The aim of CONTACT, established in March 2019, is to enable robots to understand and predict humans’ intentions, internal states, and limitations while being transparent, predictable, and adaptable in their own behaviors. To achieve this goal, we investigate the sensory, motor, and cognitive bases of human social abilities, by using robots as ideal, controllable probes to test and model the dynamics of human interaction. We then study how to integrate these abilities into a cognitive architecture, also relying on memory, motivation, and anticipation, to support social awareness, adaptability, and autonomous learning. As a result, we contribute to a more profound comprehension of human cognition through a constructive and embodied approach, with the technological goal of developing more intuitive and adaptable robots.