The ability to understand others' affective states and adjust behavior accordingly is crucial in social relationships. Enabling robots with these capabilities can advance their development as social agents and improve Human-Robot Interaction (HRI). This work introduces a cognitive framework to enhance social robots' adaptability and perception as intentional agents, aiming to foster personalized HRI. Inspired by attachment theory, the proposed framework endows robots with basic cognitive functionalities - such as perception and decision-making - and an internal motivation defined as comfort, which is dynamically influenced by the robot's needs and goals and by the stimuli it receives during the interaction. This innovative approach aims to allow robots to align their behavior with their partner's preferences and adapt accordingly. By monitoring their comfort-based motivation, robots can personalize the interaction with users, promoting more predictable and immediate exchanges.

Interacting with partners that understand our desire of closeness or space and adapt their behavior accordingly is an important factor in social interaction, since the perception of others is a fundamental prerequisite for reliable interaction. In human-human interaction (HHI), this information can be inferred by a person's attachment style - a person's characteristic way of forming relationships, modulating behavior (i.e ways to give or seek support) and, on a biological level, their hormone dynamics. Enabling robots to understand their partners' attachment style could enhance robot's perception of partners and help them on how adapt behaviors during an interaction. In this direction, we wish to use the relationship between attachment style and cortisol, to equip the humanoid robot iCub with an internal cortisol-inspired framework that allows it to infer the participant's attachment style and drives it to adapt its behavior accordingly.