Tech companies are currently focused on rolling out new updates and integrations to enhance the capabilities of artificial intelligence (AI) to improve productivity and enhance user experience. However, while AI-powered chatbots can be asked to schedule events for you or create appointments, artificial social intelligence requires further exploration and advancements, according to researchers based in China.
A recent study, published in CAAI Artificial Intelligence Research, claims that “artificial social intelligence (ASI) will play a crucial role in shaping the future of artificial intelligence (AI).” The review article explored ASI from a cognitive science standpoint, including social perception, theory of mind (ToM), and social interaction.
According to the paper, ASI has been mostly disregarded in the AI community, “with only scattered applications.” The article refers to the Defense Advanced Research Projects Agency, which believes that AI should include the human-like skill of contextual adaptation — "the capacity to reason about and adapt to various contextual inputs.”
ASI includes the ability to interpret subtle social cues such as yawning or eye-rolling, understand beliefs and intentions, and work together towards a shared goal, according to lead author Lifeng Fan, from the Beijing Institute for General Artificial Intelligence (BIGAI), an IANS report explains.
“ASI is distinct and challenging compared to our physical understanding of the work; it is highly context-dependent,” Fan said. He further explained that the most effective method is to adopt a comprehensive approach that imitates people’s interactions with each other and their environment.
The paper concluded with four directions to inspire future work on ASI. According to the research, there is need for a more holistic approach to social intelligence than existing computational models, which often focus on “a single aspect of the problem.” It also emphasises that lifelong or continuous learning is a crucial path for developing ASI.
The communication should be open-ended and interactive, which could aid in better coordination and cooperation, solving tasks, and learning strategies similar “to social intelligence in humans, and even more complex behaviour,” the research adds.
Finally, better biases, even structural biases, as a form of built-in common sense should be introduced. These biases could be “hard-coded, evolve from interactions with other agents, or be taught by humans.”