This article presents a new design framework for the specification and prototyping of geometrically and behaviorally complex materials with graded properties, coined predictive information modeling (PIM). The contribution is the development of new circular design workflows employing machine learning for predicting fabrication files based on performance and design requirements. The aim is linking endogenous capacities as well as exogenous environmental dynamics of graded materials, as an approach to material focused intelligent design systems. Using two experimental case studies, the research demonstrates PIM as an applied design framework for addressing (1) material uncertainty, (2) multi-scale data integration, and (3) cyclical fabrication workflows. Through the analysis of these models, we demonstrate research methods that are validated for design applications, review their implications, and discuss further trajectories.
