Social intelligence is the next frontier for AI, researchers say

by Tsinghua University Press

Three inextricably linked aspects of social intelligence—social cognition, theory of mind, and social interaction—are the cognitive tools that will help computer science advance artificial intelligence beyond contemporary models. Photo credit: CAAI Artificial Intelligence Research, Tsinghua University Press

Siri and Google Assistant might be able to schedule meetings on demand, but they don’t have the social savvy to independently prioritize appointments — not yet. According to researchers in China, artificial intelligence (AI) may be intelligent, but it is atrophied by a lack of social skills.

They published their overview of the current state and call for future directions in CAAI research on artificial intelligence.

“Artificial intelligence has transformed our society and our daily lives,” said lead author Lifeng Fan, National Key Laboratory of General Artificial Intelligence, Beijing Institute for General Artificial Intelligence (BIGAI). “What is the next major challenge for AI going forward? We argue that artificial social intelligence (ASI) is the next big frontier.”

ASI, the researchers say, encompasses several isolated subfields, including social cognition, theory of mind—the understanding that others are thinking from their own point of view—and social interaction. By using cognitive science and computer modeling to identify the gap between AI systems and human social intelligence, as well as current problems and future directions, the field will be better equipped to advance, according to Fan.

“ASI is different and challenging compared to our physical understanding of the work; it’s very contextual,” Fan said. “Here, the context could be as big as culture and common sense, or as small as the shared experience of two friends. This unique challenge prevents standard algorithms from addressing ASI problems in real-world environments, which are often complex, ambiguous, dynamic, stochastic, partially observable, and multi-agent.”

READ :  Using AI and simulations, Rowan engineers are helping the Army re-imagine the future of combat | mountain ash today

Therefore, according to Fan, ASI requires a comprehensive approach, since improving certain components of an ASI system does not always result in improved performance – unlike modern AI systems. Rather, ASI requires the ability to interpret latent social cues such as eye rolls or yawns, to understand other agents’ mental states such as beliefs and intentions, and to collaborate on a shared task.

“Multidisciplinary research informs and inspires the study of ASI: The study of human social intelligence provides insight into the foundations, curriculum, points of comparison and benchmarks required to develop ASI with human-like characteristics,” said Fan.

“We focus on the three most important and inextricably linked aspects of social intelligence: social cognition, theory of mind, and social interaction because they are based on established theories in cognitive science and are readily available tools for developing computational models in these areas. ”

According to Fan, the best approach is a more holistic approach that mimics how people interact with each other and the world around them. This requires an open and interactive environment and considerations on how to introduce better human-like distortions into ASI models.

“To accelerate future progress of ASI, we recommend taking a more holistic approach, just like humans do, to integrate different learning methods such as lifelong learning, multi-task learning, one/few-shot learning, meta-learning etc.,” said Fan.

“We need to define new problems, create new environments and datasets, set up new assessment protocols and create new computational models. The ultimate goal is to infuse AI with high-level ASI and increase human well-being with the help of Artificial Social Intelligence.”

READ :  Data is key to an economic future

For more information, see Lifeng Fan et al., Artificial Social Intelligence: A Comparative and Holistic View, CAAI Artificial Intelligence Research (2023). DOI: 10.26599/AIR.2022.9150010

Provided by Tsinghua University Press