Tell me who I am

I am not what I think I am, and I am not what you think I am. I am what I think you think I am.

Charles Horton Cooley, sociologist

There is something so oddly satisfying about being told who you are.

Especially by a system. Consider: you take a personality test to figure out your type. Are you an INTJ or Enneagram 3? A dating app predicts who you might like. Do you swipe left or right? A fitness tracker tells you how well you slept. At 3:45am, your sleep was disrupted. Spotify tells you what kind of listener you are. You judge others based on their Spotify Wrapped list. Now, AI systems can produce extremely convincing interpretations of our personality, emotions, preferences, and behavior. We might treat this as a technological development. Upon closer inspection, it seems to me that our desire to understand ourselves through external systems goes back further.

In 1902, sociologist Charles Horton Cooley described the “looking-glass self” theory. It was the idea that our sense of who we are develops partly through how we imagine others perceive us. We see ourselves through the reactions, judgments, and expectations of the people around us. In other words, self-knowledge has never been entirely internal. We are, after all, a mosaic of anything and everything we have ever experienced, including the people who have reflected parts of ourselves back to us.

But what happens when the mirror changes? Psychology and psychometrics have spent decades developing ways to describe and measure human characteristics that are inherently difficult to observe directly: intelligence, personality, ability, behavior, etc. These systems can be useful, but measurement requires abstraction by definition. Something complicated, continuous, and nebulous has to solidify into an identifiable category, score, trait, or variable.

This is where I think the distinction between being described and being understood is critical. Description looks for patterns, while understanding asks what those patterns mean. Spotify can identify that I listen to a particular genre of music without knowing what that music means to me. The Oura Ring 4 can recognize that I sleep poorly without knowing that I was awake because I was worried about my mom’s upcoming chemotherapy treatment. Systems can identify recurring patterns in the way I communicate without knowing the experiences that shaped them.

In conclusion: context changes meaning. Two people can exhibit the exact same behavior for completely different reasons. And the same person can exhibit completely different behavior under different circumstances. A system that becomes extremely accurate at recognizing patterns can still be working with an incomplete representation of the person producing them. Yet we find these representations so compelling because they give us something that our own experiences often lack: coherence.

Someone tells you that you are an introvert, and suddenly a collection of experiences seems to fit together. An algorithm repeatedly recommends the same kinds of things, and it becomes evidence of what you genuinely prefer. An AI system describes your behavior in language that feels intensely personal, and you may start using that language to understand yourself.

The system has moved from observing us to participating in how we understand ourselves. This creates a fierce feedback loop. We build systems to identify patterns in human behavior, which produce “objective” interpretations. We engage with those interpretations and respond to them, so our behavior changes as a result. The system receives new information and updates its model. Rinse and repeat, rinse and repeat.

In this new reality, the mirror is no longer passive, blank, and waiting. It selects what to measure, determines which patterns are meaningful, makes inferences from incomplete information, and gives those inferences back to us. AI makes this particularly complicated because it can, in a word, “speak” back to us. A test score may tell us where we fall on a specified, narrow scale, but a conversational system can tell us a story about who we are and why we behave the way we do. That story may feel remarkably useful and accurate, but resemblance isn’t the same as understanding. I think true understanding requires something that measurement cannot fully provide: context, contradiction, history, and meaning.

So the question isn’t whether technology can tell us who we are, but rather what happens when we start using its interpretation to answer that question for ourselves.