[Paper Review] Content Moderation on Social Media in the EU: Insights From the DSA Transparency Database

Chiara Drolsbach, Nicolas Pröllochs|arXiv (Cornell University)|Dec 7, 2023

Hate Speech and Cyberbullying Detection4 citations

TL;DR

This study analyzes 156 million Statements of Reasons (SoRs) from the EU's DSA Transparency Database to examine content moderation practices across major social media platforms. It reveals vast disparities in moderation frequency, heavy reliance on automation, inconsistent application of rules, and fragmented enforcement—highlighting systemic inconsistencies that undermine the DSA’s goal of harmonized platform accountability in the EU.

ABSTRACT

The Digital Services Act (DSA) requires large social media platforms in the EU to provide clear and specific information whenever they remove or restrict access to certain content. These "Statements of Reasons" (SoRs) are collected in the DSA Transparency Database to ensure transparency and scrutiny of content moderation decisions of the providers of online platforms. In this work, we empirically analyze 156 million SoRs within an observation period of two months to provide an early look at content moderation decisions of social media platforms in the EU. Our empirical analysis yields the following main findings: (i) There are vast differences in the frequency of content moderation across platforms. For instance, TikTok performs more than 350 times more content moderation decisions per user than X/Twitter. (ii) Content moderation is most commonly applied for text and videos, whereas images and other content formats undergo moderation less frequently. (ii) The primary reasons for moderation include content falling outside the platform's scope of service, illegal/harmful speech, and pornography/sexualized content, with moderation of misinformation being relatively uncommon. (iii) The majority of rule-breaking content is detected and decided upon via automated means rather than manual intervention. However, X/Twitter reports that it relies solely on non-automated methods. (iv) There is significant variation in the content moderation actions taken across platforms. Altogether, our study implies inconsistencies in how social media platforms implement their obligations under the DSA -- resulting in a fragmented outcome that the DSA is meant to avoid. Our findings have important implications for regulators to clarify existing guidelines or lay out more specific rules that ensure common standards on how social media providers handle rule-breaking content on their platforms.

Motivation & Objective

• To provide the first empirical analysis of real-world content moderation decisions in the EU using the newly established DSA Transparency Database.
• To investigate how different social media platforms enforce content moderation rules under the Digital Services Act (DSA).
• To assess the extent of automation in content moderation and its implications for transparency and fairness.
• To identify variations in moderation frequency, content types targeted, reasons for action, and types of actions taken across platforms.
• To inform regulators about inconsistencies in platform implementation of DSA obligations and the need for clearer, standardized guidelines.

Proposed method

• Collected and analyzed 156 million Statements of Reasons (SoRs) submitted by large social media platforms to the DSA Transparency Database over a two-month period.
• Categorized SoRs by platform, content type (text, video, image, etc.), moderation reason (e.g., illegal speech, pornography, misinformation), and action type (removal, visibility reduction).
• Classified moderation decisions as automated or non-automated based on platform disclosures in the SoRs.
• Quantified moderation frequency per user across platforms to compare scale and intensity of enforcement.
• Mapped variations in rule application and action types across platforms to assess compliance divergence.
• Used descriptive and comparative statistical analysis to identify patterns and inconsistencies in DSA-compliant content moderation.

Experimental results

Research questions

• RQ1 How frequently is social media content subject to content moderation in the EU?
• RQ2 How often are different types of content (text, images, video, etc.) moderated on social media platforms?
• RQ3 What are the specific legal grounds (reasons) cited for content moderation decisions?
• RQ4 To what extent are content moderation decisions automated versus manually reviewed?
• RQ5 What types of content moderation actions (e.g., removal, visibility reduction) do platforms implement, and how do they vary?

Key findings

• TikTok performs more than 350 times more content moderation decisions per user than X/Twitter, indicating vastly different enforcement intensities across platforms.
• Content moderation is most frequently applied to text and video content, while images and other formats are moderated significantly less often.
• The primary reasons for moderation are content falling outside the platform’s scope, illegal or harmful speech, and pornography/sexualized content, with misinformation being a relatively rare justification.
• The majority of rule-breaking content is detected and processed through automated systems, though X/Twitter is an exception, reporting exclusive use of non-automated methods.
• Platforms vary significantly in their moderation actions: while most remove rule-breaking content, others frequently reduce its visibility, indicating divergent enforcement strategies.
• These findings reveal substantial inconsistencies in how platforms interpret and implement their DSA obligations, undermining the regulation’s goal of harmonized, transparent content moderation across the EU.