This is a very good question, so let me give a lowbrow answer and a highbrow answer.
The lowbrow answer is that homotopies of topological spaces induce this relation on their associated chain complexes (the ones you use to define singular homology, for example), so it makes sense to call this relation homotopy. See the proof of homotopy invariance of homology in Hatcher's text, for example.
The highbrow answer is that one can axiomatize homotopy theory, which gives the notion of a model category. In a model category, we use an "interval object" to define the notion of homotopy. For example, in the category of topological spaces, the interval object is $[0,1]$. It turns out one can define an interval object for the category of chain complexes, and if one works out the abstract definition of homotopy in the category of chain complexes, we get exactly the definition you gave.
For more on the interval object, see here.