Artificial Metacognition: Giving an AI the ability to “think” about its “thinking”

Have you ever had the experience of reading a sentence several times and then realizing that you still don’t understand it? As many new college freshmen have been taught, when you feel yourself getting off track, it’s time to change your approach.

This process of becoming aware that something isn’t working and then changing what you’re doing is the essence of metacognition, or thinking about thinking.

It is your brain that monitors its own thinking, detects a problem, and controls or adjusts your approach. In fact, metacognition is fundamental to human intelligence and, until recently, has been little explored in artificial intelligence systems.

My colleagues Charles Courchaine, Hefei Qiu and Joshua Iacoboni and I are working to change that. We have developed a mathematical framework that allows generative AI systems, especially large language models like ChatGPT or Claude, to monitor and regulate their own internal “cognitive” processes. In a way…