loading . . . Introspection is an illusion created by the brain _Many have assumed that self-knowledge begins with looking inward. But psychologist**Nick Chater** argues that this is impossible: the mind does not store beliefs and desires to be uncovered, it invents them on the spot. The brain is a brilliant improviser, coming up with explanations for our behaviour the instant we are asked for them. Shattering this illusion of inner depth exposes the therapeutic search for a "true self" as a futile pursuit, renders billion-dollar consumer surveys useless, and reveals that our demand for perfectly transparent AI is deeply misguided. The task of being human is not self-discovery, Chater suggests, but the creative act of self-authorship._
For millennia, people have tried not merely to look outward at the external world, but to turn inward and examine the workings of their own minds—as if there were an inner eye that could observe our mental life.
Yet a synthesis of decades of research in psychology and neuroscience shows that the very idea of introspection is an illusion. And for a surprising reason. It is not merely that we find it difficult to accurately perceive our inner motives, beliefs, principles, and desires (or that these are repressed, as Freud suggested). The problem is more fundamental: there _are_ no such stable beliefs and desires “inside” us that can be observed and reported. Instead, the human mind is a wonderfully fluent, but profoundly deceptive, improviser: spinning stories justifying our thoughts and actions as fast as we ask questions. And these invented explanations are vague, inconsistent, and often provably wrong.
Consider, for example, the wonderfully clever experiments by Petter Johansson and Lars Hall and their colleagues at Lund University in Sweden. They gave people pairs of faces presented on cards and asked them to select their preferred face. They then handed people the card with their chosen face and asked them to explain their preference. On a small number of trials, however, by using close-up card magic, they tricked people by handing them the wrong card—the face they had _not_ chosen.
In the great majority of cases, not only do people fail to notice the switch, but they happily and fluently justify the choice they _didn’t actually make_ ; and they do so just as confidently as for the choices they _did_ make. If justifications came from introspection—looking inside our minds to discern the “true” explanation for our preferences, they would of course treat the two cases very differently—because when the cards were switched, the justification would make no sense. Yet not only do people not notice—they often explain their (apparent) choice with comments like “it’s those nice earrings” when the face they originally chose wasn’t even wearing earrings. The justifications must be mere rationalizations—improvisations fabricated in the moment they are asked for—rather than the product of genuine introspection.
Let’s extend that thought a little further. For most people, the areas of the brain concerned with language are primarily concentrated in the left hemisphere. Suppose, then, it were possible to give people a task that could only be carried out by the right hemisphere (without “telling” the left hemisphere what it was doing). In that case, the left hemisphere would have no way, in principle, of knowing—or introspecting—the true origin of the behaviour. But if our explanations of our own minds are no more than fluent fabrications, the left hemisphere should still be able to come up with some (wrong, but plausible) explanation for the right hemisphere’s choice.
Remarkably, it is possible to test exactly this prediction, as shown by cognitive neuroscientist Michael Gazzaniga in a long series of important studies with split-brain patients—people for whom the bundle of nerve fibres connecting the two hemispheres, the corpus callosum, has been surgically severed (once a treatment for severe epilepsy). People with split brains are able to lead remarkably normal lives and show only a modest IQ deficit.
___
The human mind is a wonderfully fluent, but profoundly deceptive, improviser: spinning stories justifying our thoughts and actions as fast as we ask questions.
___
The right hemisphere surveys the left half of the visual world and controls the left side of the body due to the strange cross-over wiring of the human brain. By presenting images to the left visual field and asking people to respond with their left hand, researchers could give participants tasks they could successfully perform, but to which the language centres of the left hemisphere had no access.
In one of Gazzaniga’s famous studies, a split-brain patient looks at a fixation point in the middle of the screen, and two images are shown simultaneously: to the left, a snow scene; to the right, a chicken’s claw. They are told to select, from a range of picture cards in front of them, which card best fits with the image they have seen. Crucially, they are asked to respond with their _left_ hand—controlled by the side of the brain that has seen the _left_ image, the snowy scene. It therefore picks out a related picture: a snow shovel.
