It can also help to remember that memories are far from totally accurate video recordings of an event. They are highly-lossfully-compressed, and there's some evidence that the act of recalling them is destructive, so every time a memory is recalled it is destroyed, processed, and re-written.
It isn't surprising that when a very out-of-the-usual event occurs, and the memory goes through this lossy compression and lossy storage, that when it is later reconstructed with the lossy decompression schema generated by a lifetime of "usual" experiences that it would result in phantom "usual experiences" being generated. It is very similar in process to those "deep learning" pictures that turn everything picture you feed it into bizarre mixes of dogs and sea creatures, because that's all that decompression schema understands.
> It can also help to remember that memories are far from totally accurate video recordings of an event. They are highly-lossfully-compressed, and there's some evidence that the act of recalling them is destructive
We have counter-examples of eidetic memory for vision, conversations etcetera so any model has to allow for near-perfect recall.
It isn't surprising that when a very out-of-the-usual event occurs, and the memory goes through this lossy compression and lossy storage, that when it is later reconstructed with the lossy decompression schema generated by a lifetime of "usual" experiences that it would result in phantom "usual experiences" being generated. It is very similar in process to those "deep learning" pictures that turn everything picture you feed it into bizarre mixes of dogs and sea creatures, because that's all that decompression schema understands.