Mental models and concept information: why readers need to know how it works before they act

What the research shows

The term mental model in its cognitive sense goes back to Kenneth Craik (1943), who proposed that the mind builds small-scale models of reality to anticipate events. Philip N. Johnson-Laird (1983) developed it into a theory of comprehension and reasoning: people understand a description by building a model of the situation it describes and draw conclusions by inspecting that model.

Donald A. Norman (1983) studied mental models of devices and observed that they are incomplete, unstable, unscientific and partly superstitious, yet people rely on them. In The Design of Everyday Things (first published in 1988 as The Psychology of Everyday Things, revised 2013) he distinguished three models: the designer's conceptual model, the user's mental model and the system image, everything the product presents of itself, from controls and displays to labels and manuals. Users build their model only from the system image. Where it is unclear, they face what Norman calls the gulfs of execution (how do I make it do what I want?) and evaluation (did it do it?). The theory glossary describes the system image and the gulfs as terms of their own.

David E. Kieras and Susan Bovair (1984) gave learners either a description of how a control panel device worked or only the procedures. Those with the device model learned the procedures faster, retained them better and could infer procedures they had not been taught.

What it means for documentation

Documentation is part of the system image, so it shapes the model the reader builds. Several practices follow:

• Concept before task. A short explanation of how the product works, its parts, states and operating modes, placed before or linked from the tasks, lets readers understand why each step is needed. Information typing captures this as the DITA concept next to the DITA task.
• Models that predict. The concept states causal links ("pressure builds while the valve is closed"), not only names. That is what lets the reader handle a case the procedure does not cover.
• Consistency with the interface. Terms, states and messages in the text match the human-machine interface, so the reader's model and the system image agree.
• Correcting likely misconceptions. Where users predictably hold a wrong model, for example that a stopped machine is de-energized, the concept says so explicitly.

Not every reader needs the concept first: experts often skip it, and minimalist instruction warns against long explanations before action. A concept that is short, linked and optional serves both.

Limits

The evidence comes mostly from studies of device and software learning; how much concept a given audience needs is settled by task analysis and testing. A correct mental model does not make a reader safe. Hazards, warnings and the limits of intended use are stated where they apply in the instructions for use, whatever the reader's model, and their content follows from the risk assessment and the applicable law. No cognitive argument replaces the risk assessment or legally required safety information, and concept information never makes a product compliant.