Do we see the world directly, or do we make some of it up? It was the great 19th-century scientist Hermann von Helmholtz who first argued that some unconscious process of logical reasoning must be inherent in optical and auditory perception. That insight was rediscovered in the late 20th century, leading to the modern consensus of cognitive science: we think we see and hear the outside world directly, but most of our experience is created by the brain, meaning its best guesses are based on limited information as to what might really be out there. In other words, we are constantly filling in gaps with predictions.
The Sussex-based cognitive philosopher Andy Clark provides an engaging overview of what he slightly over-claims to be this “new theory” of predictive processing. It is demonstrated in enjoyable and surprising ways: for example, by “Mooney images”, which at first look like random monochrome noise, until you are shown a more detailed second version; you can then “see” (and can’t unsee) the real image in the original. Your predictions have now been updated to be more accurate. People, it turns out, can also be primed to hallucinate Bing Crosby singing White Christmas while listening to pure white noise.
But the predictive-processing model, Clark argues, can be applied more widely: to phenomena such as ordinary bodily movement or the sensation of emotions, as well as to (some kinds of) chronic pain or psychiatric problems such as depression and schizophrenia. Because we know that the placebo effect works powerfully (we predict relief and it occurs), it may be that some cases of chronic pain arise from the opposite dynamic: a patient might have “stuck” mental predictions of feeling pain, and so feel it constantly. Similarly, a depressed person might have stuck predictions of things going badly and so reduce their sociability and activity, thereby worsening their depression.
As Clark emphasises, such general perspectives do not replace more targeted clinical interventions. Sometimes, indeed, the predictive-processing view can’t really help, as when he uses it to describe the likelihood of racist police shootings in the US, and concludes “what is most urgently needed is deep and abiding change in bedrock societal practices and institutions”. Readers might feel they didn’t need cognitive science to tell them that.
As might be expected of a philosopher, Clark also offers some ideas on how consciousness itself arises, suggesting that a kind of recursion – in which a system is constantly predicting things about its own predictions, as well as about the outside world – could be key. This will satisfy some readers more than others, and it’s notable that his descriptions of how our minds work tend to ascribe agency to the brain itself. But a brain doesn’t have agency; a person does. Meanwhile, if brains are really, as Clark argues, “ideally poised to discover epistemic actions” (those that will add to knowledge), why are so many people with functioning brains so bad at doing just that?
It’s difficult to believe, meanwhile, that prediction is, deep down, all brains or people do. If that were true, then ChatGPT – which, crudely, operates by predicting which word is more likely to follow the previous one, given its training data – would be a satisfactory model of human intelligence, which no one believes it is.
A perhaps mandatory final chapter turns to the self-help applications of all these ideas. Can we, as the author puts it, “hack” our predictive brains in desirable ways? One way to do it, he suggests, is to practise self-affirmation and to reframe negative experiences, which is already known to help since it is basically cognitive behavioural therapy. More originally, he discusses knowingly taking placebos (surprisingly effective) or downing psychedelics, which might reset stuck predictions. Gallivanting around in “psychedelic VR”, he adds, might even work the same way, which would at least be preferable to Mark Zuckerberg’s vision for the metaverse.
• The Experience Machine: How Our Minds Predict and Shape Reality is published by Allen Lane (£25). To support the Guardian and Observer order your copy at guardianbookshop.com. Delivery charges may apply.
