This talk explores the potential of conceptual spaces as a novel layer in robot architecture to enhance the interaction and communication of children diagnosed with Autism Spectrum Disorder who use social robots’ assistance. Drawing on insights from cognitive theories and empirical observations, we propose a model that utilises conceptual spaces to better explain and thanks to this simulate the unique cognitive processes of autistic individuals. Application of our model will facilitate more effective interventions using robots, thereby enabling social engagement and reducing the communication barriers faced by autistic children. The work is in collaboration with Antonio Chella (Palermo) and Peter Gärdenfors (Lund).