Self Reflection
“No one can make you feel inferior without your consent.”
Eleanor Roosevelt
Self Reflection is a body positivity project combining hardware and software to help users visualize the effect that self talk has on how we see ourselves.
Prototype Testing
How does this work?
Audio and video are captured through a webcam attached to the mirror. The mirror continuously displays the viewer on a screen behind the reflective surface.
The audio is transcribed using speech recognition and fed to a CBT-trained chatbot via DialogFlow API.
This produces a sentiment score, represented on a scale of -1.0 (negative sentiment) to 1.0 (positive sentiment). This triggers a two-fold response with both audio and visual components.
Conversational response: The mirror responds to the viewer with thought-provoking prompts about the way they spoke to themselves, using both text transcription and text-to-speech audio feedback. For example, if the user criticized their looks, the CBT-informed response might be “What makes you say that about yourself?” or “Are there things you do like about your appearance?” which encourages the user to further examine their negative self perception.
Image response: Based on the sentiment score, the video feed is then passed through a generative adversarial network (GAN) which alters the person’s reflection in real time according to the user’s positivity or negativity. The sequenced images will either appear crisper and brighter in response to positive verbal input, or will be distorted by various types of pixel manipulation in response to negative input, making it increasingly difficult for the user to see themselves as they truly are until positive verbal input is reintroduced.
These integrated models run on a Raspberry Pi embeded in the screen behind the mirror’s surface, and will engage with the user until the discussion is concluded and a new conversation begins. Throughout these interactions, the models employ continuous learning to iteratively improve the fidelity of future conversations.
Designed with lengthy user research and cognitive behavioral therapy (CBT) techniques in mind, Self Reflection is a study in Human-Computer Interaction (HCI) which transforms the positive or negative things the viewer says about themselves into a corresponding visual effect. For instance, if the viewer calls themselves ugly or says they don’t like themselves, their image in the mirror becomes more distorted, whereas positive speech gives the viewer a clearer image of themselves as they truly are - beautiful and worthy of appreciation, even (and especially) from themselves.
Google Colab notebook running on a Raspberry Pi embedded in the screen
Workflow Diagram
Many thanks to Futurity for their support of this project. To check out what’s happening with Self Reflection now, or to see any of Futurity’s very cool moonshot projects, head over to futurity.systems!