Mental model

A mental model is the picture in someone's head of how a product works. Nielsen Norman Group defines it as "what the user believes about the system (web, application, or other kind of product) at hand." It is built from past experience, other products, what people have been told and what they have seen this product do. It is a belief, not a fact, and it is often incomplete or wrong in ways that still work well enough day to day.

Why it matters

People act on their mental model. If it matches how the system really behaves, they predict outcomes correctly and feel in control. If it does not, they make errors, get surprised and blame themselves or the product. A common example is the difference between saving and syncing: a user who thinks "save" writes to the cloud may lose work when the file only lives on one device.

Mental models are shaped heavily by other products. NN/g links this to Jakob's law: users spend most of their time on other sites and bring those expectations with them.

How to apply it

NN/g describes two ways to close a gap between a user's model and the system: make the system conform to users' mental models, or improve users' mental models so they more accurately reflect the system.

Do:

• Research the models people already hold, through interviews, card sorting or watching them think aloud.
• Follow established conventions where your product works like others (cart icons, swipe to delete, back buttons).
• Where your product works differently, say so plainly at the moment it matters, not only in a help page.
• Give immediate, visible feedback so people can update their model as they go.

Don't:

• Assume your own model is the user's. The team knows the internals; users only see the surface.
• Reuse a familiar pattern for something that behaves differently, such as a trash icon that deletes permanently with no undo.
• Rely on a one-time tutorial to correct a deeply held expectation.

Common mistakes

• Designing to the architecture. Menus organized by internal team or database structure match the company's model, not the user's.
• Ignoring variation. Different user groups can hold different models. Experts and newcomers often need different explanations.
• Treating misuse as user error. Repeated "wrong" use is usually evidence of a mental model the design has encouraged.

In AI products

AI products are especially hard to model because their behavior is probabilistic and changes over time. People often arrive with models borrowed from search engines, human assistants or science fiction, and those set the wrong expectations. Google's People + AI Guidebook warns that mismatched mental models can lead to unmet expectations, frustration, misuse and product abandonment, and advises teams to be up front about what the product can and can't do the first time someone uses it. It suggests explaining what the AI can do, what it can't, how it may change and how to improve it. The AI-specific version of this topic is covered under AI mental models.