conditioning

How sensitive a mathematical problem's own answer is to a small change in its input -- a property of the problem itself, not of whatever method is used to solve it.

A problem is well-conditioned if a small relative change in its input produces only a comparably small relative change in its exact output, and ill-conditioned if a small input change can produce a disproportionately large output change. Conditioning is a property of the mathematical problem alone, independent of the algorithm used to solve it.

Examples

Finding the roots of $(x-1)(x-1.0001)\approx0$ is ill-conditioned: nudging one coefficient by $10^{-8}$ can move a root by orders of magnitude more than that.