Assessing Synthetic Voices for Mental Health Chatbots

Considering the possibility to exploit information and communication technologies (ICT) and specifically speaking chatbots, in the mental health domain, a study is proposed aimed at testing the perceptual features of different synthetic voices considering some fundamental aspects of human–computer interaction, namely users’ acceptance and expectations. More specifically, the effect of synthetic voices’ gender and quality on user’s preferences were investigated. The study involved 40 participants, recruited in Northern Ireland, split into two groups: mental health experts and participants living with depression and/or anxiety. Six synthetic voices, three females and three males, characterized by different quality levels were developed for the experiment, exploiting free online synthesizers. The Virtual Agent Voice Acceptance Questionnaire (VAVAQ) was used to collect data regarding preferences toward the different synthetic voices. Results showed that participants’ preferences seem to be affected by both the gender and the quality of a synthetic voice. In particular, participants preferred female voices and high-quality voices. Results also seem to suggest that the quality of a synthetic voice could have a stronger impact on users’ evaluations compared to voice’s gender.