Modeling Trial-and-Error Navigation With a Sequential Decision Model of Information Scent

A team of researchers has published a new paper, 'Modeling Trial-and-Error Navigation With a Sequential Decision Model of Information Scent,' challenging the classic 'information scent' theory in human-computer interaction. The classic theory assumes users see all links on a page before deciding, but the new model from Xiaofu Jin, Yunpeng Bai, and Antti Oulasvirta argues navigation is a sequential process under memory constraints. Users don't scan entire pages; they inspect strategically, looking 'just enough' given their time budget, which often leads to clicking the wrong link.

This sequential model incorporates both local (page-level) and global (site-level) scent cues, both constrained by human memory. The key insight is that to avoid wasting time, users make premature selections without inspecting everything, leading to the trial-and-error behavior (wrong turns and backtracking) commonly observed in website navigation. Comparisons with empirical data show the model successfully replicates these real-world navigation patterns, providing a more accurate explanation for why users struggle with ambiguous or deeply nested links.

The findings have significant implications for UX designers and AI developers. For designers, it underscores the need to create information architectures that support quick, error-prone navigation rather than assuming perfect user rationality. For AI, particularly agents built for web navigation or RAG systems, the model offers a more human-like framework for predicting and simulating user behavior, which could lead to more intuitive assistive tools and better testing protocols for website usability.