Introduction: Face It or Replace It?: Why Computational Metaphors Fall Short and Why We Need a New Approach
Louise Barrett
We are living creatures, so comparisons to machines and computational devices will always fall short as metaphors for how we make sense of the worlds in which we live.
When Shakespeare wrote that Juliet was the sun, he didn’t mean she was a giant ball of flaming gas, or that she lived eight million miles away from earth. He meant she was the center of Romeo’s world, the life-giving force without which he could not survive. That’s how it goes with metaphors. We don’t apply them literally; we use them to capture a certain quality of the object in question, a way of seeing through one thing to gain a fresh appreciation of another. Similarly, in science, we use metaphors to help us think about the things we don’t understand in terms of the things we do. Without the metaphor, we would have no name to describe the abstract concepts and complex processes we have to grapple with. Using a metaphor turns the abstract into something concrete and provides us with a guide to discovery.
I raise this issue because we seem to have lost control of our metaphors when thinking about brains and cognition. Descriptions of brains as computers, and neural activity and associated psychological processes as information processing, are now taken quite literally. For example, in an op-ed for the New York Times, published a few years ago, Gary Marcus bluntly told us all to “face it, your brain is a computer,” strongly implying that using metaphors is not a means to kick-start new ways of thinking, but a way to identify the “right” way of thinking about things. For Marcus, Descartes’ idea of the brain as a hydraulic pump, for example, was just one in a long list of failed metaphors that could never give us the right answer. But Descartes was not wrong as such, he was simply using the best metaphor he had at his disposal to make familiar what was otherwise deeply unfamiliar and difficult to grasp. Obviously, times change, knowledge grows, we think differently, and we adopt new and better metaphors. Descartes was a man of his time, and if we have a better metaphor than he did, it’s because we have gained more knowledge of the brain. Marcus goes further than this, however, suggesting that the computational brain is not simply a more useful metaphor than a hydraulic pump, but precisely the right comparison: Brains are computers, and those who object to this idea are, quite simply, wrong. All we have to do, he says, is figure out what kind of computers they are.
But here’s the thing: The objections that Marcus raises and then dismisses against the computer metaphor - that computers are serial, stored-program machines whereas brains are parallel; that computers are digital and brains are analog; that computers cannot generate emotions like humans - is like assuming Shakespeare was, in fact, referring to Juliet’s resemblance to a giant ball of gas. Such objections make his argument sound more plausible but fail to capture the metaphor’s essence. The heart of the original computer metaphor is the idea that brains construct an internal model of our environment, and that this internal representation is what allows us to act efficiently and effectively. Like computers, brains take certain inputs, manipulate these in various ways to generate our perception of the world, and then compute a set of outputs that instruct our bodies how to act. This is the job of the brain, so the argument goes, because the inputs our senses receive are too impoverished to allow us to cope with the world around us. The flat, upside-down image on our retina, for example, has to be converted into our dynamic three-dimensional view of the world. Our contact with reality is therefore indirect, via the representational model our brain builds, and not with the world itself. This view, as it turns out, was more or less what Descartes suggested as well, so the gap between Descartes’s and the computer metaphor is not so great as Marcus makes it seem.
The idea that brains are biological computers and that cognition is a process of computation is a legacy of the original artificial intelligence project of the 1950s. Its guiding assumption was that intelligence could be modeled by computers, and it focused on precisely the kinds of representation-heavy abilities, like language skills and mathematics, that were thought to epitomize intelligent behavior. This approach failed to deliver on its promise. As Marcus admits, we are no closer to understanding the brain in computational terms than we were 60 years ago. So perhaps we should change our metaphors, and so change the job description of the brain.
Instead of thinking of brains as representational, we can view them as “performative”: Their job is not to model the world around us, but to guide and control the actions of our bodies in an inherently dynamic, unpredictable world. Rodney Brooks, the MIT roboticist, makes the argument this way: Four billion years of evolution were, in the main, spent refining the perception-action mechanisms that guide effective action in the world. It took an enormously long time to build insect-level intelligence, while those things we think of as highly intelligent human capacities - language, logic, mathematics, chess - evolved very rapidly in less than a blink of evolution’s eye. These distinctively human, representation-heavy capacities must therefore have been pretty easy to implement once the former was in place, Brooks argues. This means we would do better trying to understand how whole animals cope with changeable environments, rather than assuming that investigating the computing power of brains in isolation will allow us to understand how flexible, adaptive behavior is produced.
