Plan a drive across town and you’ll form a route before you leave. But you don’t actually drive the route you planned — you drive whatever the road gives you. A light changes, a lane closes, a pedestrian steps out, someone lets you merge. You arrive having executed almost none of your original plan step by step, and yet you’d say, accurately, that you got where you were going on purpose.
That gap between the plan you made and the actions you actually took is the starting point for one of the more influential ideas in how we understand human-computer interaction: the concept of a situated action.
Two Views of Human Action
Researcher Lucy Suchman, writing in her 1987 book Plans and Situated Actions, laid out two competing pictures of how human behavior works. The first, dominant in cognitive science and early artificial intelligence, treats plans as prerequisite to action — you form a plan, then carry it out, and anyone watching can understand your behavior by figuring out the plan behind it.
The second view, which Suchman argued for, treats action as fundamentally situated: shaped moment to moment by the specific, contingent details of the circumstances you’re actually in, which no plan formed in advance could fully anticipate. Plans still exist and still matter — but as a loose resource you draw on, not a script you follow.
Suchman’s Key Insight
The core claim isn’t that planning is useless. It’s that the relationship between a plan and the action it describes is much looser than it appears after the fact. A plan is good at expressing an intention and orienting you toward a goal. It’s bad at anticipating every specific circumstance you’ll encounter on the way there. So the actual, moment-to-moment doing of anything — driving across town, cooking a familiar recipe, having a conversation — is necessarily improvised against real conditions that the plan never specified.
Suchman summarized this with a line that’s since become the most quoted phrase from the book: plans are “a resource for action,” not a determinant of it.
Why This Matters for Language
Linguists will recognize a close cousin of this idea in pragmatics, the study of how meaning depends on context rather than existing fully formed in the words themselves. Words like “here,” “now,” “that one,” or “in a minute” don’t have fixed meanings — they point at something in the immediate situation, and you can only interpret them by being situated in that same context. A conversation isn’t the execution of a script either; it’s built turn by turn, each side adjusting to what the other just did, in a way no transcript written in advance could have predicted.
Situated action theory essentially generalizes this insight from language to action as a whole: almost everything people do is context-dependent in the same way that indexical words like “here” and “now” are.
From the Lab to the Screen
This reframing had a direct, practical effect on how technology gets built. If action is situated rather than planned, then a system designed around a rigid model of “the steps the user will follow” is going to fail the moment reality diverges even slightly from that model — which, per Suchman’s argument, is almost always. This is part of why user-centered design and contextual, ethnographic research — watching how people actually behave in their real environments, rather than designing from an assumed sequence of steps — became central to serious interface design in the decades after Suchman’s book appeared.
What This Means for You
When a piece of software feels rigid — when it insists you follow its steps in its order, and breaks down the moment your real situation doesn’t match — you’re experiencing the planning model directly, and its limits. The next article in this series looks at exactly that breakdown in action, through the study that gave Suchman’s ideas their most famous illustration: a 1980s office photocopier that couldn’t keep up with the people trying to use it.