"We Have Always Done It This Way"

In the AI Story

The phrase was most famously delivered on 60 Minutes in 1983, when Morley Safer asked Hopper whether the computer revolution might be winding down. Her response combined the nanosecond wire with the phrase, arguing that the field had "barely built the Model T" and that the resistance Safer might sense came not from technology slowing but from institutions defending investments they had already made. Safer took the phrase seriously enough to use it as the interview's organizing principle.

The three-layer structure Hopper diagnosed — technical objection, cultural objection, philosophical objection — maps cleanly onto every subsequent widening of computing access. In 1952, machine-language programmers objected to compilers on all three grounds. In 2025, senior software engineers objected to AI coding tools on structurally identical grounds. The fact that the arguments recur verbatim, seventy years apart, is not evidence that the arguments are correct. It is evidence of the pattern Hopper was naming.

The phrase's diagnostic value is precisely that it does not require judgment about the truth of specific technical claims. The senior developer may be right that AI-generated code is sometimes less efficient, less elegant, or less maintainable than hand-written code. The claim can be granted without affecting the deeper question, which is whether the inefficiency justifies maintaining the barrier against the population of potential builders the barrier excludes. Hopper's phrase forces the conversation to that deeper question.

The phrase's power lies in its refusal of the retreat into tradition. It treats "we have always done it this way" as an argument that has failed to meet the burden of proof — the burden of demonstrating that the old way is better than every possible new way. When that burden cannot be met, the phrase instructs, the correct response is to step aside, because the door is opening whether the defenders approve or not.

Origin

Hopper used versions of the phrase throughout her lecturing career, particularly from the late 1960s onward. The specific formulation "the most dangerous phrase in any language" became her signature line, repeated in Congressional testimony, university lectures, and television appearances.

She framed it explicitly as diagnostic: a way of identifying institutional failure modes rather than a complaint about particular people or organizations.

Key Ideas

Resistance as expertise pathology. The people most likely to resist a new tool are the ones whose expertise in the old tool is most developed, because the new tool threatens what their expertise credentialed.

Three-layer structure of resistance. Technical objections, cultural objections, and philosophical objections appear together at every widening of access, and each is partially valid but jointly insufficient.

Burden of proof. "We have always done it this way" is not an argument; it is the absence of an argument, a confession that the speaker cannot envision a future different from the present.

Pattern recognition across decades. The phrase's diagnostic value compounds over time — recognizing the pattern in one transition accelerates recognition of it in subsequent transitions.

The door is opening regardless. The phrase's sharpness lies in its recognition that institutional resistance does not determine outcomes; it only determines whose reputation survives the transition.

Debates & Critiques

A line of critique runs that Hopper's phrase is itself an oversimplification — that sometimes "we have always done it this way" encodes real accumulated wisdom, and that the dismissal of tradition can be a form of its own triumphalism. Chesterton's fence applies: before tearing down the fence, you should know why it was put up. The defense of Hopper's phrase is that she did not claim tradition was always wrong. She claimed that tradition could not be sustained on the grounds of tradition alone — that the burden of proof rested on those defending the current practice to demonstrate its superiority, not on those proposing change to justify their proposal. The distinction matters for the AI debate: the senior developer is not obligated to disprove that AI tools might be worse in some respects, but she is obligated to demonstrate that the ways they might be worse are sufficient to justify the exclusion of the population the tools would include.