New Study Highlights Opportunities for Artificial Emotional Intelligence

New Study Highlights Opportunities for Artificial Emotional Intelligence

How will artificial intelligence (AI) impact the future of the mental health industry? One pioneering group at the Massachusetts Institute of Technology (MIT) is applying emotion AI to improve mental health and overall quality of life. Recently the Affective Computing Research Group at the MIT Media Lab published a new studythat provides empirical evidence that empathetic artificial intelligence (AI) machine learning can counterbalance the adverse effects of anger on human creative problem solving.

The MIT study involved over a thousand participants to play a word guessing game, Wordle, to see how anger and empathy impacts performance. Those assigned to the anger elicitation condition performed poorly compared to the control group. The research shows that an empathic AI agent can reduce the negative impact of anger on creative problem solving.

Emotion AI, also known as artificial emotional intelligence, is a subset of AI that reacts, simulates, quantifies, and comprehends human emotions. Affective computing is computing that relates to, arises from, or deliberately influences emotions. The field was introduced in 1995 in a paper authored by Rosalind W. Picard, Sc.D., FIEEE, a MIT professor of media arts and sciences, Principal Investigator at MIT’s Jameel Clinic, the founder and director of the Affective Computing Research Group at the MIT Media Lab, and co-founder of MIT spinouts Affectiva (now part of Smart Eye) and Empatica. In the paper, Picard suggests that computers need the ability to recognize and express affect in order to interact in a natural and intelligent manner with humans.

“We’re looking at opportunities where AI can be of immediate help,” said Picard. “Since there aren’t enough therapists for people suffering from depression and anxiety—the two largest sources of mental illness—we’re looking at what can be done with technology to support those people. And in particular, we’re thinking about not just healthcare, which is misnamed because it’s just really sick care, but about real healthcare—before people are depressed, diagnosed with Major Depressive Disorder or anxiety disorder.”

One of the current research projects at MIT’s Jameel Clinic is creating personalized AI machine learning for improving mental health. Roughly 22 percent of undergraduates at MIT suffer from severe or moderately severe depression according to MIT. The project goals are to develop a method to monitor changes in Major Depressive Disorder levels and forecast the trajectory, identify the key variables, and enable individualized time-adaptive treatments. The aim is to use personalized digital medicine to enable the delivery of optimal MDD prevention or treatment to underserved populations, breaking down the barriers of time, cost, availability, and trust.

Last year MIT unveiled a study at the International Conference on Machine Learning (ICML) 2021 that improved the accuracy of AI machine learning algorithms for predicting clinical Hamilton Depression Rating Scale scores (HDRS-17). The Hamilton Rating Scale for Depression is a questionnaire developed by Max Hamilton at the Department of Psychiatry, University of Leeds, that was published in the Journal of Neurology, Neurosurgery, and Psychiatry in 1960. It is commonly used to assess clinical depression, also called Major Depressive Disorder (MDD), a serious mental health disorder that can limit a person’s ability to perform daily activities. The MIT researchers used mixed effects random forests (MERF) and some labelled patient data when training the algorithm. As a resulting, their AI outperformed standard random forests and personal average baselines when predicting clinical Hamilton Depression Rating Scale scores.

This is just one of the many MIT research projects at the intersection of AI machine learning and mental health. Other projects include a scalable AI recommender system to customize therapy recommendations to improve mental health based on the user’s current context, past preferences, and similar users, a study that examines whether humans or AI are better at detecting deepfake videos, the creation of three AI computer vision modeling methods to predict a driver’s stress levels based on the driving scene, a digital mental health intervention game called The Guardians: Unite the Realms that engages users to complete behavior activation interventions, and many more research projects.

“With respect to mental health and AI, I think there are a lot of opportunities,” said Picard. “We have a lot to learn from the psychological community, we are not trying to replace them. I think building an AI that replaces people in general is a really bad idea. What we’re looking at is how to build the kind of AI that augments and expands what people can do.”

Images Powered by Shutterstock