The patient is then asked why they chose the snow shovel. It turns out that, due to the strange crossover wiring of the human brain, the left visual field and the left hand are both wired through to the _right_ half of the brain. But the language centres are overwhelmingly on the _left_ half of the brain. And the left half of the brain saw a chicken, not a snowy scene at all.
So when the person is asked to explain why they chose the snow shovel, you would expect that they should be baffled. The left half of the brain knows nothing about the snow scene and how it led to the choice of a snow shovel. The left half of the brain should surely be dumbfounded.
But not at all. The patient’s language centres chirp up without hesitation: “Oh, that’s simple. The chicken claw goes with the chicken, and you need a shovel to clean out the chicken shed.’ Ingenuous. Convincing. But totally wrong: the real reason the left hand selected the shovel is nothing to do with the chicken’s claw—the right side of the brain (controlling the left hand) saw only a snow scene after all. But the language system fabricates a plausible story out, as it were, of thin air.
Gazzaniga has a lovely metaphor: that the brain’s language centres are its PR office, not its CEO; and the real shock is that it turns out that the CEO never talks to the PR office! It simply acts, while the PR office scrambles together the most convincing running commentary that it can. Yet the PR office is so fluent and compelling that we are fooled into thinking it has a direct line to the CEO; in reality, it is no more than a ceaseless imaginative spinner of stories. It is not introspecting—it is inventing.
___
We learn about other people, albeit provisionally and imperfectly, from _interacting_ with them, not through inspecting the contents of their brains.
___
I set out this argument in my 2018 book, _The Mind is Flat_. Since then, the rise of generative AI has provided a striking embodiment of the same idea. We can ask “large language models” (LLMs) like ChatGPT, Claude or Gemini to explain their responses. But in doing so they do not—and cannot—reach “inward” to recover the true causes for their responses. Rather, LLMs are merely lucid storytellers, inventing explanations on demand—explanations that have no relation to activity across billions of artificial neurons that really drive their behavior.
The human mind is no different: we are each post-hoc commentators on, and apologists for, the results of the electrical and chemical activity of the intricate network of roughly one hundred billion neurons that compose each human brain, and whose true workings are as opaque to us as the biochemistry of digestion.
Introspection, then, is an illusion. It may be rather dizzying to realize that no inner eye with which can probe our supposed inner life—and indeed, that we have no inner life (beyond the flow of conscious experience) to probe. But why does it matter? Let me pick out three ways (of many) in which it matters a lot. __
First, from self-discovery to authorship. Our lives are complex, and we continually struggle with conflicting aims and objectives. It is very tempting to imagine that only by looking inward,we can discover who we really are. But we can no more introspect the operations of our brain than we can the workings of our pancreas or liver by introspection. What we can do, though, is actively reflect on and construct our lives: to create, test, and revise the stories we tell about ourselves. These narratives help us make sense of what we do and guide what we do next. The task is not self-discovery, but the creative activity of self-authorship.
SUGGESTED VIEWING The real me With Joanna Kavenna, Steve Taylor, Jack Symes, Frank Tallis, Anneli Jefferson
Second, beware surveys! We need to give up the idea that we can find out what other people feel and believe simply by asking them. If introspection is unreliable, then people’s reports about their own minds are not windows onto underlying states. Instead, we must understand both ourselves and others by attending to patterns of behavior—what people actually do, over time and across contexts. Market researchers, take note: focus on what people _do_ , not _why_ they think they do it.
Finally, we should rethink AI explainability. Generative AI systems have long been “black boxes” from the point of view of their users. But so are people. Indeed, as we’ve seen, the illusion of introspection tells us that we are black boxes _even to ourselves_. Yet this does not prevent us from successfully predicting, coordinating with, and relying on others—working out who has which expertise, who to trust, and who is on our side. We learn about other people, albeit provisionally and imperfectly, from _interacting_ with them, not through inspecting the contents of their brains. So if we can interact—and converse—with an AI agent (and, one day, a physical robot) in a way that allows us to “get along” and work together successfully, that’s all we need. The running commentary from the AI as we work together won’t be a revelation of its actual inner workings, but will be fabricated in the moment. But that’s nothing new—humans have been improvising stories about our own minds for as long as we’ve had language in which to express them. https://iai.tv/articles/introspection-is-an-illusion-created-by-the-brain-auid-3573