Effects of Adding Friction to Attention-Capture Patterns in TikTok

Paper Title

Effects of Adding Friction to Attention-Capture Patterns in TikTok

Publication Info

• Topic area: Quantitative evaluation of interventions on attention-capture mechanisms in social media interfaces.
• Keywords: Attention-capture damaging patterns, TikTok, friction-based interventions, user engagement, eye-tracking, user experience, sense of agency, social media design, behavioral metrics, digital well-being.

Background and Problem

• Problem / challenge: Attention-capture damaging patterns (ACDPs) in social media exploit psychological vulnerabilities, prolong engagement, and undermine user agency. Existing interventions primarily focus on external mechanisms and lack quantitative evidence on internal design modifications.
• Significance: Understanding and mitigating ACDPs is critical for improving user well-being, reducing compulsive usage, and informing ethical design practices in social media platforms.
• Motivation and related work: Prior studies have explored external interventions (e.g., browser plugins) and qualitative methods but lack quantitative, platform-specific approaches to directly modify ACDPs. Few studies have examined the impact of friction-based interventions on internal mechanisms, leaving gaps in understanding their effects on engagement, user experience, and agency.

Solution

• Proposed approach: High-fidelity TikTok prototype with friction-based interventions targeting three ACDPs: autoplay, infinite scroll, and social investments.
• Novelty:
  1. Development of a high-fidelity TikTok prototype to measure ACDP interventions with behavioral and attentional metrics.
  2. Integration of eye-tracking data to analyze attentional engagement in short-form video interfaces.
  3. Holistic evaluation combining behavioral data, eye-tracking metrics, user experience (AttrakDiff), and agency (SoAS) measures.
• Procedure and key techniques:
  • Friction-based modifications applied to autoplay (manual play required), infinite scroll (increased scroll friction), and social investments (concealed metrics).
  • Controlled laboratory study with 48 participants using a TikTok-like prototype and eye-tracking device.
  • Behavioral engagement measured via posts viewed, video completion rates, and interaction metrics.
  • Attentional engagement analyzed through gaze metrics across defined areas of interest (AOIs).
  • User experience and agency assessed using validated questionnaires (AttrakDiff and SoAS).

Results

• Concrete findings:
  • Disabling autoplay reduced posts viewed per session (median: 36 → 24) and repetitive video watching, increasing deliberate engagement.
  • Increased scroll friction further reduced posts viewed (median: 36 → 27) but caused high error rates in scroll gestures (median success rate: 0.994 → 0.640).
  • Obstructing social investments had negligible effects on engagement metrics but reduced interface attractiveness (AttrakDiff ATT median: 5.0 → 4.8).
  • Eye-tracking data showed fewer but longer visits to content AOI across all interventions, indicating more focused engagement.
• Advantage over baselines:
  • Quantitative evidence of reduced compulsive engagement patterns through friction-based interventions.
  • Novel insights into attentional engagement using eye-tracking data.
• Experiments / evaluation:
  • Within-subjects study design (N=48) with counterbalanced conditions.
  • Behavioral metrics, eye-tracking data, and questionnaire responses analyzed across four conditions (control and three interventions).
  • Statistical tests (Wilcoxon signed-rank, paired t-tests) applied with Holm-Bonferroni correction.
• Limitations and future work:
  • Controlled laboratory setting limits generalization to real-world usage contexts.
  • Lack of personalized content in prototype may affect ecological validity.
  • Short task durations may not capture long-term effects or dissociative states induced by ACDPs.
  • Future research should explore gradual friction interventions, alternative design patterns, and adaptive ML-based solutions.

Summary

This study quantitatively evaluated friction-based interventions on attention-capture damaging patterns (ACDPs) in TikTok using a high-fidelity prototype and eye-tracking data. Disabling autoplay and increasing scroll friction reduced compulsive engagement patterns but negatively impacted user experience and agency. Obstructing social investments had minimal behavioral effects but reduced interface attractiveness. Eye-tracking revealed more focused attentional engagement across interventions. Findings highlight trade-offs between reducing problematic usage and preserving user experience, emphasizing the need for gradual, adaptive interventions and collaborative research to redesign social media platforms for ethical and sustainable use.