Consider a mouse who lives its whole life in a maze in a lab. If you asked the mouse about its understanding of the universe, it might say "it's 3 inches high and 3 inches wide, has two left turns and three right turns, and there's cheese at the end". Of course the scientist outside the maze has a different--and much broader--conception of the universe. This is because the mouse's powers of observation are limited by its enclosure in the maze. The scientist knows that her model of the universe is more accurate, but the mouse's is consistent with the observable data, and useful for its purposes.
Or consider pre-Galilean humanity. Their experience of the world was consistent with a flat Earth: that model matches the observable data, and is useful enough for travel on human scales. Once the technology and analytic capacity to observe the stars and planets was invented (and a few centuries later, accepted), the flat-Earth model was discarded in favor of a broader one.
Is there a word, phrase, or theory that describes this manner in which (1) a model can be consistent with all local observable data, but later, (2) in the face of more data, broadened, adapted, or replaced?
I think there are a few interrelated ideas here, so any pointers to relevant ideas or literature are appreciated!