Picture the region between y=x and y=x2 for 0≤x≤1, a thin crescent. Slice it into vertical strips. At x=0.5 the line is at 0.5 and the parabola at 0.25, so that strip is 0.25 tall. Every strip's height is top minus bottom, x−x2, and adding strips of width dx gives ∫01(x−x2)dx=61.
The method works wherever the curves sit, above or below the axis. Shift both curves down by 5 and each strip keeps its height, so the area is unchanged. That is the difference from a single definite integral, which counts area below the axis as negative. Here the subtraction measures a length, and a length is positive as long as you take the upper curve minus the lower one.
The limits usually come from solving f(x)=g(x), since the curves pinch the region closed where they meet. If they cross inside the interval, top and bottom swap, so split the integral at the crossing. For y=x3 and y=x on [−1,1], the cubic is on top on [−1,0] and the line on top on [0,1]. One integral of x−x3 over [−1,1] gives 0, because the two pieces cancel, while splitting gives the true area 41+41=21.
Think of it like this. Two strings pinned across a page, one above the other, with paint poured between them: the paint needed is the area. At each x the depth of paint is the top string's height minus the bottom string's, however far above the edge of the page the strings sit. Where the strings cross, the paint pinches off and a new pool begins.
Core idea. Area between two curves is ∫ab[f(x)−g(x)]dx where f is the upper curve over [a,b].