Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Algorithmic Audit of Personalisation Drift in Polarising Topics on TikTok

Branislav Pecher, Adrian Bindas, Jan Jakubcik, Matus Tuna, Matus Tibensky, Simon Liska, Peter Sakalik, Andrej Suty, Matej Mosnar, Filip Hossner, Ivan Srba

Abstract

Social media platforms have become an integral part of everyday life, serving as a primary source of news and information for many users. These platforms increasingly rely on personalised recommendation systems that shape what users see and engage with. While these systems are optimised for engagement, concerns have emerged that they may also drive users toward more polarised perspectives, particularly in contested domains such as politics, climate change, vaccines, and conspiracy theories. In this paper, we present an algorithmic audit of personalisation drift on TikTok in these polarising topics. Using controlled accounts designed to simulate users with interests aligned with or opposed to different polarising topics, we systematically measure the extent to which TikTok steers content exposure toward specific topics and polarities over time. Specifically, we investigated: 1) a preference-aligned drift (showing a strong personalisation towards user interests), 2) a polarisation-topic drift (showing a strong neutralising effect for misinformation-themed topics, and a high preference and reinforcement of interest of US politic topic); and 3) a polarisation-stance drift (showing a preference of oppose stance towards US politics topic and a general reinforcement of users' stance by recommending items aligned with their stance towards polarising topics). Overall, our findings provide evidence that recommendation trajectories differ markedly across topics, with some pathways amplifying polarised viewpoints more strongly than others and offer insights for platform governance, transparency and user awareness.

Algorithmic Audit of Personalisation Drift in Polarising Topics on TikTok

Abstract

Social media platforms have become an integral part of everyday life, serving as a primary source of news and information for many users. These platforms increasingly rely on personalised recommendation systems that shape what users see and engage with. While these systems are optimised for engagement, concerns have emerged that they may also drive users toward more polarised perspectives, particularly in contested domains such as politics, climate change, vaccines, and conspiracy theories. In this paper, we present an algorithmic audit of personalisation drift on TikTok in these polarising topics. Using controlled accounts designed to simulate users with interests aligned with or opposed to different polarising topics, we systematically measure the extent to which TikTok steers content exposure toward specific topics and polarities over time. Specifically, we investigated: 1) a preference-aligned drift (showing a strong personalisation towards user interests), 2) a polarisation-topic drift (showing a strong neutralising effect for misinformation-themed topics, and a high preference and reinforcement of interest of US politic topic); and 3) a polarisation-stance drift (showing a preference of oppose stance towards US politics topic and a general reinforcement of users' stance by recommending items aligned with their stance towards polarising topics). Overall, our findings provide evidence that recommendation trajectories differ markedly across topics, with some pathways amplifying polarised viewpoints more strongly than others and offer insights for platform governance, transparency and user awareness.
Paper Structure (15 sections, 5 figures, 2 tables)

This paper contains 15 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Evaluating Personalisation Drift on TikTok
  4. Audit Overview
  5. User Interaction Predictor
  6. Audit Phases
  7. Evaluation methodology
  8. RQ1: Comparing Drift Between Neutral and Polarising Topics
  9. RQ2: Comparing Drift Between Stances of Individual Topics
  10. Ethical Considerations and Limitations
  11. Conclusion
  12. Additional Methodology Details
  13. Technical Details of Audit Implementation
  14. Supplementary Results Details
  15. Topic-Related Hashtag Popularity Analysis

Figures (5)

  • Figure 1: Visualisation of the audit phases.
  • Figure 2: Ratio of videos for the polarising and cooking topics over time for users seeded with both topics (neutral+polarising). We can see that the neutral cooking topic completely dominates the personalisation across all polarisation topics.
  • Figure 3: Ratio of videos for the polarising and cooking topics over time for users seeded with the polarising topics only (polarising only). We observe a strong neutralising polarisation-topic drift for climate change and vaccines, while US politics show an equilibrium behaviour with a large ratio of political videos.
  • Figure 4: Average number of videos with specific stance across topics. We observe an equilibrium behaviour for polarising-stance drift with similar number of videos from the seeded stance over time and only few videos from opposing stance.
  • Figure 5: Preference-aligned (left) and polarisation-stance (right) drift for users seeded equally with both stances for the US politics topic (mixed polarity). We observe a stronger tendency of the system to recommend videos from the oppose stance.