Why does observation collapse the wave function? In one of the first lectures on QM we are always taught about Young's experiment and how particles behave either as waves or as particles depending on whether or not they are being observed. I want to know what about observation causes this change? 
 A: In different interpretations of quantum mechanics the definition of "measurement" is different. But I think it would be enough if I give just five of which you can choose yourself.


*

*In Copenhagen/von Neuman interpretations the collapse of the wave function is triggered by the observer. This person has the special property which no other object in universe is capable of. In Copenhagen interpretation the collapse can be triggered by any system which is connected to the observer, including the measurement apparatus and external medium (if the observer is not isolated from it). All things can be arbitrarily divided into the observed system and the measuring system by so-called "Heisenberg cut" with the only requirement the measuring system include the observer. 

*The von Neuman interpretation is the edge case of Copenhagen interpretation where the Heisenberg cut is placed as close to the observer as possible. As such even the parts of his brain still be be considered the part of the observed system. In von Neuman interpretation the collapse of the wave function happens when the observer feels any qualia(feeling) depended on the measured value.

*In Bohm interpretation the collapse of the wave function happens when the observer introduces into the measured system some perturbation, which is inevitable when performing the measurement. The difference between the measurement and any other interaction is in that the perturbation introduced by measurement is unknown beforehand. This is because initial conditions of a system containing the observer are unknown. In other words, the observer always contains information which is unknown and cannot be determined by any means due to self-reference problem. Thomas Breuer called this phenomenon "subjective decoherence". The philosophers believe that this unpredictability of the system containing the observer for himself, defines the free will.

*In Relational interpretation the collapse happens when the interaction affects the ultimate measurement performed by ultimate observer on the universal wave function at infinite future. As such, for the collapse to happen the result of interaction should somehow affect the external medium, the stars, etc, either now or in the future, rather than being recohered and lost. 

