Modeling the Intuitiveness of Mobile GUI Navigation by Understanding and Simulating Users' Attention Distribution

Analyzing users' attention distribution is an effective way to evaluate whether the primary UI content captures users' engagement. This has led to the development of eye trackers that gather data on users' attention, as well as computational models that simulate this distribution on individual UI screens. However, there is limited research on how users' attention distribution correlates with the intuitiveness of goal-oriented UI navigation. Additionally, simulating user attention as they reason and identify the correct UI element to navigate to the target screen remains unexplored. To address this gap, we introduce an AI-driven model, the UI Link Transformer (UILT), which predicts users' attention distribution as they navigate from the current UI screen to the target UI screen. This model helps designers to evaluate the intuitiveness of UI navigation. Our initial study aimed to understand how users typically identify the UI element that links two consecutive UI screens and how this identification relates to the intuitiveness of the UI navigation design. The insights gained from this study offer designers actionable recommendations to improve the intuitiveness of mobile GUI navigation, with a focus on users' attention distribution. Moreover, the dataset collected during this study supports the development of the UILT. Building on the insights and data from the initial study, we designed, trained, and evaluated the UILT.