You can probably answer this question without thinking too hard about it. Being aware of our own thought processes — a capacity called introspection — seems to come naturally to us. Yet among philosophers and psychologists, introspection is a subject of active debate.
Some think it’s closely related to consciousness. Others argue that it’s an elaborate illusion. It’s tricky to test theories of introspection in the lab — or to pin down a precise definition in the first place.
Now, after over a century of studying introspection in humans and animals, researchers are beginning to explore the phenomenon in unlikely new subjects — large language models. These systems, which are based on mathematical structures called neural networks, excel at cognitive tasks that until recently seemed exclusive to humans. Can they introspect as well?
In 2025, Jack Lindsey, a researcher at Anthropic, devised a way to test introspection in language models by injecting “thoughts” into their neural networks — in a kind of targeted digital neurosurgery. Lindsey’s experiment demonstrated that models could at least sometimes detect and describe these injections. It looked a lot like introspection.
Lindsey’s paper was deliberately limited in scope — it didn’t make broader claims about consciousness. (In particular, it did not imply that language models have subjective first-person experiences.) Even so, it caused a stir online. Researchers rushed to replicate Lindsey’s study with other models. In the year since, a clearer picture has emerged.
Some aspects of the phenomenon Lindsey observed have simpler explanations unrelated to introspection. Others have proved harder to explain away, but whether they should count as introspection remains hotly contested. It’s a heady time for the field, as researchers devise real-world tests for questions that were once considered philosophical.
Extract — continue reading at the source.