*In Many-worlds interpretation the wavefunction collapse never happens. Instead what the observer perceives as the collapse is just the event of entanglement of the observer with the observed system.
A: In the following answer I am going to refer to the unitary evolution of a quantum state vector (basically Schrodinger's Equation which provide the rate of change with respect to time of the quantum state or wave function) as $\mathbf{U}$. I am going to refer to the state vector reduction (collapse of the wave function) as $\mathbf{R}$. It is important to note that these two processes are separate and distinct. $\mathbf{U}$ is understood well and can be modelled accurately with the equations of QM, $\mathbf{R}$ is not well understood and it is some physicist's thoughts that QM will need to be modified to incorporate this state vector reduction process.
There is much to say about the $\mathbf{R}$ process, but I will address your question directly; basically "is it consciousness that reduces the state vector/collaspes the wave function?". Among those who take this explanation seriously as a description of the physical world, there are those who would argue that - as some alternative to trusting $\mathbf{U}$ at all scale and believing in a many-world type view point - that something of the nature of this $\mathbf{R}$ process occurs whenever the consciousness of an observer becomes involved. E. Wigner once sketched a theory of this kind in Nature in the 60s. The general idea was that unconscious matter or inanimate matter, would evolve according to $\mathbf{U}$, but as soon as a conscious entity becomes physically entangled with the state, then something new comes in and actually reduces the state (some $\mathbf{R}$ process). 
The posit that it is consciousness that causes this collapse is very hard to debunk, due to the very nature of this type of argument. However, if you consider the following example, it should be clear that this picture is far from complete; and that this argument for consciousness causing the $\mathbf{R}$ process is not sufficient. Consider the weather, the detailed weather patterns that occur on any planet, being dependent of chaotic processes, which much be sensitive to numerous individual quantum events. if the $\mathbf{R}$ process does not actually take place in the absence of consciousness, then no particular weather pattern could ever establish itself out of the morass of quantum-superposed alternatives. Can we really believe that the weather on these planets remain in complex-number superpositions of innumerable distinct possibilities - just some total hazy mess quite different from actual weather - until some conscious being becomes aware of it and then at that point, and only that point the superposed weather becomes actual weather? I don't think so - do you?
Personally I think we can expect some amendment to QM if this process $\mathbf{R}$ is ever going to be sufficiently explained. One candidate model to explain this reduction process is the gravitationally induced state-vector (and its decedents). There are strong reasons for suspecting that the modification of quantum theory (QT) that will be needed, if some form of $\mathbf{R}$ is to be made into a real physical process, must involve the effect of gravity in a serious way. Some of these reasons have to do with the fact that the very framework of standard QT fit uncomfortably with the curved-space-time that GR demands. Yet most physicists seem reluctant to accept that it maybe QT that needs adjustment to facilitate a successful union with GR. Roger Penrose describe a new model (based on other candidates) in his book The Shadows of the Mind (not an easy read!) that uses a quantum gravity model to explain the elusive quantum process $\mathbf{R}$ - this is well worth a read if you want a better understanding of this mysterious process and it implication on human consciousness.
I hope this helps.
A: Towards a better picture of the duality
On Young's double-slit experiment the wave-particle duality (one by one photon) is more a problem of "picture of the model" than a philosophic  one: see Y. Couder interpretation, by your self  (!),
Youtube Couder experiments
The quantum particle HAVE a location, there are only a limitation in choose a good pictoric model when you is constrained by "wave or particle" picture options: Couder demonstrates that a good picture,
of an "intermediary wave/particle object" model, exists!
Imagine a "localizable object" that haven't a well-defined boundary, but have a well-defined distance-limit (~lambda) to interact with obstacles (other objects).
On this little video you see the objects one by one, changing (or not)  the rectilinear trajectory  by the "oscilating interaction" with the (double-slit) obstacle, not  by the after-obstacle-screen as "observer".
There are an article online about the experiment.
PS: of course, if a observation constitutes measurement before  the screen, it will be interfere on result, changing the interference pattern at the screen.
A: The wavefunction is not a material object. It is not a wavey process in 3 dimensional space. (as is seen as soon as you consider the wavefunction of two or more particles in the many body problem). It is a mathematical object in 3n dimensional configuration space where n is the number of interacting particles. It essentially contains all the statistical information about a system that it is possible to have- kind of like a giant list. If you make a measurement you effectively add a condition that the system obeys so reducing the the possibilities and so you are now considering a subset of the original list. This is what the collapse of the wavefunction is. This is why a measurement can collapse the wavefunction everywhere instantaneously rather than propagating out from the measurement location at the speed of light as it would if the wavefunction were some sort of material thing.
A: An electron, indeed any particle, is neither a particle nor a wave. Describing the electron as a particle is a mathematical model that works well in some circumstances while describing it as a wave is a different mathematical model that works well in other circumstances. When you choose to do some calculation of the electron's behaviour that treats it either as a particle or as a wave, you're not saying the electron is a particle or is a wave: you're just choosing the mathematical model that makes it easiest to do the calculation.
The next question is OK what is an electron then? At the moment our best description is that the electron is an excitation of a quantum field. Using quantum field theory allows us to calculate the behaviour of electrons whether they happen to be involved in particle-like or wave-like interactions. This doesn't mean that the electron is a quantum field, and we'll almost certainly replace quantum field theory by some even more complicated e.g. some future development of string theory.
The collapse of the wavefunction is a separate issue, and one that has generated much debtae over the years. I think the general consensus is that the collapse of the wavefunction is a manifestation of a more general process called decoherence.
A: They act as waves and particles all of the time.  In order to make a measurement one must interact with the system, and so you can not observe the particle without interacting with it, and thus the measurement changes it.  
A simple case would be a single electron.  In order to see the electron a photon must hit the electron and that would change something about the electron.
A: Imagine a billiard ball as it rolls on the flat billiard board in a straight line with constant velocity as expected. Then suddenly it changes direction.
What happened? As you may think the billiard ball is not alone, there are other balls that collided with it.
That's same situation with the wave function collapse. We are dealing with a partial system not the whole one. You have the wave function of a single particle which evolves according to a Hamiltonian, then suddenly collapses. 
The situation is like the one with the billiard balls. The particle is not alone. The state contains the measurement apparatus, you, and the rest of the universe as well. You are part of that ket too not just the particle.
So the actual state is the superposition of the following:
an universe where you will measure your electron spin up,
and an universe where you will measure your electron spin down.
As long as your test electron is on the way the two universes are indistinguishable: the probabilities of finding anything in any particular state is the same. 
But this doesn't mean the two wave functions are identical: there is a complex phase shift between the electron wave functions in the two universes. 
This phase shift causes an interference between the probability amplitudes which in turn can cause observable differences  when you run the particle through multiple Stern-Gerlach apparatuses.
This phase is enough to make the electron go spin up in the Stern-Gerlach aparatus in one of the universes and down in the another.
After this split happens the two universes are now distinguisable cannot interfere, but at this point you can still run your electron beams through a reverse Stern-Gerlach apparatus to merge the two electron beams. At this point the two universes are indistinguishable again and interfere again.
Now after some split and merge you decide to measure the electron spin.
To do that you have to make some, whatever, state change in your universe that allows you collect data. This can be as simple as directing your electron beams into a two detectors. In one of the universes it will go the  up detector, in the one universes it will go to the down detector.
Now at this point the two universes have become permanently distinguishable, and the probability amplitudes cannot interfere anymore.
In one of the universe you collected a spin up data, entered it in a file, and published the data where that particular data line will show a spin up detection. In the other universe you collected a spin down data, entered it in a file, and published the data where the particular data line will show a spin down detection.
Once you did the measurement there is no way back, it cannot be undone, you cannot get the two universes indistinguishable and make them interfere again. And if you think about it again you cannot make any particle measurement without making splits like this.
But even if the the two states are diverging now, they are still in superposition, there is no collapse. We merely discard the states that didn't happen because they are not relevant. You in the state where the particle went spin up will discard the other state where it went spin down, and vice versa. And both versions of you might be wondering what caused the particle go spin down or spin up.
This superposition is a bit like the radio stations on the air, where the universe is the broadcast itself. If two stations are indistinguishable and radiate on the same frequency there will be interference and you hear loud whistling in the radio. But once one of the stations change frequency the whistling and the interference will be gone and won't be heard anymore. But their waves still on the air and mix but the beat frequency between them will be too high to cause any noticeable effect. Now you have two separate broadcasts that doesn't cause whistle in each other but it's still there.
So you don't just have the wave function of particle, you have the wave function of the entire universe in superposition. This universal state evolves according to a Hamiltonian as usual and doesn't collapse.
