Design and evaluation of an AI avatar for speech based completion of health screening questionnaires : a case study with the Autism-Spectrum Quotient (AQ)
Ojeabulu, Jude; Abara, Callistus (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025121737602
https://urn.fi/URN:NBN:fi:amk-2025121737602
Tiivistelmä
Currently, mental health assessments, especially those for neurodiverse individuals often rely on text based formats that are impersonal, and these can feel inaccessible, clinical or even lack the necessary empathetic support that is necessary for a comfortable experience. This in turn leads to the hinderance of accessible and trustworthy screening.
In response, this thesis investigates the design and implementation of Auralis, an empathetic AI voice companion developed to help users navigate through the Autism-Spectrum Quotient (AQ). Our research investigates how speech enabled AI can make self assessment more comfortable, accessible, and especially ethically mindful of the whole assessment process. Integrated into the Auralis system are speech recognition, text to speech, and a finite state dialogue manager. These are powered by an empathy engine developed to foster a calm, neuro affirming, and user centered experience.
The prototype was developed using Python, FastAPI, and Streamlit alongside OpenAI's GPT 4o models, that gives room for more natural hands free interaction. Human centered AI principles were adhered to; informed consent and GDPR compliant privacy were the foundational pillars. Evaluation results demonstrated that Auralis achieved high user satisfaction and effective empathetic delivery. Participants described the system as calm, trustworthy, and non judgmental, confirming its potential to improve accessibility for neurodivergent individuals. Overall, this research contributes to the field of trustworthy AI by showing how emotional intelligence and ethical safeguards can coexist within and help digital healthcare systems to create technology that listens to while comforting and the same empowers neurodivergent patients.
In response, this thesis investigates the design and implementation of Auralis, an empathetic AI voice companion developed to help users navigate through the Autism-Spectrum Quotient (AQ). Our research investigates how speech enabled AI can make self assessment more comfortable, accessible, and especially ethically mindful of the whole assessment process. Integrated into the Auralis system are speech recognition, text to speech, and a finite state dialogue manager. These are powered by an empathy engine developed to foster a calm, neuro affirming, and user centered experience.
The prototype was developed using Python, FastAPI, and Streamlit alongside OpenAI's GPT 4o models, that gives room for more natural hands free interaction. Human centered AI principles were adhered to; informed consent and GDPR compliant privacy were the foundational pillars. Evaluation results demonstrated that Auralis achieved high user satisfaction and effective empathetic delivery. Participants described the system as calm, trustworthy, and non judgmental, confirming its potential to improve accessibility for neurodivergent individuals. Overall, this research contributes to the field of trustworthy AI by showing how emotional intelligence and ethical safeguards can coexist within and help digital healthcare systems to create technology that listens to while comforting and the same empowers neurodivergent patients.