Workshop Program

Date: September 16th, 2025.

Location: ICDL 2025 conference, Faculty of Electrical Engineering CVUT in Prague.

Time: The workshop will start at 14.00 CEST (GMT+2) .

Total duration: 3.5 hours.

Accepted Contributions

Touch as a Stress Buffer: A Bio-Inspired Robot Model for Adaptive Social Interaction Based on Hormone Regulation

Lorenzo Aloe, Salvatore Anzalone, Elisabetta Zibetti, Lola Cañamero

Stress can easily disrupt social behaviour. Drawing on the social buffering hypothesis and the impact of touch on stress dynamics, this paper proposes a robot model in which touch modulates two synthetic hormone signals, cortisol and oxytocin, to regulate stress, affecting the robot’s behavior.

Can a Robot Become Addicted? Modeling Addictive Behavior Using Incentive Salience Theory

Maria Gabriela Valeriano, Letícia Mara Berto, Ricardo Gudwin, Alexandre Simões, Esther Colombini

Motivation is central to mobility, as animals must continually choose actions that best serve their goals. Hence, any goal-directed movement can be understood as motivated behavior. However, there is no consensus on a unified theory of motivation, but the aspects of want and like are consistently present in the literature. In this work, we bridge motivation research with neuroscience, demonstrating how incentive salience theory reconciles  apparent contradictions within the wanting-liking dichotomy and offers a possible explanation for addiction. We present the formulation of a  computational model along with a proposal for experimental evaluation. Finally, we present future directions for a more comprehensive and integrative modeling of motivational processes.

Here We Go Again: Social Context Recognition and Prediction for Real-World Social Robots

Dario Pasquali, Francesco Rea, Giulio Sandini, Alessandra Sciutti

Autonomous robots operating in real-world, dynamic social environments must be able to quickly learn and adapt to their surroundings, the people within them, and evolving social dynamics. We previously introduced a novel cognitive architecture for the iCub robot and demonstrated its ability to autonomously segment multimodal perceptions and form in-memory associations of first-person experiences. In this manuscript, we build upon those results to evaluate whether the architecture’s associative memory can: (i) form stable associations of the robot’s experiences over time; (ii) accurately label new experiences based on previously stored ones; and (iii) leverage the temporal distribution of past experiences to forecast the most likely social context at a given moment. Our findings support the feasibility of using the architecture’s associative memory to enable context-aware behavior in real-world social environments.

For any questions, please contact us at aicogdev.workshop@gmail.com