The Evolution of Useful Things

In the AI Story

The book's significance in Petroski's body of work is structural. To Engineer Is Human (1985) had argued that failure drives engineering progress in high-stakes structures like bridges and buildings. The Pencil (1990) had demonstrated the same principle in a mundane artifact. The Evolution of Useful Things generalized the principle across the entire designed world, establishing that the form-follows-failure dynamic is not specific to dramatic engineering cases but operates continuously in the humble objects that surround daily life.

The book's method is historical and taxonomic. Petroski traced the development of specific artifacts through patent records, manufacturing archives, user complaints, and the physical evidence of surviving examples from different periods. The method enabled him to reconstruct the sequence of variations that produced the current form and to identify the specific failures that drove each modification. The reconstruction is often surprising: artifacts that appear to have always existed in their current form reveal themselves, under historical investigation, to have undergone dozens of significant changes, each responding to some inadequacy that users experienced and manufacturers corrected.

For the AI era, the book establishes a framework that makes visible what AI-generated design changes and what it cannot change. AI can produce variations at extraordinary speed — far faster than any historical evolutionary process. What AI cannot do, the book's argument implies, is perform the selection step that gives the evolutionary process its corrective power. Selection in Petroski's framework is performed by use — by the direct encounter between the artifact and the person whose needs it is supposed to serve. This encounter cannot be simulated, because the user's frustration, discomfort, or dissatisfaction is the signal that drives the next variation, and the signal is embodied and contextual in ways no dataset can fully capture.

The book's most frequently cited cases — the fork, the zipper, the Post-it note — have become staples of design education. But Petroski's deeper argument, often underweighted in popular reception, is that the evolutionary process produces not just better objects but better designers. Each participant in the process — the maker, the user, the manufacturer — develops, through direct encounter with failure, the calibrated sensitivity that constitutes design judgment. The objects are the visible deposits. The judgment is the invisible deposit. AI preserves access to the first. What happens to the second when the process that produced it is increasingly mediated by tools that bypass use-based selection is the question the book's framework poses to the current moment.

Origin

The Evolution of Useful Things was published in 1992 by Alfred A. Knopf. Its method extended the approach Petroski had established in The Pencil (1990) and applied it to the broader category of ordinary artifacts. The book drew on patent records, manufacturing histories, and the substantial literature on specific artifacts that Petroski had accumulated over years of research. Its reception was strong in design education but less widely noted in engineering circles, where Petroski's more dramatic failure-focused work (To Engineer Is Human, Design Paradigms) attracted more attention.

Key Ideas

Useful design is evolutionary. The forms of ordinary artifacts emerge through variation, selection, and retention — a process structurally similar to biological evolution, operating across distributed networks of designers and users.

Selection is performed by use. The evolutionary mechanism requires the direct encounter between artifact and user. This encounter generates the frustrations, discomforts, and failures that drive the next variation. The mechanism cannot be simulated because the signals are embodied and contextual.

The Post-it note is a case study in failure-as-opportunity. Spencer Silver's weak adhesive was, by the specification it was designed to meet, a failure. The transformation of this failure into a product required human recognition — Art Fry's hymnal bookmark frustration — that the failed property matched an unmet need. No optimization algorithm makes this recognition; it requires embodied human experience.

The AI era accelerates variation without accelerating selection. AI can generate more design variations in an hour than the historical evolutionary process produced in a decade. What AI cannot accelerate is the use-based testing that gives the evolutionary process its selective power. The result is an imbalance between variation-generation and variation-testing that the book's framework makes legible.

Debates & Critiques

The book has been criticized for underweighting the role of genuine creative breakthrough — the moments when designers do see further than the evolutionary process has yet reached. Defenders of the book's framework argue that such breakthroughs, on close examination, almost always turn out to be extensions of existing evolutionary trajectories rather than ruptures with them. The contemporary debate applies the book's framework to AI-generated design: whether AI can run evolutionary processes computationally, simulating millions of variations tested against virtual conditions, or whether simulated selection is categorically different from use-based selection in ways that make it unable to replicate the developmental work Petroski documented. The question remains open, but the book provides the analytical tools through which it can be posed precisely.