In the Cave: How Automation Changes the Way We Interact with Our World

Nicholas Carr has titled his latest book, which investigates the dangers of widespread automation, The Glass Cage. The title alludes to a shift in the display of information in an airplane’s cockpit from an older analog model to an electronic LCD screen. Among other things, the turn to electronic display rendered the plane’s flight engineer obsolete. But with increased automation in the cockpit, even the pilot, who spends an average of only three minutes per flight actually working the controls, now seems redundant. The glass cockpit has become a glass cage.

Readers of Carr’s interesting and disturbing book will perhaps come away with a different image, also suggested by his title: the image of Plato’s Cave. As in the cave, where image substitutes for reality and connection to the real world is lost, so in Carr’s rendering does ubiquitous automating technology threaten to leave us in a mere “shadow of the world.” The Platonic overtones are unmistakable, though Carr does not explicitly acknowledge them.

Carr returns often to the condition of modern pilots. The degree to which their work has been automated is considerable, and both the benefits and the costs of that automation are apparent. Air travel is vastly safer now than even sixty years ago—far safer, in fact, than driving. Automation also cuts costs: in those same sixty years, the professional staff of a typical airliner has shrunk from five, with a navigator, a radio operator, and a flight engineer, to two, with only the pilot and co-pilot remaining. Not a benefit for the professionals, but surely a less costly way to fly.

On the other hand, some dangers must be chalked up to automation, as disuse of manual skills causes those skills to atrophy. In an emergency, pilots can show diminished quality of response, occasionally resulting in tragic and lethal pilot error.

The occasional and preventable accidents that result might seem like a small price to pay for a safety record that, on balance, is more than impressive. But Carr is after something bigger than a look at the safety risks that automation may pose when automation complacency and bias lead professionals (such as doctors) to overlook or ignore essential data. For Carr, more central than such risks is the way that technology negatively affects our agency, our cognition, and our connection to the world.

Automation and Agency

Let us start with the question of agency. What impact does automation—the replacement of human activity by computer-guided technology—have? One consequence is a decrease in skill. Like the pilots who do not need to exercise any manual flying skills, other professionals also suffer erosion of talents that were once deemed essential. So, for example, architects once prized the ability to sketch; now, computer-aided design (CAD) allows architects immediately to move to design without first working through a series of sketches of increasing adequacy. Indeed, CAD software cleans up errors and inadequacies in first drafts, moving the architect even further away from a need for what was once considered forms of basic competency.

Agency is further affected by distancing agents from the world in which they work. For doctors, this can mean paying attention to a tablet screen rather than a patient. For a pilot, it can mean no longer having the tangible mechanical connection to the various parts of the aircraft he controls. This distancing is tied to the diminution of skills: Part of any difficult skill is recognizing what is salient in the world to its exercise. But distance from the world erodes that recognitional capacity. Just ask the Inuit, whose reliance on GPS screens has corrupted a centuries-old ability to navigate seemingly marker-less landscapes of snow and ice with uncanny accuracy.

Agency is eroded when we rely on software designed by others because that software implicitly realizes the presuppositions of its authors: They are the creators who tell the machine what to look for and how to respond, removing from the lives of professionals such as doctors or lawyers the need—and therefore the ability—to judge for themselves what is important in any given situation.

Automation and Autonomy

Carr frequently speaks of the way that technology diminishes autonomy. He should not be misunderstood to be speaking only of autonomy in the modern liberal sense: doing what one wants, unencumbered by obligation or connection to others. On the contrary, when a doctor’s professional autonomy and judgment are replaced by reliance on a mediating computer screen, the doctor’s real human connection to her patient and her real exercise of human agency is diminished.

Proponents of automation frequently promote what Carr calls the “substitution myth.” This term refers to the idea that automation merely takes the place of some discrete activity that imperfect humans used to do. Automation, on this view, leaves everything as it was before, save that it lightens the workload for some overburdened person. But this is false.

Rather, automation brings with it a change in how we think. Automation complacency and bias are important here. Our reliance on automation leads to errors of attention and inference. But the threat is even more worrisome.

Is Automation Making Us Stupid?

Consider the role, discussed at length by Carr, that effort and friction play in learning and remembering. Subjects asked to memorize pairs of antonyms do better if they are initially asked to fill in missing letters of one of the words—for example, instead of being given “hot” and “cold,” they are given “hot” and “c—.” Subjects asked to recall immediately what they have been shown do better at later recall than subjects who merely receive the information over a series of sessions and are only later asked to recall it.

This phenomenon also occurs in the acquisition of tacit knowledge: the struggle to ride a bike, drive a stick shift, play on pitch, learn a foreign language, or cook a risotto takes place through time; it involves inevitable failure and effort, but such a struggle results, when successful, in a deep internalization of “tacit” knowledge. This knowledge imparts the ability to draw on what is known without even bringing that knowledge to the level of full and articulated consciousness.

But if, as Carr argues, automation turns us largely into observers, then its effects may be expected to be deleterious where the development of tacit knowledge and memory are concerned. And this suggests, as do many other points made by Carr, that automation is making us stupid (Carr himself suggested this in a 2008 essay in The Atlantic, “Is Google Making Us Stupid?”).

Carr pursues these points via a discussion of “embodied cognition”: the idea that thinking, for us embodied beings, requires doing. That is, thinking requires an extension of mind out into the physicality of world in a bodied engagement with that world. Think of the way an infant learns: by touching, tasting, grasping, breaking. Such engagement with the world should not stop at infancy. Nor, we might add, should it be cut off in infancy; although not discussed in this book, readers should leave worried if they have been allowing their young children unfettered access to tablets, smartphones, or computers.

What Happens When Work Doesn’t Connect Us to the Physical World?

What are the long-term consequences of these failures of agency and cognition? I shall focus on just one.

Work, like play, is a basic aspect of human flourishing. We are inclined to deny this in the face of work that is tedious, monotonous, unskilled, and boring, work to which it is impossible to be attentive and mindful. But at its best, work allows us engagement with the materials of the world in ways that utilize long-cultivated skill and hold our attention, giving us a sense of flow, which Carr discusses, following psychologist Mihaly Csikszentmihalyi. Good work surely connects us more deeply to the world around us, and gives us opportunity to impress our character on that world in lasting ways.

But real work requires agency and cognition. It is impossible without autonomous engagement with the things of the world, without planning, without awareness of how the world is and of how it offers or threatens to push back against our intention. It is impossible without the developed skills that enable us to overcome that pushback, even transform it according to our will so as personally to shape the product that is the proximate end of our work, whether that product be a safe flight, a medical diagnosis and cure, or a new building.

What is left when work has been evacuated of agency and cognition? We return to Plato, and his image of human beings chained by the neck and able only passively to watch the shadows, secure from the sun. Carr’s honest engagement with the perils of automation should make us eager for the climb back out of the cave. Carr offers only a few suggestions as to how that might be done, directing us, for example, toward attempts to create more human-centered automation, in which persons are engaged rather than passive.

The topic clearly requires more attention. But Carr deserves recognition for articulating the need for that attention in a smart and focused, if ultimately rather ominous, way.