The role of computational intelligence (CI) in science and engineering has traditionally been to serve as an assistive tool for predictive analytics and decision making. The growing capacity of modern algorithms to process and uncover hidden patterns in large volumes of data promises to transform them into scientific discovery machines. However, a purely data-driven model does not guarantee compatibility with basic scientific principles that must underlie any new discovery. This special session is thus focused on state-of-the-art research that augments CI algorithms by melding data with scientific knowledge. Such knowledge is traditionally expressed as formal differential equations describing known physical laws, and more recently, in the form of approximate models of intuitive physics. The resulting physics-informed CI can not only carry useful inductive biases to learn better surrogate models, but also enable the direct solution of complex differential equations, the solving of inverse problems with hidden physics, and more efficient search for novel yet optimized engineering designs.