Modeling user attention continuity and interaction mechanisms in continuous interfaces of digital platforms

Abstract

The paradigm shift from paginated to continuously scrolling interfaces has fundamentally altered how users interact with digital content. However, a theoretical framework for quantifying attention continuity in such environments remains absent. This study introduces the Attention Continuity Model (ACM), which quantifies users’ probability of sustained interaction through a composite metric—the Attention Continuity Index (ACI)—now formulated as a non-linear function incorporating quadratic terms for content similarity (S) and novelty (N) to capture their hypothesized inverted-U relationships, along with a linear term for interaction intensity (I). Furthermore, we propose, as a theoretical design specification for future empirical implementation, a Feedback Loop Mechanism (FLM) specifying how real-time user behaviors could dynamically inform content presentation. A 2 × 3 between-subjects controlled experiment ( N = 120) compared infinite scrolling versus paginated interfaces across three content types (similar, novel, mixed). Results demonstrated that infinite scrolling interfaces significantly prolonged session duration (18.5 vs. 12.3 min, t(118) = 11.47, p < .001, Cohen’s d = 2.10) and increased interaction frequency (45.2 vs. 28.7 actions, t(118) = 12.35, p < .001, d = 2.27). The ACI was substantially higher for infinite scrolling (M = 0.78, SD = 0.11) than paginated interfaces (M = 0.52, SD = 0.14). Mixed content (70% similar, 30% novel) yielded optimal attention continuity. The interface type manipulation significantly reduced bounce rates from 28.6% to 12.3%. The FLM’s real-time adjustments are proposed as a future direction rather than a validated empirical finding. This research provides a quantifiable framework for attention modeling in digital platforms, offering theoretical contributions to human-computer interaction and practical guidelines for ethical interface design.

The authors' abstract, as published at the source. Scientific Reports, 2026 · DOI ↗