Bedeviled: A Shadow History of Demons in Science
Source: https://press.princeton.edu/books/hardcover/9780691175324/bedeviled ↗
Canales writes the history of imaginary beings in science — Maxwell's demon sorting molecules, Laplace's demon predicting the universe, Descartes's evil genius deceiving the thinker — and shows that these thought experiments were never mere illustrations but active tools that shaped the development of thermodynamics, information theory, quantum mechanics, and artificial intelligence.
Each demon was introduced to test the limits of a theory and ended up revealing something the theory could not contain.
The book connects figures from Leibniz through Boltzmann and Szilard to Landauer and Bennett, tracing how the demon concept migrated from physics to computation as the relationship between information and entropy became clear.
Canales's approach — taking metaphorical constructs seriously as drivers of scientific progress — is itself a demonstration of how conceptual tools do real intellectual work.
The result is an unusual history of physics told through its own fictional characters.
Central argument
Canales argues that the fictional demons of scientific thought experiments — Maxwell's, Laplace's, Descartes's — were not decorative illustrations but generative conceptual instruments that actively shaped the development of thermodynamics, information theory, and computation. Each demon was constructed to probe the boundary of an existing theory and invariably exposed something that theory could not account for, forcing its revision or replacement. The book's central finding is that the lineage from Maxwell through Szilard to Landauer and Bennett reveals a deep structural link between information and entropy — a connection that the demon concept, precisely because it was fictional and therefore unconstrained, was uniquely positioned to make visible.
Critique
Canales's method — treating metaphorical constructs as drivers of scientific progress — risks overclaiming the causal role of conceptual figures relative to the mathematical, experimental, and institutional pressures that also shaped these fields. A thoughtful reader might ask whether the demons were genuinely generative or whether they became legible as such only in retrospect, once the theories they supposedly drove had already been developed by other means. The history may be more a coherent narrative imposed on a messier record of parallel developments than a strict account of conceptual causation.
Why it matters for product
Product leaders routinely dismiss thought experiments and hypothetical framings — the imaginary user, the edge case scenario, the 'what if we had no database' constraint — as soft or preparatory thinking, but Canales's argument suggests these fictional constructs do real structural work by exposing the limits of current models before those limits become costly failures in production. More concretely, the demon framework offers a direct analogy for how information and friction interact in product systems: every workflow, funnel, or data pipeline has an implicit Maxwellian assumption about what sorting is costless, and surfacing that assumption is a strategic act. A CPO who takes this seriously will invest in the deliberate construction of conceptual pressure-tests — not just user research or A/B tests — as a distinct mode of product discovery.