> It made me think how does this happen in the brain is the brain reading data from the memory where it shouldn't?
Kinda, yes. But the analogy breaks down, like the other user noted, because human memory is not stored in buffers, and not copied around from buffer-to-buffer. It's just not a good way to reason about the brain.
The mental model I use to think of it is "overconnectedness". Pattern-matching first fires on spurious associations more than usual, and the regions of the brain that would normally filter that out are turned down. The result is an almost dream-like erratic association of surface meanings and deep meanings in an incoherent way. Signals turning inwards, amplified in loops until the noise turns into imagined patterns.
The analogy to dreaming, unlike a computer, probably has some real substance to it. When you dream, certain parts of the brain become more inhibited, while other parts become disinhibited and start speaking more to each other. Same with the state experienced on a drug like LSD, for that matter.
This lines up with what we understand of the drugs that seem to intersect with this. Drugs that affect certain kinds of neurochemical receptors, particularly serotonin and dopamine receptors, can either induce LSD-style hallucinations (for drugs that increase signalling) or or, in some schizophrenic people, reduce or even largely shut off hallucinations and delusional thinking (if they reduce signalling).
There's definitely more to the puzzle than just this piece, though. (For one thing, anti-psychotic drugs are pretty good at shutting down hallucinations but not so good at addressing the rest of the scattered thinking typical in schizophrenia. Which suggests there's more than just a senses-gazing-inward thing going on. It's also quite possible schizophrenia is actually several distinct-by-cause but symptomatically related diseases.)
