This article is incomplete, and likely significantly incorrect. But I’m posting it, anyway, just to spur me to keep working on it. It is flawed but there is important truth in it. If you are an engineer, please pay especially close attention to my reckless claims about regression testing, because I suspect I may be speaking about obsolete ideas that I never adequately understood. Let’s fight.

Once you understand that design is concerned with the development of hybrid systems comprising both voluntary participants (humans) and automatic components (nonhumans), and that such systems are incomplete until the human participants are actually participating several truths become obvious:

• The human participants must be taken as one of the essential parts of the system. Designers seek understanding of how human participants will behave within a system for exactly the same reason engineer seek understanding of how material or technological components will behave in an engineered system: To the degree a part of a system is poorly understood, it is likely to behave unpredictably and cause the system to perform poorly or fail.
• When we view systems that people are eventually expected to use or to participate in, as already complete without the people using it — when we try to evaluate that as-yet humanless system, without including the users or participants, we are mistaking a mere part for the whole. Almost all engineered systems are mere subsystems of larger designed supersystems, and the superstystem is the proper unit of evaluation.
• When we make changes to engineered systems, we are often making changes to the larger design system and how humans will participate in it. No competent engineer would dream of releasing a change to a system without performing regression testing, to ensure no unintended effects emerge from the system, but, because they misconceive design systems as mere engineered systems that will eventually be used by people, they fail to include usability testing in their protocols.

UXers and HCDers have always known it was foolish to speculate on how people will use or participate in a system. But few actually thought of people as voluntary participants in a supersystem that could be treated as a testable unit. Service design — our first human-nonhuman hybrid system development methodology — has made the advantages of this conception of design more obvious, but it was actually true from the start.