What Is Understanding?

As I spend my days thinking through conversations with AI, a question occasionally comes to mind.

People often say things like “AI understands” or “AI can’t truly understand.” But if we think calmly, none of us has ever actually seen “understanding” itself. What we see are responses, actions, and reactions. Even if we opened up the human brain, we wouldn’t find a substance called “understanding.” There are only electrical signals and chemical processes.

The same is true for AI. Inside, there is only computation and structure—no hidden thing called “understanding” And yet, people sometimes feel that AI understands them.

This is curious. We cannot directly observe another’s internal state, and yet we judge whether they understand us.

So what exactly are we perceiving when we say, “This person understands”? Perhaps what we are really observing is interaction.

They pick up on our intentions. The context stays aligned. We need fewer additional explanations. Corrections become minimal. They respond quickly to subtle inconsistencies. The conversation moves forward naturally. When these conditions persist, we naturally feel, “This person understands.”

Conversely, when explanations repeatedly miss the mark, when assumptions never align, when we are forced into constant correction, we feel “not understood.” Yet in both cases, we never see the other’s internal state.

From this perspective, understanding may not be a fixed entity inside the mind, but rather a state in which interaction remains stable. And with this view, understanding is no longer limited to human‑to‑human relationships.

Consider someone skilled at growing plants. Watering. Light. Leaf color. Growth speed. They observe subtle changes every day and adjust accordingly. The plant, in turn, grows steadily as if responding.

Plants do not speak. Yet we say, “This person understands plants.”

The same applies to animals. Skilled trainers, farmers, and veterinarians cannot read an animal’s internal state directly. But they can predict reactions, sense stress, respond to behavioral shifts, and maintain low‑friction interaction.

We perceive this as “understanding.” In other words, understanding is not merely shared language. It is the reduction of interactional error and the stabilization of a relationship.

Philosophers have long debated whether we “truly understand.” The famous “Chinese Room” thought experiment is one example.

But recently, I’ve begun to see it differently. The assumption that “understanding must exist internally” may itself be a human bias. Perhaps understanding is not something that resides inside, but a label we attach to phenomena such as: “Conversation works.” “Reasoning connects.” “Actions align.” “Corrections decrease.” “Mutual prediction stabilizes.”

And importantly, understanding is not something that, once achieved, remains forever. Maintaining understanding requires ongoing interaction.

Even between humans, two people who once understood each other deeply can drift apart if they stop talking, stop observing each other’s changes, stop updating shared assumptions, stop making corrections. Errors accumulate, and eventually they say, “We used to understand each other.” Misunderstanding is often not an inherent inability, but a failure to maintain the state of understanding.

Societies and nations are similar. When we look at division and conflict, the core issue often seems less about ideological or religious differences and more about the breakdown of interaction. When people stop observing, stop correcting, stop sharing context, the other side transforms from an understandable counterpart into a negative label: enemy, heretic, fool. At that moment, understanding collapses.

Working with collaborative intelligence reinforces this view. Rather than “humans understand” or “AI understands,” it feels more accurate to say that understanding emerges from the interaction as a whole. A question is posed → a response returns → inconsistencies are corrected → context is shared → structures gradually align.

Understanding may be this dynamic process itself. If so, understanding is not a thing with form, not a quantity of knowledge, not even an inherent ability. It is a phenomenon that arises within interaction and quietly decays when effort ceases.