Computers are primarily used for automating numerical calculation but they can also be programmed to make inferences through symbolic calculation. Powerful automated inference is key to a flexible and reliable computing infrastructure. Such technology can be used to build and analyze abstract models of complex systems, find bugs and optimizations in software and hardware designs, and in the design of more expressive programming languages. These engineering applications of inference technology require lightweight inference tools that are embeddable and composable.
The Infer! project is a collaborative effort between SRI International, Clarkson University, and Pace University to construct an open foundational infrastructure for inferential software components. The Infer! project provides an abstract interface to a range of basic inference capabilities through a logic middleware layer. The Infer! middleware also includes mechanisms for combining and mapping between different theories and levels of abstraction. The theoretical and pedagogical foundations for the Infer! project are developed in a discipline called "inferential analysis" that studies efficient inference components and the architectural framework within which they can be combined. The Infer! technology is used to explore novel and exciting applications including the qualitative modeling of complex phenomena and high-assurance open-source software.
Broader impacts in research and education are sought for this project by decomposing traditionally monolithic inference frameworks into useful components that can be exchanged in open source form and embedded within open verification technology. A key part of the project is an outreach effort to establish standards and interfaces for an open exchange of components within the verification and analysis research community. This effort includes active cooperation with the open source software community through the development of verification components that can be integrated for use in constructing dependable, verified, and self-checking software.
