Sometimes We Need Something Between Us

In her study of a manufacturing company, Beth Bechky describes an engineer asking an assembler why there were scratches inside a machine's chamber. The assembler explained that a screw caught when a plate was lifted. He repeated the explanation and gestured, but the engineer still did not understand. Eventually, he fetched the plate and showed what happened. With the part in front of them, the engineer could connect the explanation to the assembly process. 1

The difficulty came from their different experiences of the same machine. The engineer understood its design; the assembler knew what happened while putting it together. The physical part gave them a way to connect those perspectives.

This helps explain why a sketch, model, or prototype can be useful in a conversation. General descriptions leave room for people to imagine different things. Drawing a sequence requires decisions about what happens first and what follows. Building a model gives shape to relationships that might otherwise remain vague. Other people can then inspect those decisions and explain where their understanding differs.

Cognitive scientist David Kirsh describes how external representations give people a stable object of thought. A drawing stays available while we examine a detail, change something, or return to an earlier question. Relationships that would otherwise have to be remembered and imagined can be inspected and rearranged. 2

However, the representation must make sense to the people using it. In Bechky's study, engineering drawings were often too abstract and unfamiliar to help assemblers connect them with their work. The machines and their parts provided a more useful common reference. A drawing that is clear to its author may still be difficult for someone else to use. 1

There is also a difference between letting people see a representation and letting them contribute to it. Someone may recognise a missing step or a practical constraint, but that knowledge can only affect the work if the group is willing to change what it has drawn. A shared artifact becomes useful when people can use it to explain what they know and work through the consequences together.

Prototypes let people explore an idea through use. Trying a sequence of screens can reveal a confusing transition that seemed straightforward in a description. But the questions a prototype can answer depend on what has actually been built. A clickable demonstration may help us understand an interaction while telling us little about whether the underlying system will work reliably.

Stephanie Houde and Charles Hill make this distinction in What Do Prototypes Prototype?, separating a product's role, its look and feel, and its implementation. Their approach encourages choosing a prototype for the question being investigated and making its limits clear to the people using it. 3

Where AI-assisted development reduces the effort of making a working example, it can bring this kind of investigation earlier into a project. The team still needs to decide what to examine and involve people whose experience can reveal something the makers have missed.

The useful outcome is a change in understanding that affects the next decision. Someone can explain a concern more clearly, the group can revise an assumption, or a question that needs further investigation can become visible. Putting something between us gives those differences a place to be examined while there is still time to act on them.

Further reading