This project was selected to represent MIT at the Microsoft Design Expo 2020 on ‘Healthier Futures.’
A tele-mental health platform that uses generative AI avatars for preserving a person’s privacy and identity while allowing for expressivity.
Mental health concerns are growing, and the need for digital solutions are on the rise—especially in the area of tele-counseling services such as hotlines, chatrooms, online communities, and mental health apps. However, patients generally have concerns around both their privacy and their data security. This is particularly because mental health issues can be very stigmatizing.
One common method for preserving patient privacy is leveraging “anonymity,” which is particularly helpful for those who feel vulnerable sharing information about their health. Anonymity can help people to avoid feelings of embarrassment or shame. It allows people to be more open and honest, knowing that their identity is secure. However, anonymity is typically associated with a number of disadvantages as well. From a therapist’s perspective, there’s a higher chance of misunderstanding. When they cannot see the other client’s face, emotional information that would normally be put forth by their facial expressions is lost. Additionally, there is a lower sense of empathy and trust that can develop between the two parties. As such, those surveyed about anonymity with regards to online interventions for mental health have even mentioned that the anonymous experience can feel very impersonal or dehumanizing.
We present a preliminary prototype of iNonymize, a tele-mental health platform that uses generative AI avatars to preserve a user’s privacy and identity while allowing for expressivity (i.e., the transmission of emotion through facial expressions and voice). Our approach uses generative AI to allow someone to swap the look of their face and change the sound of their voice. In this way, emotional signals can be carried forward, while not being tied to the user’s identity.
We also envision this pipeline being useful at different levels. Beyond the experience of helping people connect with a counsellor, iNonymize can also be used to the facilitate and strengthen interpersonal connections in support groups, and help with community outreach.
- Application Concept:
Generative AI Avatars for Empathetic Anonymity in Tele-Mental Health
- Visual Mock-Up:
Mobile app to help people anonymously connect with mental health professionals
- Early Technical Prototype:
Pre-rendered avatars generated using the First Order Motion Model by Siarohin et al. 
Microsoft Design Expo 2020 on "Healthier Futures" (October 29 2020)
 Siarohin, Aliaksandr, Stéphane Lathuilière, Sergey Tulyakov, Elisa Ricci, and Nicu Sebe. "First order motion model for image animation." arXiv preprint arXiv:2003.00196 (2020).