The Unbuilt Bridge

In the AI Story

The unbuilt bridge is Petroski's positive concept — his category for engineering wisdom operating as intended. It is easy to document failures: bridges that fell, buildings that collapsed, systems that catastrophically malfunctioned. It is harder to document the bridges that should not have been built and were not — the ambitions that were deferred because the profession recognized, honestly, that its understanding was insufficient. These non-events leave no ruins, no commission reports, no memorials. They exist only in the proposals that were shelved, the designs that were modified, the projects that waited for the understanding to catch up before proceeding.

The judgment the unbuilt bridge embodies — the capacity to distinguish between what the formulas permit and what experience warrants — is what Petroski considered the most valuable and most endangered form of engineering intelligence. It is the engineer saying: this is possible in principle and irresponsible in practice, because the principles have not been validated at this scale, and the consequences of discovering their limits during construction or operation would be measured in human lives.

The judgment is fallible. It can be excessively conservative, leading engineers to avoid designs that would have been safe. It can be miscalibrated, firing in response to superficial similarities between a current design and a past failure that are not structurally relevant. But its average effect, across the history of the profession, has been protective. The bridges that were not built because engineers felt uneasy about them include, almost certainly, some that would have stood. They also include some that would have fallen. The unbuilt bridge is the profession's way of accepting the cost of excessive caution — some forgone capability — in exchange for the benefit of avoiding catastrophes whose cost is measured in lives.

AI changes this calculus because it shifts the default. In the pre-AI era, designing a bridge was expensive, time-consuming, and required significant institutional commitment before the first calculation. The cost of proposing a design was high enough that proposals were self-selecting: only designs the engineer believed, based on experience and judgment, to be both feasible and safe were proposed. The friction of the design process acted as a filter. In the AI era, generating a design is cheap and fast. The cost of proposal has dropped dramatically. The filtering function of the pre-AI design process has been weakened. Designs that would not have survived the old filter — because the engineer would have hesitated, felt the boundary between validated and speculative, decided the ambition exceeded the understanding — now arrive on desks as completed outputs, formatted and detailed and apparently ready for review. The reviewer must supply the filter that the process no longer provides.

Origin

The concept of the unbuilt bridge as a positive category — as engineering wisdom rather than engineering failure — was developed across Petroski's career, most explicitly in Engineers of Dreams (1995), which examined the work of great American bridge builders including Lindenthal. Petroski's treatment inverted the typical historical narrative, which celebrates built structures and treats unbuilt proposals as curiosities. For Petroski, the unbuilt proposals were often the clearest evidence of mature engineering judgment — the recognition, by the engineers closest to the work, that a design whose building would be possible would nonetheless be wrong.

Key Ideas

Unprecedented is a synonym for untested. The fact that a configuration can be calculated does not mean it can be built safely. The calculation operates within a theoretical framework; the safety depends on whether the framework's assumptions have been validated at the relevant scale.

The unbuilt bridge is wisdom, not failure. It represents the profession's judgment that the gap between capability and safety exceeds what the factor of safety can bridge. This judgment is the most valuable form of engineering intelligence because it operates precisely at the boundary between the possible and the catastrophic.

AI removes the friction that produced the filter. When designing was expensive, only designs the engineer believed to be safe were proposed. The cost of proposal acted as a self-selection mechanism. AI collapses this cost, which means every design that satisfies specifications now arrives for review, regardless of whether the engineer would have hesitated to propose it in the pre-AI era.

The reviewer must supply the hesitation the process once provided. In the AI era, the judgment about whether a design should be built — as distinct from whether it can be — falls to the reviewer, who must evaluate designs she did not generate. This form of review is structurally harder than review of a design she constructed herself, because the assumptions are invisible when you have not encountered them through construction.

Debates & Critiques

Proponents of AI-augmented design argue that the ability to rapidly explore parameter spaces no human could examine manually enables the profession to validate previously unprecedented designs through exhaustive simulation — that unprecedented and untested are no longer synonymous when AI can simulate millions of conditions. The argument has merit within the space of phenomena the simulation models. The Petroski objection is that the simulations operate within existing physical theory, and the failures that matter most involve physics the theory does not yet include. No amount of simulation can validate against conditions the simulation does not know to model. The unbuilt bridge category exists precisely for designs where the relevant physics has not yet been experimentally validated at the required scale — and AI, however powerful, cannot bridge this gap because the gap is not computational but empirical.