Usability Heuristics, Jakob's Law, and AI with Jakob Nielsen | Episode #91

Jakob Nielsen began working in usability in 1983, when text-based mainframe computing was common, the Macintosh had not yet launched, and few companies practiced user testing. Four decades later, he sees usability as dramatically better. Graphical interfaces made computing more visible, the web gave businesses a direct financial reason to improve online experiences, and artificial intelligence is changing how people tell computers what they want.

The progress does not make established UX principles obsolete. Nielsen argues that human characteristics remain stable even as interfaces change. The practical challenge is to apply those principles to new forms of interaction, use limited research budgets wisely, and keep observing what people actually do.

What you'll learn

• How AI shifts interaction from step-by-step commands toward intent and outcomes

• Why Nielsen's usability heuristics still apply across changing technologies

• How Jakob's Law turns familiar conventions into a design advantage

• Why small studies and repeated iterations can beat one highly precise test

• How observation, mentorship, and competitive research build UX judgment

Design AI around intent, context, and human review

Traditional software requires users to choose commands and specify the steps a computer should take. Nielsen describes AI interaction differently: people state an intended outcome, review what the system produces, and refine the request. His image-generation example shows the pattern. A person can ask for a scene, change one of its subjects, and request a different size without manually performing every editing operation.

That flexibility also makes the experience less predictable. AI-generated words and interface details can vary with the user's situation and context, so designers cannot control every pixel or sentence in advance. Nielsen connects this challenge to earlier transitions. Responsive design already required teams to stop treating one fixed screen as the complete experience, while service design coordinates interactions that vary from one employee or customer encounter to another. AI raises the level of variability, but it does not remove the need for design.

Research with AI also requires judgment. Automated sentiment analysis will not be perfectly accurate, but human analysis is not perfectly accurate either. Nielsen's preferred model combines AI productivity with human experience, critical thinking, and supervision. The tool can extend a researcher's capacity, while the person remains responsible for inspecting its output rather than letting it operate without review.

Apply enduring principles without copying every detail

Nielsen defined his current set of 10 usability heuristics in 1994 after examining real usability problems across telephone systems, mainframes, and personal computers. He calls heuristics broad rules of thumb rather than exact design standards. Visibility of system status means people need to understand what a system is doing, even when a speech interface must communicate that status without a screen. User control and freedom means they need ways to redirect or change what happens. Recognition rather than recall reduces dependence on memory by presenting available commands or choices.

The specific implementation can change while the human need remains. A familiar icon may disappear as technology evolves, but people still benefit when a system helps them recognize an action instead of requiring them to remember it unaided.

Jakob's Law adds a second practical principle: people spend most of their time with interfaces other than yours. Their broader experience forms expectations about where controls appear and how common tasks work. Product teams should identify those conventions rather than inventing a different interaction without a reason. Nielsen recommends testing with five users as an early check. If four expect an element in the upper-right corner and the design places it at the bottom, the pattern deserves attention.

The law leaves room for invention when a category is genuinely new and stable expectations have not formed. Nielsen points to early ecommerce, when the shopping cart convention still had to be discovered, and to emerging three-dimensional interfaces where object manipulation conventions remain unsettled.

Move quickly, observe users, and iterate

Nielsen's discount usability approach grew from a constrained university budget. A mentor encouraged him to treat limited resources as an opportunity, leading him toward methods that rely on small samples and informed professional judgment. The goal is not to claim statistical precision from five participants. It is to identify useful direction, improve the design, and run another round.

He illustrates the tradeoff with an icon test in which one of five participants interpreted a bulletin-board image as laundry. A large study could estimate that misunderstanding more precisely, but it would consume time and money. The team could instead use judgment, move ahead, and watch for the issue in later rounds. Fast studies make three, four, or five iterations possible before release, allowing the final product to improve even when each individual test is imperfect.

For new UX professionals, Nielsen puts observation first. Watch real people use products, including alternatives from other industries that solve similar design problems. Mentors, a small set of newsletters, and conferences can add perspective, but direct user behavior remains the foundation. Modern remote and cloud research tools make that observation easier and less expensive than the lab setups used earlier in his career.

Listen to Episode #91

The lasting lesson is to seek reliable direction rather than false certainty. Familiar patterns, broad heuristics, small tests, and repeated observation give teams a disciplined way to design for both established interfaces and emerging AI experiences.