Modeling Modular Software Structure for Human Understanding

People form internal mental models of the things they interact with in order to understand those interactions. This psychological insight has been used by the human-computer interaction (HCI) community to build software systems that are more intuitive for end users, but it has only been informally applied to the problems of software designers, programmers, and maintainers. Conventional programming languages still do little to help client programmers develop good mental models of software subsystems. To address this problem, we have developed the Abstract and Concrete Templates and Instances (ACTI) model of modular, parameterized software subsystems. This model of software structure addresses the needs of human software engineers who must reason about collections of interacting software parts during design, maintenance, and evolution. ACTI is diierent from other module systems and models of software in several ways. In ACTI, a subsystem never has any implicit dependencies, and never depends directly on any external deenitions|all external dependencies are described through an explicit interface. In addition, a subsystem speciication is meaningful by itself, even without respect to any implementation. Finally, a subsystem is more than just a collection of types and operations; it also includes: an explicit model of behavior, an explicit model of all external dependencies, a collection of deenitions used to construct and describe these models, and (potentially complex) substructure. There are strong parallels between ACTI and other research on the understanding of modularly structured physical devices, particularly Functional Representation.