Researchers have been analysing how humanoid robots relate to people and say teaching the machines how to understand emotion may be essential in getting the best from them.
Professor Bouchra Senadji is Head of Engineering at James Cook University and co-author of a study investigating the role of empathy in long-term human-robot interactions (HRI).
She said robots are increasingly used to support childhood learning and development - including emotional skills - and in senior care to support older adults.
"But studies involving humans interacting with a social robot over several meetings have shown the robots can have difficulties in sustaining human engagement over time," said Professor Senadji.
She said that previous research seeking to improve the situation pointed to the important role of empathy in long-term HRI.
"While empathy involves both sharing in another person's feelings and understanding how another person feels, it is particularly important that a robot understands the emotions of its user in order to respond appropriately," said Professor Senadji.
She said that emotions expressed by the robot's human partner are used as cues to evaluate the individual's emotional state. This emotional state is then used to decide the robot's response.
"We found that the robots studied used a variety of techniques, including measuring eye gaze, analysing vocal patterns and facial cues and posture to gauge how their interaction with humans was going.
"Whatever the behavioural model used, all studies ultimately relied on the emotion state of the user to inform a robot's response," said Professor Senadji.
She said the findings highlight how important understanding emotion is in informing a social robot's responses and helping sustain user engagement.
"This suggests that social robots need to be able to learn from their users and continuously adapt their response. To get the best out of these tools we have to recognise that a robot's behavioural response has to be based on the emotion state of the person it's dealing with and optimise the machine's abilities to judge this and respond appropriately," said Professor Senadji.
Link to paper here.
