From TikTok to Telegram: Cross-Platform Efficacy and User Acceptance of Erroneous and Flawless Misinformation Interventions

Paper Title

From TikTok to Telegram: Cross-Platform Efficacy and User Acceptance of Erroneous and Flawless Misinformation Interventions

Publication Info

• Topic area: Cross-platform evaluation of misinformation interventions under varying conditions.
• Keywords: misinformation, social media, interventions, cross-platform, TikTok, Telegram, X, user acceptance, fact-checking, accuracy.

Background and Problem

• Problem / challenge: Misinformation interventions are often evaluated under ideal conditions, but real-world systems are prone to errors, such as false positives and negatives, which undermine their efficacy and user trust.
• Significance: Addressing misinformation is critical for reducing harm caused by misleading content on social media, especially as platforms increasingly rely on automated or community-driven interventions.
• Motivation and related work: Prior research has focused on text-based misinformation and platform-specific interventions, often under ideal conditions. There is limited understanding of intervention robustness across modalities (e.g., video, voice) and platforms (e.g., TikTok, Telegram, X), as well as the impact of errors in interventions.

Solution

• Proposed approach: A large-scale online experiment evaluating five misinformation interventions (inoculation, accuracy prompt, community note, fact-check, indicators) across three platforms (TikTok, Telegram, X) under flawless and erroneous conditions.
• Novelty:
  1. Cross-platform evaluation of misinformation interventions across different modalities.
  2. Analysis of the impact of intervention errors on efficacy and user perception.
  3. Comparison of user acceptance metrics (helpfulness, annoyance) for various interventions.
  4. Insights into the robustness of interventions across platforms and modalities.
• Procedure and key techniques:
  • Conducted an online experiment with 1,004 participants from the U.S.
  • Participants were exposed to 18 social media posts (9 misinformation, 9 accurate) with or without interventions.
  • Interventions were tested in both flawless and erroneous conditions.
  • Statistical analysis included cumulative link mixed models (CLMMs) and ordinal logistic regression to assess efficacy and user acceptance.

Results

• Concrete findings:
  • Fact-checks and indicators significantly reduced perceived accuracy of misinformation, while community notes had weaker effects.
  • Inoculation and accuracy prompts did not significantly influence accuracy ratings or sharing intentions.
  • Errors in interventions nullified their efficacy in reducing misinformation accuracy ratings.
• Advantage over baselines:
  • Flawless fact-checks and indicators outperformed other interventions and the control group in reducing misinformation accuracy ratings.
  • Community notes showed some efficacy but were less impactful than fact-checks and indicators.
• Experiments / evaluation:
  • Platforms: TikTok (video), Telegram (voice), X (text-image).
  • Metrics: Accuracy ratings, sharing intentions, perceived helpfulness, and annoyance.
  • Findings were consistent across platforms and modalities, indicating robustness.
• Limitations and future work:
  • Combined false positives and negatives into a single category; future work should disentangle these effects.
  • Simulated error rates were higher than real-world systems; more realistic modeling is needed.
  • Cross-platform content conversion may not fully replicate real-world conditions.
  • Future studies should explore trust as a dimension of user acceptance and evaluate interventions in real-world settings.

Summary

This study evaluates the efficacy and user acceptance of five misinformation interventions (inoculation, accuracy prompt, community note, fact-check, indicators) across three platforms (TikTok, Telegram, X) under both flawless and erroneous conditions. Fact-checks and indicators were the most effective in reducing perceived misinformation accuracy, while errors in interventions critically undermined their efficacy. User acceptance was higher for interventions providing transparent and integrated information, such as fact-checks and community notes, but these were also rated as more annoying. The findings highlight the importance of designing accurate and trustworthy interventions and demonstrate their robustness across platforms and modalities, offering valuable insights for combating misinformation in real-world social media environments.