According to recent studies on mental health, one in ten children are affected by a serious psychological problem, with future projections showing an alarming increase of this trend. That’s why we have taken the initiative to explore new models of engagement and investigate the potential use of empathy applied to human–machine interaction.
In mental health everybody is different, so it is important that every voice is heard.
During a 5 month deep-dive we gained first-hand insight into non-intrusive mechanisms for prevention of depression in childhood and adolescence, collaborating with key experts in the field, children, and parents through an open and co-creative process.
By bringing together design thinking, Artificial Intelligence and the principles of crowdsourcing, FINE enables a digital friend to react empathetically to a child’s emotional state.
A machine learning ‘empathetic’ model has been trained to read and react to emotions appropriately, with a corresponding family hub displaying the child’s and family member’s collective mood over time, acting as a central trigger to the habit-forming routine of talking about emotions at home; and encouraging the kind of positive behavior change that leads to a preventative and collective caretaking of how one feels.
A joined-up ecosystem of three different working prototypes that combine to create a comprehensive ‘empathy system’
Close collaboration between Method, Fitzrovia Youth in Action, NHS Tavistock Trust and MIND in the London area
Open co-creation process with rich cross-cultural teams in different European locations
Making children the experts
An open and non-stigmatizing approach was key in our research process. With co-creation sessions, we invited children and parents to share their experiences and create low-fi prototypes of future concepts. This had a strong influence on our design concepts, which received positive praise in a series of subsequent validation exercises.
No emotion in Artifical Intelligence, yet
As part of our research, we found that no dataset currently exists which would enable to teach a machine learning model how to be empathetic. As a result, we designed and created an application that captures empathetic responses to emotional stimuli. Through a crowdsourcing initiative, we generated enough data to train a model over time on how to respond to different moods detected by analyzing the human face. This showcases the potential use cases of new forms of human machine interaction; one we can experience right now.