Reading intent off an interface
Lesson 4 of 24, level 01Index
Open any app you use daily and look at the single most prominent button on the home screen. Somebody fought for that. Somebody else lost.
Prominence, friction and default
Three levers explain most of what you see. What is big and central is what they want you to do. What takes extra taps is what they tolerate. What is pre-selected is what most people will end up doing, because almost nobody changes a default.
Every screen trades something away. Simplicity costs power users, density costs beginners, and there is no layout that wins on both. So the question is never “is this good”, it is “who did they decide to lose”.
An interface is a set of priorities made visible. Learn to describe what it encourages before deciding whether you like how it looks.
Use your chosen course project throughout. The additional examples below are fictional practice cases; transfer the method to your own evidence.
Start from one person's task
Choose a specific user and a concrete job, such as finding last month's spending. Describe the starting state, what information they need and what counts as completion. The same dense screen may help a returning expert while overwhelming a first-time user; a critique without a task cannot resolve that difference.
Trace prominence, defaults and friction
Mark the largest element, the primary action, preselected choices and steps that take extra effort. These features suggest what the interface prioritises, but they do not prove the designers' intentions. Phrase intent as a hypothesis and use observable interaction costs as the evidence you can defend.
Separate a usability issue from taste
“I prefer green” does not predict behaviour. “The primary action and secondary action have identical emphasis, so the next step is unclear” is testable. Rank findings by whether they block the task, slow it or merely change appearance. Keep a note of what already works so your redesign does not remove it accidentally.
Three kinds of critique
- Observation: What is visible
- Hypothesis: What it encourages
- Test: What a user actually does
How Instagram handled it
The 2012 home screen had one job: show photos from people you follow, in order.
The camera button moved to the centre, then to a swipe. Chronological went away. Stories took the top strip, then Reels took the bottom centre slot that used to be post. Each move traded a little creator intent for a lot of consumption time, which is the metric Meta monetises. You can read the business strategy off the tab bar.
Prominence is a statement of priority, and priority is usually a statement about money.
Doing it with AI, and where it breaks
Screenshot the interface, give it to a vision model, and ask what the screen is optimising for and what it makes harder. Then argue with it and write down every point where you disagree.
Vision models describe what is visually salient, not what is strategically important. They will tell you the hero image is the focus and miss that the real decision is a pre-ticked checkbox in 11px grey. Salience and intent are different things.
Your AI workbench
Start with your own notes or clearly labelled practice data. Remove private details before sharing. Replace the placeholders, run the prompt in your chosen AI tool, and keep the output beside its source.
Review this screenshot for a user trying to [specific task]. Separate observations, hypotheses about intent and questions requiring a usability test. Identify prominence, defaults and friction. Give three findings with severity and an observable test. Do not invent user behaviour or declare the design bad because of visual taste.
Before you use the output
- Check the screenshot's actual state
- distinguish visible evidence from inferred intent
- test the highest-severity finding with a real task.
Stuck? Try this next
If the AI invents hidden screens, explicitly restrict it to visible evidence. Provide additional screenshots only after naming the missing state.
Keep a brief AI log: input used, useful output, what you checked, and what you rejected. The decision remains yours.
Build it
A one screen teardown: what it optimises for, what it trades away, and who lost the argument.
Figma or ExcalidrawA vision model
The worked solution
Stuck, or want to compare?
There is a worked solution. Try the build first. Reading it before you have attempted anything is the fastest way to learn nothing.
Checkpoint
If you did the build, these take two minutes. If you cannot answer one of them, that is the part to go back to.
Moving on marks this lesson complete. Finish the build first, it is the part that counts.