Fairly recently excellent quantum eraser experiments, such as the Mach-Zehnder interferometer with cold helium atoms, have been carried out, and in 2007 the highly advanced quantum eraser by Giulino Scarcelli et al. with ‘two-photon imaging’ has completely confirmed to the experimenters – and to me – that speaking of a photon that travels a path is not right. There exists definitely only a non-material quantum state wave until the measurement is done.

Entangled photons carrying their joint information
The experimenters took advantage of the fact that the state wave of the two photons – signal and idler – is a single entangled wave, which means that the state wave of the idler photon reflects, as it were, the passage of the signal photon through the physical double slit in a kind of ‘ghost’ image. This ghost image can then be used to ‘look at’ the slit without measuring and disturbing the signal photon itself.
Here too, the information about which slit was chosen by the signal – after the slit passage but still before detection – is erased randomly by a semi-transparent mirror. This ensures that half of the idler photons, as determined by objective quantum coincidence, will lose the information about the passage of the entangled signal through the double slit. This 50-50 random distribution is provided by the semi-transparent mirror NPBS.

After the semi-transparent mirror, the idler photon, which carries the slit information of the signal photon, arrives at one of the two conical receptors PT or PR. The PT receptor cone is almost completely covered with a cover having only a small hole in the middle. The slit information that the idler carried is irretrievably lost. The PR receptor is virtually uncovered so that the slit information that the idler carries is preserved. The photon detection together with the slit information is then passed on via a fiber optic cable to detector D2, which in turn passes it on to the coincidence counter CC. Thus, whether or not the slit information is lost is carried out by the semi-transparent mirror NPBS.
Back in time?
The setup is such that the signal photon should already have passed the slit at the moment the idler photon reaches the semi-transparent mirror NPBS. Thus, the ‘action’ to delete the slit information is made after the slit has already been passed. Yet that delete ‘action’ has the effect that prior to that action the signal photon will not appear in the slit. When the delete action is not carried out, the photon will appear in the slit. This is called retrocausality, a retroactive effect in time.

The figure below shows the result of the experiment. There is interference – wave behavior – when the information about the chosen slit is irreversibly erased by the passage of the idler through NPBS. There is no interference – particle behavior – if that information is preserved by the reflection of the idler on NPBS.

In my opinion the most parsimonous interpretation is that the eventual observation, the processing of the information, creates history. There was no material photon traveling at all.
Is it about information?
What I still miss is an experimental version where it will become even clearer that it is about information. If only the irrevocable erasure of information is the cause of the shown wave behavior, erasure can just as well happen after the detectors, by electric QNRG controlled switches. When it is an effect of the information, the outcome should be identical – the interference appears. The practical design of such an experiment is even a lot simpler and easier to understand.