Modeling human decision-making in task-heavy experiences is often a daunting and intimidating task for most designers. Early in my career, I learned best practices for arresting and visualizing complex branching logic and I still rely on those best practices today. I have yet to experience an AI agent or tool that can accurately mirror complex branching logic with any level of reliable detail. In fact, the AI tools I experimented with consistently achieved mild mediocrity and never consistency.
Using AI for User Research
The workflows, artifacts and tools typically used in classical UX practices for this stage are being augmented, replaced and hastened by Artificial Intelligence. Click here to see some of the new tools and techniques I’m experimenting with.
My preferred approach…
➤ By hand, using the Jesse James Garrett Visual Vocabulary
Did you know that humanity’s fascination with logic dates back to Socrates? From Plato, Leibniz, Boole… the investigation into human logic is a fascinating story. In 2000, Jesse James Garrett introduced a set of best practices using geometric shapes and symbols to clearly communicate complex branching logic.
Adopting an AI tool for flow diagramming would be a fantastic leap—if that tool can demonstrate adherence to best practices. For me, flows mark the shift from abstract thinking to concrete solutions. Because they’re a critical artifact for communicating user experience, I hold a high standard for the tools I use to create them.
Best practices
➤ Democratization killed best practices
When “democratization” became a trend, and anyone with a six-week UX bootcamp certificate was applauded for saying empathy 20 times in a Design Thinking workshop, I knew the craft of UX was on borrowed time. And now, it’s evaporating—accelerated by AI.
Before diving into the successes and shortcomings of AI tools—and their potential in flow diagramming—it’s important to outline the set of best practices I used for evaluation.
As a UX generalist, when AI falls short, I can resort to old techniques inspired by the following best practices and principles.
| Principle | Description | By Hand / Manual | AI Capable |
|---|---|---|---|
| Diagram Types | There are two types of diagram models that can be produced: - Sitemaps (in support of navigation and browsing architectures) - Task flows (in support of jobs to be done) | Y | N |
| Macrostructure vs. Microstructure | Flow diagrams should focus on what we call macrostructure, providing just enough detail to enable team members to get the “big picture”. Details of an interface should not appear in the diagram — if you find yourself drawing buttons and fields, you’re probably loading the diagram down with excess detail. The detail that should be included involves a simple conceptual model encompassing both information architecture and interaction design: - The system presents the user with paths. - The user moves along these paths through actions. - These actions then cause the system to generate results. | Y | N |
| Repeatable / Reliable | The symbology used in the diagrams have clear definitions and each diagram uses the symbology the same way every time a new diagram is created. | Y | N |
| RE: Interaction Design | When describing interaction design, the diagram should emphasize how the user flows through defined tasks, and what the discrete steps are within these tasks. | Y | N |
| RE: Color | Adding color to diagrams can lead to unnecessary distractions. The idea is to arrest conditional logic and show flow and/or relationships. Color is not required do accomplish this and only adds more time to build the diagram than is necessary. | Y | Sorta |
| Follows Boolean Concepts | Conditional logic is rooted in binary choices that are chained together to create a end-to-end flow. Concepts of binary exploit these terms: AND, OR, NOT | Y | N |
| Incorporates and Interprets Swim Lanes | The ability to represent swim lanes (different environments) involved to complete a task. | Y | N |
Simple elements: pages, files and stacks thereof
Basic concepts: conditional elements
