Quantum eraser

Wheeler's delayed-choice experiment was designed to make this vivid, and his own conclusion was the opposite of the popular one: no elementary phenomenon is a phenomenon until it is registered, so there is no earlier fact about the route for a later decision to alter. Critics grant that this is consistent and press the usual complaint — the account explains the correlations by declining to describe what produced them.

Marking the path entangles the particle with a second system, and interference between two paths is lost exactly when some other system could tell them apart. Erasing measures that system in a basis that both routes answer equally, so the branches it defines are superpositions of the routes rather than one route each. The apparent puzzle — fringes appearing after the fact — is the ordinary observation that a sum of two opposite fringe patterns is flat.

In de Broglie–Bohm the guiding wave always passes both slits. Marking the route makes the two branches of the wave occupy different states of the marker, so they no longer overlap and the guidance equation stops producing fringes. That is a permanent fact about each individual run. What sorting by an erased marker does is pick out the runs whose marker later gave one answer, and within that sub-ensemble the trajectory endpoints happen to be distributed as fringes — a statistical selection, not a retroactive change to any trajectory.

In standard quantum mechanics erasing gets harder as the record spreads into more degrees of freedom, but nothing forbids it in principle — the obstacle is bookkeeping. GRW and CSL make it a matter of physics: past the collapse scale the superposition is genuinely destroyed, so the subsets can never show fringes again however the marker is interrogated. That is a difference of prediction rather than of story, which is why this family is the one experiments can constrain.

For a QBist the delayed choice is not even puzzling: an agent's expectations are updated when the agent learns something, and updating a probability about a past event has never required influencing it. The fringes are a feature of a conditional expectation, not of the screen. Critics accept that this dissolves the mystery cleanly and ask the standing question — what must the world be like, that agents' gambles are constrained in exactly this way?

Rovelli's account makes the delayed choice unremarkable by denying the assumption that generates the paradox: that there is one observer-independent list of what happened, fixed at the moment of detection. Facts are established in interactions, relative to the systems that interact. Comparing the marker's record with the screen's is a further interaction, and it travels at the speed everything else does. The cost is the absolute account itself.

This is one of the places the framework machinery earns its keep. Both families are legitimate and each assigns ordinary probabilities to complete histories; what the formalism forbids is the single-framework question the paradox needs, namely whether the particle went through one slit and produced fringes. Griffiths argues that the delayed-choice puzzles dissolve once that combination is recognised as ill-formed rather than mysterious. The standing objection to the whole approach is that nothing in it says which framework describes the world, only which are permitted.