Can AI understand human emotions?

If we are to use AI to interpret, for example, heart rate and brain activity and predict people’s emotions, the programming must be done differently, says Eduardo Gutierrez Maestro.
Can AI systems grasp the complexity of human emotions? Eduardo Gutierrez Maestro at Örebro University is researching how machine learning can be applied to health data and how it can learn that humans do not experience one emotion at a time.
More information about the research at Centre for Applied Autonomous Sensor Systems (AASS).
“If we want AI systems to act based on our emotions, we need to consider context. For instance, people won’t experience the same feeling when watching a comedy film after receiving bad news. I often compare it to the film ‘Inside Out’, where memories are described as a blend of different memories. It’s the same with emotions,” says Eduardo Gutierrez Maestro, a researcher in computer science at Örebro University.
Eduardo Gutierrez Maestro combines different scientific disciplines to enhance machine learning. The idea of emotional priming, derived from social psychology, explains how previous experiences shape our emotions. In computer science, a common issue is the lack of clear labels when AI systems learn from content such as images or other data. There is limited research on how machine learning should tackle these two challenges.
AI can help older people in living longer at home
“If we are to use AI to interpret, for example, heart rate and brain activity and predict people’s emotions, the programming must be done differently. Then AI systems can monitor patients in a hospital even more effectively and raise the alarm if something is about to happen. Older people can stay at home longer with the help of, for example, a robot that better monitors their emotional state, improving their quality of life,” explains Eduardo Gutierrez Maestro.
The traditional method is to encode a single, specific emotion as data so an AI system can learn that emotion. But emotions are not unambiguous. Eduardo Gutierrez Maestro’s research aims to teach data models, combinations, or distributions of different emotions.
Enabling the monitoring of mental health
“The possibilities for monitoring mental health and well-being are continuously expanding. My research contributes by proposing a methodology that enables models to better understand human emotions. This will allow these models to be developed more effectively in the future,” says Eduardo Gutierrez Maestro.
Örebro University has been researching how robots can promote good health for over 15 years, and AI and robotics are now utilised in both healthcare and the private homes of older adults.
Text: Björn Sundin
Translation: Jerry Gray