Searcharxiv⌕ Search

arXiv subjects

Luka Bekavac

Publications and source records attributed to Luka Bekavac.

4 recordsLinked to original sources

A Code-Level Audit of X

The transparency regime of recommender systems on Very Large Online Platforms (VLOPs) is undergoing significant change under the EU Digital Services Act (DSA). The DSA requires platforms to disclose information about their recommendation pipelines and grants vetted researchers access to non-public data, which may extend to recommender code. However, it is not known to what extent such code releases enable substantive and meaningful third-party auditing of a production recommender system. We address this question by proposing a code-level audit methodology for recommender systems and applying it to the X platform, which has voluntarily open-sourced its recommender pipeline. Our methodology is a backwards-walk traversal of the repository, tracing each decision-relevant component from the user-facing timeline back to its upstream producers, checking that it is reachable from the served pipeline, and classifying each under a four-tier inspectability scheme. Applying this procedure, we identify 13 transparency dead-ends on the live serving path: models, services, label lists, and configuration values that the public code depends on but the release does not contain. We find that the components that decide which posts are filtered and how they are ranked fall into the two least inspectable tiers, a pattern that intensifies with the 2025 integration of the proprietary Grok language model into moderation, annotation, and ranking. We further show that the release keeps 2023 code that the 2025 pipeline no longer reaches, with 36 parameter references that would pass for dead-ends without a reachability check. Finally, we discuss which additional artifacts would be required for meaningful auditing.

cs.CY↗

Drowning in AI Slop: How Social Media Platforms (Do Not) Label AI and Deepfake Content under EU law

AI labels are emerging as a primary safeguard for transparency about AI-generated content on social media, including under the EU Digital Services Act and AI Act. Yet, limited systematic evidence exists on how platforms implement such labels in practice. We conduct a legally grounded audit of AI labelling across Instagram, TikTok, X, and YouTube, drawing on the European Commission's July 2026 guidelines on deepfakes. To this end, we analyse platform policies and detection approaches, 10,722 posts collected via systemic-risk and AI-related keywords, an expert-annotated subset of 500 posts, and controlled uploads to the four platforms of outputs from ten popular generative AI tools. We find that AI labelling is now broadly established. All four platforms apply labels automatically and a greater share of labels are platform-applied than in earlier audits. However, coverage remains incomplete where labelling arguably matters most: only 33% of expert-identified deepfakes in systemic risk contexts carried a platform-applied AI label, while reaching a median of 160,000 views. In controlled uploads of AI-generated content carrying standard AI provenance signals, platforms labelled only 61% of uploads, and commonly strip those signals after uploading. Overall, platform rules, label designs, detection approaches, and reporting diverge substantially. In response, we identify concrete opportunities to improve the uptake, clarity and efficacy of AI labelling.

cs.CY↗

Platforms' Research API Data Access: What Users See vs. What Researchers can Retrieve

Article 40(12) of the Digital Services Act (DSA) requires Very Large Online Platforms (VLOPs) to provide vetted researchers with access to publicly accessible data. While prior work has identified shortcomings of platform-provided data access mechanisms, existing research has not quantitatively assessed data quality and completeness in Research APIs across platforms, nor systematically mapped how exactly current access provisions fall short. This paper presents a systematic audit of research-access modalities by comparing data obtained through platforms' Research APIs with data collected about the same platforms' user-visible public information environment (PIE). Focusing on two major platform APIs-the TikTok Research API and the Meta Content Library-we reconstruct the full information feeds for two controlled sockpuppet accounts during two election periods, and benchmark these against the data retrievable for the same posts through the corresponding Research APIs. Our findings show systematic data loss through three classes of platform-imposed mechanisms: scope narrowing, metadata stripping, and operational restrictions. Together, these mechanisms implement overlapping filters that exclude large portions of the platforms PIE (up to ~50%), strip essential contextual metadata (up to ~83%), and impose severe technical constraints for researchers (down to ~1,000 requests per week). Viewed through a data quality lens, these filters primarily undermine the completeness of research-accessible data, resulting in a structurally biased representation. We conclude that, in their current form, Meta & TikTok Research APIs fall short of supporting meaningful, independent auditing of systemic risks as envisioned under the DSA.

cs.CY↗

Change Your Perspective, Widen Your Worldview! Societally Beneficial Perceptual Filter Bubbles in Personalized Reality

Extended Reality (XR) technologies enable the personalized mediation of an individual's perceivable reality across modalities, thereby creating a Personalized Reality (PR). While this may lead to individually beneficial effects in the form of more efficient, more fun, and safer experiences, it may also lead to perceptual filter bubbles since individuals are exposed predominantly or exclusively to content that is congruent with their existing beliefs and opinions. This undermining of a shared basis for interaction and discussion through constrained perceptual worldviews may impact society through increased polarization and other well-documented negative effects of filter bubbles. In this paper, we argue that this issue can be mitigated by increasing individuals' awareness of their current perspective and providing avenues for development, including through support for engineered serendipity and fostering of self-actualization that already show promise for traditional recommender systems. We discuss how these methods may be transferred to XR to yield valuable tools to give people transparency and agency over their perceptual worldviews in a responsible manner.

cs.HC↗