University of Arizona to develop socially intelligent AI agent

The agents will gather information about social interactions to help real world teams achieve their goals

Researchers from the University of Arizona (UA) have been granted US$7.5m to build an artificial intelligence agent that is fluent in social cues and human interactions, hoping they can use this valuable data to help real world human teams recognise and achieve their goals.

The funding comes from the Defense Advanced Research Projects Agency (DARPA) as part of DARPA’s Artificial Social Intelligence for Successful Teams programme.

“The goal of the ASIST programme is to develop artificial intelligence with a ‘theory of mind’, and create AI that is a good teammate to humans,” said Adarsh Pyarelal, principal investigator and research scientist at UA’s Machine Learning for Artificial Intelligence Lab.

While AI agents such as Google Assistant, Siri and Alexa already exist, their skills lie in finding and presenting information. These devices are not yet able to read the complex nuances of human social cues and interactions.

“The thing that makes a human a good teammate is having a good sense of what other people on their team are thinking, so they can predict what their teammates are going to do with some level of certainty,” Mr Pyarelal explained. “We’re trying to build AI with some social awareness and this kind of theory in mind.”

Advertisement


You might also like: New digital start-up to promote sustainable international education


The project, called ToMCAT (Theory of Mind-Based Cognitive Architecture Teams), has been funded for four years. Working with one to four human players on their teams, the agents will complete specially tailored Minecraft missions, using digital and physical sensors to collect data on individual players and their interactions as they progress through the game.

The study will use a network of webcams and microphones to track the eye movements, facial expressions and voices of human players, while an electrocardiogram machine will monitor their heart rates. Researchers will also use two brain monitoring techniques to measure players’ brain activity. Eventually, the AI agents should have gathered enough information on the team’s goals and social dynamics to offer meaningful suggestions for improvement via text.

“So everything comes full circle – the agent observes, it learns, and then, if needed, it can intervene to help the team,” added Pyarelal.

Robert C. Robbins, president of UA, said: “Artificial intelligence and machine learning are increasingly present and important in our everyday lives, and the development of new technologies and understanding in this area is a key focus for the University of Arizona. With DARPA’s support, our talented researchers have a tremendous opportunity to take the next step toward making AI even more beneficial to our daily routines and interactions.”

Advertisement

JOIN THE LIVE WEBINAR

3 Strategies To Avoid Data Disaster at Your Academy Trust