The Economics of Attention

A. In 1971, the economist Herbert Simon made an observation that seemed counterintuitive at the time but has since become almost a truism. A wealth of information, he wrote, creates a poverty of attention. Simon was writing decades before the internet, yet he had identified the mechanism that would come to define the digital economy. When information is scarce, its value is high and the cost of consuming it is low. When information becomes abundant, the calculation inverts: the scarce resource is no longer the information itself but the finite human capacity to attend to any of it.

B. The commercial implications took some time to become apparent. Early commentators on the internet assumed that abundance would democratise knowledge, and in certain respects it has. But abundance also created a market in which the primary commodity being traded was not content but the attention that content could capture. Advertising-funded platforms do not sell products to their users; they sell their users' attention to advertisers. Once this is understood, a great deal of otherwise puzzling design becomes legible. Features that appear to serve users may in fact serve the extraction of attention, and the two objectives coincide only intermittently.

C. Consider the phenomenon of infinite scroll, introduced in 2006. Before its invention, reaching the bottom of a page required a decision: click for more, or stop. That decision point was a natural boundary, and boundaries permit reflection. Infinite scroll removes the boundary entirely, and with it the moment at which a user might reasonably conclude that they have finished. The engineer who designed the feature has since expressed regret, estimating that it costs humanity a substantial number of collective lifetimes each day. Whether or not that figure withstands scrutiny, the underlying point is difficult to dispute: small alterations to the architecture of choice produce large aggregate effects.

D. Critics of the attention economy often frame the problem in terms of addiction, and the vocabulary of addiction has proved rhetorically effective. It is, however, imprecise. Addiction in its clinical sense involves tolerance, withdrawal and continued use despite substantial harm, and while some heavy users of social platforms exhibit patterns resembling these criteria, most do not. A more accurate framing may be that these systems exploit ordinary features of human cognition that operated perfectly well in the environments where they evolved. Variable reward schedules, social monitoring and a bias towards novelty are not pathologies. They become problematic only when engineered environments amplify them beyond any function they once served.

E. This distinction matters because it bears on what solutions are plausible. If the problem is addiction, then the appropriate responses are clinical and individual: treatment, abstinence, willpower. If the problem is instead a mismatch between evolved cognition and engineered environments, then individual remedies are unlikely to suffice, much as individual dietary discipline has proved an inadequate response to the industrial production of food engineered for overconsumption. The historical parallel is instructive. Regulation of tobacco advertising achieved more than decades of exhortation to smokers, not because smokers lacked willpower but because the environment had been deliberately constructed to defeat it.

F. Some economists have proposed that attention be treated as a form of property, with users compensated for the time they surrender. The proposal is elegant in theory and troublesome in practice. Attention is not straightforwardly alienable: I cannot transfer three hours of my attention to another person in the way I might transfer three hours of my labour. Nor is it clear that monetisation would improve matters. A market in which users are paid to attend might simply intensify the competition for their attention rather than relieving it, in the same way that paying for blood donation has, in some jurisdictions, reduced rather than increased overall supply by displacing the altruistic motive.

G. What seems more promising, if less dramatic, is a shift in the metrics that platforms optimise for. Time spent is a poor proxy for value delivered, yet it remains the dominant measure because it is easy to count and correlates with advertising revenue. Some firms have experimented with alternative measures, asking users afterwards whether the time was well spent, and adjusting recommendations accordingly. Early results suggest that engagement falls modestly while reported satisfaction rises. Whether shareholders will tolerate that trade is a separate question, and probably the decisive one.

H. None of this is to suggest that the outcome is predetermined. Attention has been commercially valuable for as long as there have been newspapers, and the printing press provoked anxieties about distraction and shallow reading that read almost identically to contemporary complaints. What is different now is the granularity of feedback available to those designing the systems. A newspaper editor of the nineteenth century could observe circulation figures monthly; a platform today observes the behaviour of individual users continuously, and can adjust in response within hours. That asymmetry between the sophistication of the system and the awareness of the person using it is the genuinely novel element, and it is the element that any serious response will have to address.