You need a language which enables generalization. Sadly, most computer languages lack good facilities for this. Human language works, though:

A monad is some parametric type, T, along with two functions called (in Haskell anyway) "return" and "bind". Return "injects" values into the type taking values of type A to values of type T(A). Bind transforms values of type T(A) into values of type T(B) using a function like A -> T(B).

Then these two functions must follow a few rules.

That's a monad. The Haskell fragment above describes the signatures of those functions, notes how they relate to the parametric type, and also produces a facility for overloading `return` and `(>>=)` ("bind") and even working with them when lacking a concrete choice of type `T`.