Another way of looking at it is that the force acting on a mass (e.g. due to a spring) is generally a function of the displacement, F = F(x). Generally, F(x) can be a complicated function. If the force has an equilibrium point, say x = 0, then for small displacements about x = 0 the force can be approximated by
[tex]F(x) = F'(0) x + F''(0) x^2/2 + \cdot \cdot \cdot[/tex]
which is just a Taylor series expansion (remember, F(0) = 0 since x = 0 is an equilibrium point). If we restrict ourselves to sufficiently small values of x then [itex]F(x)[/itex] is well-approximated by keeping only the leading term so we write [itex]F(x) = F'(0) x[/itex]. Again, since we're dealing with motion about an equilibrium then F'(0) must be a negative number (i.e. the force is a restoring force).
It is convenient to represent F'(0) with a number like -k where k is positive and is often simply called the "spring constant." The process I've just described is called "linearization" and is generally applicable when dealing with small oscillations. Of course, there are many situations where the linear approximation does not hold such as when a spring is stretched or compressed in such a way that it becomes permanently deformed (i.e. the mass will not return to it's starting point!)
Since the linear equations do apply to many situations and provide considerable insight into how things work the time it takes to learn and use them is time well spent.