Avatars have been found to be useful tools to overcome communication barriers in people affected by Autistic Spectrum Disorders (ASD) and to help them understand and express emotions. However, it has been shown that the success of the interactions is highly dependent on the subject's identification with the avatar. In this study, we assess the variables that may influence that perception in children under 10, in the context of the largely multi-ethnic Ecuadorean society. The results reveal that, unlike previous studies showed for young adults, the ethnic traits displayed by the avatars are not a critical factor, as the the quality of the interactions was more influenced by the perception of the avatars' appearance, their similarity with the kids' peers and, above all, the ability of the human model who controls the avatar to use a pleasant voice, to succeed in making his/her questions and responses fully understood, and to master the non-verbal communication transmitted through gestures and voice.