Brooks was the pioneer of an alternative behavior-based robotics, which demonstrated convincingly that flexible, adaptive, intelligent behavior does not require a representation-heavy computational model to achieve. He built robots with bodies that could sense and act in ways that allowed the robot to “use the world as its own best model” and didn’t worry about providing them with any kind of central processing unit or “brain.” After all, why go to the expense of representing the world, when the world itself contains all the information needed? As inventor of the Roomba, the vacuum cleaner that uses precisely these principles to clean your house all by itself, the strength of Brooks’ approach is clear: Ten million Roomba owners can’t be wrong.
Rather than viewing the brain as the only factor responsible for adaptive behavior, this “embodied” view of cognition regards the brain as part of a dynamic system in which brains, bodies, and environment all come together to generate intelligent action. That is, cognitive processes are grounded in action, and are not purely brainbased phenomena, but also exploit both the body and environment so that problems can be solved more cheaply than via energetically expensive brain tissue. If you’ve ever wondered why a seahorse looks so boxy and square, for example, it is because a tail with a square cross-section enables firmer anchoring to marine plants than does a circular one, which reduces the need for fine motor control. The importance of bodies is also apparent among animals that are well endowed with brain tissue, such as New Caledonian crows. These birds have the ability to both make and use tools, but it is not apparent that their brains are any larger or more complex than those of the other members of the crow family. What New Caledonian crows do possess, however, are very straight bills and forward-facing eyes that improve their ability to grasp and visually guide the stick-like tools they use to acquire food, giving them the edge over other crow species.
Marcus, and other proponents of the standard cognitivist view, give the impression that the representational-computational brain is the only game in town. Although we humans can and do deal with representations (after all, by reading the words on this page, this is precisely what you are doing now), we need to entertain the possibility that this is a nifty trick that humans developed over the course of our own particular history, and not something that all brains everywhere have evolved to do. It may also be useful to continue using computational-representational models in certain areas of neuroscience and psychology, while recognizing that such models are, in all probability, wrong about the way brains actually work. Brains are not magical, as Marcus rightly says, but that doesn’t mean they must obey the laws of computation. Work in artificial life, behavior-based robotics and embodied cognition suggests we are reaching the limits of the computer metaphor, and it may be time to try something new.
Perhaps another useful metaphor, as Daniel Nicholson suggests, is to think of organisms as more like candle flames than machines. A flame has an enduring and stable form, but this comes about as a result of a continually ongoing set of dynamic processes. This metaphor highlights a crucial difference: Machines can only take part in processes, but organisms are processes. You can turn a machine off, and it will still be there, but not so an organism (well, it will for a bit, but we call that being dead). In this alternative view, organisms are thus seen as leaky processes spreading out across brain, body, and world. Similarly, we may be able to advance our understanding of ourselves and other organisms - and hopefully improve all planetary life - by recognizing the need for a similarly leaky process of knowledge production, one that regards disciplinary boundaries as entirely porous and focuses on examining common areas of interest from a variety of intersecting perspectives. The essays in this book make for an excellent start.
For Sources and Further Reading
Barrett, L. (2015). Beyond the brain: How body and environment shape animal and human minds. Princeton, NJ: Princeton University Press.
Barrett, L. (2016). Why brains are not computers, why behaviorism is not satanism, and why dolphins are not aquatic apes. The Behavior Analyst, 59(1), 9-23.
Brooks, R. (2002). Robot: The future of flesh and machines. London: Penguin Books.
Leary, D. E. (Ed.). (1994). Metaphors in the history of psychology. Cambridge, UK: Cambridge University Press.
Marcus, G. (2015). Face it, your brain is a computer. New York Times, June 27, 2015. https://www.nytimes.com/2015/06/28/opinion/sunday/face-it-your-brain-is-a-computer.html
Nicholson, D. J. (2018). Reconceptualizing the organism: From complex machine to flowing stream. In Nicholson, D. J., & Dupre, J. (Eds.), Everything flows: Towards a processual philosophy of biology (pp. 139-166). New York: Oxford University Press.
