Induced Demand (Cognitive Labor)
Induced demand describes the phenomenon—well-documented in transportation economics—by which increasing the supply of a resource (highway lanes) generates new consumption (more driving) that fills the added capacity, leaving congestion unchanged or worse. The mechanism operates because latent demand exists but is constrained by cost; remove the constraint and demand expands. Heilbroner's simulation applies this to cognitive labor: AI tools reduce the time-cost of knowledge work, enabling workers to complete tasks in forty minutes that previously required four hours. The naive expectation is that freed time becomes leisure. The actual outcome, documented in the Berkeley study, is that freed time fills with additional tasks—either because the worker volunteers (the internal imperative to achieve) or because managers observe the efficiency gain and adjust expectations upward. The productivity gain does not reduce work; it induces new work, and the induction operates faster than the worker's capacity to establish boundaries, producing the intensification rather than relief that optimistic predictions assumed AI would deliver.
In the AI Story
The concept originates in urban planning and transportation economics—most famously in Anthony Downs's 'law of peak-hour traffic congestion' (1962) and subsequent empirical studies showing highway expansion fails to reduce congestion because new lanes induce new trips. The mechanism requires three conditions: first, latent demand exceeding current capacity; second, a binding constraint (cost, time, or inconvenience) suppressing that demand; third, the removal of the constraint, which releases demand that fills the new capacity. All three conditions obtain for knowledge work in the AI age. Latent demand for cognitive output has always exceeded supply—there were always more emails to write, more analyses to run, more features to ship, more research to conduct. The constraint was the time and mental effort required. AI reduces both, and the latent demand rushes in.
Heilbroner's framework reveals induced demand as an instance of the broader dynamic he called, in 21st Century Capitalism, the 'treadmill'—the mechanism by which every productivity gain is absorbed into higher expectations rather than converted into reduced effort. The treadmill operates through comparative advantage competition: when everyone's productivity rises together, no one gains positional advantage, but everyone must work at the new higher level to maintain their relative position. The developer who produces twice as much with AI tools does not earn twice the wage or work half the hours; she maintains her position in a labor market where everyone else is also producing twice as much. The productivity gain accrues to consumers (who get more software), to platform owners (who capture subscription and API revenue), and to employers (who get more output per employee)—but not, systematically, to the worker in the form of reduced hours or increased autonomy.
The policy implication is that productivity gains do not automatically translate into welfare improvements for workers, and that converting gains into welfare requires deliberate institutional intervention. The eight-hour day was such an intervention—a collective decision to refuse the induced demand for more labor, to establish a maximum regardless of what the market would bear. The contemporary equivalent would be institutional structures preventing AI-enabled productivity gains from being automatically channeled into intensified expectations: protected time for judgment development, maximum response latency requirements for asynchronous work, seasonal rhythms building in recovery periods, and compensation structures decoupling pay from hours and linking it to output quality rather than quantity. These structures do not exist at scale, and their absence is why the Berkeley researchers observed intensification rather than relief. The induced demand will continue operating until the institutions constraining it are built.
Origin
The induced demand concept in transportation is formalized in Anthony Downs, 'The Law of Peak-Hour Traffic Congestion,' Traffic Quarterly 16:3 (1962), and empirically demonstrated in hundreds of subsequent studies. The extension to cognitive labor is implicit in Juliet Schor's work-hours research and explicit in the Berkeley AI ethnography. Heilbroner did not use the term but described the underlying mechanism across his work: that productivity gains in capitalist economies are channeled into growth rather than leisure because the institutional and cultural infrastructure directs them there.
Key Ideas
Constraint removal releases latent demand. Demand for cognitive output was never fixed but suppressed by cost; AI reduces cost and demand expands to fill every freed hour, preventing the conversion of productivity into leisure.
Comparative treadmill. When everyone's productivity rises together, relative positions remain unchanged and everyone must work at the new higher level to maintain standing—the gains accrue to consumers and infrastructure owners, not to workers as reduced hours.
Institutional intervention is required. Converting productivity gains into welfare improvements for workers does not happen automatically; it requires deliberate structures (maximum hours, protected rest, compensation decoupled from time) that the AI age has not yet built.
The pattern is Jevonsian. Efficiency in the use of a resource increases total consumption of that resource when demand is elastic and latent—a principle operating across coal, highways, and now cognitive labor with identical mechanism and identical outcome.
Appears in the Orange Pill Cycle
Further reading
- Anthony Downs, 'The Law of Peak-Hour Traffic Congestion,' Traffic Quarterly 16:3 (1962)
- Juliet Schor, The Overworked American (Basic Books, 1991)
- Leslie Perlow, Sleeping with Your Smartphone (Harvard Business Review Press, 2012)
- Xingqi Maggie Ye and Aruna Ranganathan, 'AI Doesn't Reduce Work—It Intensifies It' (HBR, Feb 2026)