonstrated by the Quantum eraser experiment and the Delayed choice quantum eraser.

If the measuring device measures the pathway of the particles, but then erases this data before humans beings have access to it, the screen shows wave interference. Therefore measurement itself does not collapse the wavefunction. Only measurements that would give humans access to the information of the particle's pathway through the double-slit collapse the wavefunction.

Even more disturbing: in the delayed choice experiment, human choices can retroactively determine the outcome of the experiment. If humans are unaware of both the measurement and the pattern on the screen, they can create the outcome of the experiment even though it already happened.

The experiment happened yesterday, do you want the outcome on the screen to show two columns of particles? Then look at the measurement data first.

The experiment happened yesterday, do you want the outcome on the screen to show wave interference bands? Then erase the measurement data and look at the screen.

To answer the original question, physicists can only be referring to human beings when they refer to "the observer", since human beings are the only ones who can perform the various double-split experiments. We couldn't know if Schrödinger's cat's observations do anything. Even if we were to invent an artificially intelligent/conscious robot, or learn to communicate with cats, or meet an advanced race of aliens who told us the wavefunction collapses when they observe it as well, it would not be clear if this happened to them too only because they interacted with us! The delayed-choice experiment suggests that even us meeting the aliens could retroactively determine how they have viewed the experiment for millions of eons.
posted by dgaicun at 3:01 PM on November 8, 2008

Other QM threads have also repeated its the measurement.

a robot made out of meat : "When they put the "camera" next to one slit, it could only be in one place to interact with the camera. Consciousness does not come into this... In QM the "observer" that you always hear about is anything that needs to interact with the particle in a very location-dependent way, not a human brain."

And this thread was the same question:

Jimbob: ""Observed" doesn't mean "viewed by a human". I always thought it really means "measured""

alexei: "it is not the human observer in particular that makes it "choose a path", but the interaction with a photon (which, incidentally, makes it possible to observe it) which does so"
posted by dgaicun at 3:37 PM on November 8, 2008

Another thread with the same repetition:

bhnyc: "consciousness has nothing to to with quantum mechanics. QM is about measurement... an "observer" is just something that records a measurement. it can be mechanical."

ook: ""Observe" is a terrible word in this context, since it implies a conscious observer and causes this sort of confusion, but we're stuck with it by now. "Measure" or "interact" would've been a better choice."

The one comment here by a quantum physicist gets it right:

"First off, I'm a quantum physicist (well, graduate student), and I've asked other quantum physicists (graduate students) how much they knew about quantum erasers. Everyone, so far, has said zip, zero, nada... What the quantum eraser "does" is restore a fundamental interference after a measurement has been taken. . . "
posted by dgaicun at 11:04 AM on November 9, 2008