Who Are We Designing For?

A woman involved in a motor vehicle collision is significantly more likely than a man to be seriously injured, even when the crash is of comparable severity. That isn’t because women are poorer drivers or drive less safely. For decades, vehicle safety testing relied primarily on crash test dummies modelled on the average male body. Seatbelts, airbags, and other safety features were optimized around the body of an average male occupant¹. Only recently have regulators begun requiring crash testing using female occupant models.

Medicine provides a similar example. Women experiencing heart attacks are more likely to have their symptoms overlooked or misdiagnosed, in part because cardiovascular research and clinical teaching have historically treated men as the default patient, while women experience a broader range of symptoms that do not fit the “classic” presentation².

Or consider voice recognition software. For years, speech recognition systems performed better for men’s voices than women’s. Researchers have since confirmed measurable gender bias in automatic speech-recognition systems and traced part of the problem to the data and design choices on which those systems were built. The technology wasn’t intentionally designed to exclude women, but it was designed around an incomplete picture of the people expected to use it³.

These examples aren’t really about cars, medicine, or artificial intelligence. In each case, the people designing the system believed they were designing for the “typical” person. The problem was that the people they had in mind were less representative than they thought. None of these examples happened because someone asked, “How can we make this work less well for women?” They happened because the crucial question was not asked: Who are we designing this for?

Engineers wanted to build safer cars. Physicians wanted to reduce heart attacks. Software designers wanted systems that worked. None set out to disadvantage women, but they designed for the people they understood best, using the data they had, and asking the questions they thought to ask.

What these examples have in common is that certain assumptions became so embedded in the design of the system that they stopped looking like assumptions. The average male body came to represent the default body. The male presentation of a heart attack became the classic presentation. A male voice became the voice the technology was designed to understand.

The law is no different. Lawyers are trained to identify discrimination after it occurs. Governance asks us to go back a step further. When we draft legislation, advise boards, litigate what the reasonable person would have done, or design complaint processes, regulatory schemes, workplace policies, and contracts, we are making decisions about people we may never meet. Every legal rule reflects assumptions about how people will behave, what risks they will face, and what experiences are ordinary enough to be anticipated.

Assumptions are unavoidable in this process, but the challenge is recognizing that they are assumptions in the first place. Much of our legal system developed at a time when legislators, judges, and lawyers were overwhelmingly men, and when the lives and experiences that informed legal reasoning were correspondingly narrower. The law has evolved significantly since then, but legal concepts often outlive the circumstances in which they were created.

We still ask what a “reasonable person” would have done, how an “ordinary” employee would react, or what risks a “typical” consumer would foresee. Those standards strive for objectivity, but they do not exist in a vacuum. They are shaped by the perspectives we bring to them, the experiences we treat as representative, and sometimes the experiences we have not yet learned to see.

We often think of governance in terms of oversight, accountability, and compliance. Those are all important. But governance also means questioning the assumptions that shape our decisions before they become embedded in the systems we create. We cannot eliminate assumptions, but we can be more deliberate about testing them. Sometimes the most important governance question is not whether a system is working as intended, but whether it was designed with the right people in mind.

1. "The Crash Test Bias: How Male-Focused Testing Puts Female Drivers at Risk" | Consumer Reports | Oct 23, 2019

2. Changing the way we view women's heart attack symptoms | Amercian Heart Association | Mar 6, 2020

3. Towards inclusive automatic speech recognition | ScienceDirect | March 2024