Interactive visualizations for adolescents to understand and challenge algorithmic profiling in online platforms

TL;DR

This paper tackles adolescents' limited visibility into how online platforms profile and use their data. It presents Algorithmic Mirror, a situated visualization tool that ingests real watch histories from YouTube, Netflix, and TikTok to reveal cross-platform datafication and long-term inferences, employing semantic clustering with text embeddings and UMAP, plus a temporal timeline. Through a 27-participant user study, the work demonstrates that personalized, data-driven visualizations foster emotional engagement, scale-aware understanding, and self-reflection, driving a sense of digital agency and demands for transparency and control. The findings offer design and policy implications for enabling adolescents to understand, challenge, and influence the personalized inferences shaping their online selves, with potential for sustained behavioral change and more transparent platform practices.

Abstract

Social media platforms regularly track, aggregate, and monetize adolescents' data, yet provide them with little visibility or agency over how algorithms construct their digital identities and make inferences about them. We introduce Algorithmic Mirror, an interactive visualization tool that transforms opaque profiling practices into explorable landscapes of personal data. It uniquely leverages adolescents' real digital footprints across YouTube, TikTok, and Netflix, to provide situated, personalized insights into datafication over time. In our study with 27 participants (ages 12--16), we show how engaging with their own data enabled adolescents to uncover the scale and persistence of data collection, recognize cross-platform profiling, and critically reflect algorithmic categorizations of their interests. These findings highlight how identity is a powerful motivator for adolescents' desire for greater digital agency, underscoring the need for platforms and policymakers to move toward structural reforms that guarantee children better transparency and the agency to influence their online experiences.

Figures (7)

• Figure 1: Algorithmic Mirror Interface. The four main components of the interface are highlighted with dashed lines.
• Figure 2: Algorithmic Mirror upload user flow. Users can drag and drop exported watch histories from three supported platforms. Once uploaded, the automated processing is initiated, and users receive an email when the processing is complete.
• Figure 3: Algorithmic Mirror pre-processing pipeline. User watch history exports contain sparse and platform-specific data. The pre-processing pipeline enriches these exports by filling in missing information, concatenating disparate sources into a unified dataset, and harmonizing heterogeneous content descriptions for each video before passing them to the embedding and topic extraction modules.
• Figure 4: Cross-Platform Data Integration. The top-left shows Netflix data, the top-right TikTok data, and the bottom-left YouTube data. The bottom-right overlays all three, where intermingled clusters reveal semantically similar content descriptions across platforms, illustrating how a persona is constructed across different services.
• Figure 5: Top: Zoom and Pan feature. Users can pan across the Mirror to explore different regions of content. Zooming in reveals less frequent subtopics within higher-level themes. At high zoom levels, contour lines indicate dense clusters, which fade as individual items appear—first by title, and eventually by thumbnail. The dot color indicates the platform where each video was watched. Bottom: Timeline Slider feature. Users can drag the slider to explore how their viewing content evolves over time, or click play to watch the progression unfold automatically. The starting point can be adjusted to focus on specific periods and reveal the most common themes within that time frame.