While Abstract State Machines (ASMs) provide a general purpose
development method, it is advantageous to provide extensions that
ease their use in particular application areas. The talk focuses
on such extensions for the benefit of a ``refinement calculus'' in
the area of data-intensive systems. We show that providing typed ASMs helps to exploit the existing logical formalisms used in this area to define a ground model and standard refinement rules. We also show that the extensions do not increase the expressiveness of ASMs, as each typed ASM will be equivalent to an ``ordinary'' one.