Australian researchers develop AI that understands human emotions

Australian researchers have unveiled a method to make artificial intelligence (AI) systems more emotionally aware, laying the groundwork for more empathetic and effective human-machine interactions.

Not yet applied to physical robots, the research could soon influence emotionally intelligent systems in mental health support, customer service, and education worldwide, according to a release from the Edith Cowan University (ECU) in Western Australia published on the Science Media Exchange website on Wednesday.

Researchers at ECU have moved beyond conventional techniques that analyze single facial images. Instead, they train AI by presenting sets of related facial expressions, allowing machines to interpret emotions in a richer, more human-like context, the release said.

"Just like we don't judge how someone feels from one glance, our method uses multiple expressions to make more informed predictions," said ECU PhD student Sharjeel Tahir, who is also the study's lead author.

This approach boosts reliability, even when faces appear at different angles or under varying lighting, said Tahir.

Exposing AI to diverse visual cues within structured groups not only increases accuracy but also keeps the technology efficient, said co-author ECU PhD student Nima Mirnateghi.

The team is now focused on developing artificial empathy, enabling AI agents to respond appropriately to human emotions, and making AI decision-making more transparent and understandable.

The work is laying the foundation for machines that can truly understand and respond to people's feelings, according to ECU senior lecturer and AI expert Syed Afaq Shah who led the research team.