The Dreams of Reason: The Computer and the Rise of the Sciences of Complexity
Source: https://archive.org/details/dreamsofreasonco0000page ↗
Pagels, a theoretical physicist, wrote this book just before his death in a mountaineering accident, and it stands as one of the earliest and most lucid accounts of the transition from reductionist physics to the sciences of complexity.
He traces the lineage from Shannon, Wiener, and von Neumann through to the cellular automata of Wolfram, the genetic algorithms of Holland, and the self-organisation models of Kauffman, arguing that the computer was not merely a tool but a new way of thinking about natural systems.
The book appeared four years before Waldrop's Complexity and covers much of the same intellectual territory from a physicist's perspective rather than a journalist's.
Pagels is unusually clear about what complexity science can and cannot explain, and his writing carries the authority of someone who understood both the mathematics and the philosophical stakes.
It remains an underappreciated bridge between the information theory era and the complexity era.
Central argument
Pagels argues that the computer transformed science not by accelerating calculation but by introducing a new epistemological mode: complex systems — from immune responses to economies — can only be understood through simulation and iteration, not closed-form equations. Drawing a lineage from Shannon's information theory through von Neumann's automata to Wolfram's cellular automata and Kauffman's self-organisation models, he contends that reductionism reached its explanatory ceiling precisely where interesting natural phenomena begin, and that the sciences of complexity represent a genuine paradigm shift rather than an extension of classical physics. The computer is the instrument that made this shift possible, because it allows scientists to construct and observe emergent behaviour that cannot be derived analytically.
Critique
Pagels writes with the confidence of a physicist who believes the new science of complexity will eventually yield rigorous, generalisable laws, but nearly four decades later that promise remains largely unfulfilled — complexity science has produced powerful metaphors and simulation tools without converging on predictive theory comparable to classical physics. A thoughtful reader might argue that Pagels underestimates the risk of what later critics called 'complexification': the tendency to redescribe hard problems in the language of emergence and self-organisation without actually explaining them. His account of what complexity science cannot explain, while unusually honest for its era, does not fully reckon with the possibility that the paradigm shift he celebrates might be more methodological than explanatory.
Why it matters for product
For a CPO, the deepest practical import of Pagels's argument is that product systems — user behaviour, adoption curves, team dynamics, market fit — are complexity-class phenomena, meaning they will systematically resist the reductionist instrumentation that dominates product analytics: funnels, attributed conversion, A/B tests that isolate single variables. The implication is not to abandon measurement but to design for observation of emergent patterns, which in practice favours longitudinal behavioural cohorts, qualitative system mapping, and organisational structures loose enough to evolve rather than execute against a fixed plan. Pagels's insistence that the computer is a thinking tool rather than merely a calculating one also reframes what product teams should actually be building with their data infrastructure — not dashboards that confirm hypotheses, but environments that surface unexpected system behaviour.