Algorithmic Transparency Summary

1. Overview of Recommendation Systems

Vexor utilizes a multi-layered recommendation architecture designed to personalize user experiences while maintaining safety, fairness, and regulatory compliance. The system integrates collaborative filtering, behavioral analysis, semantic content classification, and real-time engagement modeling. Our objective is to surface relevant, high-quality, and safe content tailored to user interests.

These algorithms operate within strict governance frameworks aligned with the EU Digital Services Act (DSA), GDPR, and global transparency standards. Vexor does not sell or transfer algorithmic decision-making data to third parties.

2. Key Ranking Signals

Vexor’s models analyze a combination of behavioral signals, contextual relevance indicators, and content attributes to determine which videos appear in recommendations, the For You Feed, category feeds, and search results.

• Watch Time: Primary indicator of viewer satisfaction.
• Engagement Metrics: Likes, shares, comments, rewatches.
• Creator–Viewer Relationship: Following status, past interactions.
• Content Topics: Detected using machine learning classifiers.
• Language & Region: Helps align content with cultural context.
• Device & Session Signals: Basic usage data to optimize real-time recommendations.
• Safety Filters: Prioritize safe and age-appropriate content, especially for minors.
• User Preferences: Categories followed, muted topics, blocked accounts.

Vexor continuously evaluates its models for bias, fairness, representation gaps, and false-positive/false-negative behavior across demographic segments.

3. Excluded Signals

To protect user rights and prevent discriminatory outcomes, the following categories are explicitly excluded from recommendation calculations:

• Race, ethnicity, nationality, or cultural background
• Religion or philosophical beliefs
• Political ideology or voter preferences
• Financial status, income, or wealth indicators
• Sexual orientation or gender identity (unless voluntarily disclosed for personalization)
• Medical or biometric data
• Any data classified as sensitive under GDPR

Vexor does not engage in micro-targeting based on sensitive attributes and does not build shadow profiles or speculative behavioral predictions.

4. User Controls

Users maintain full control over how their data informs personalization. Vexor offers transparent tools to manage, modify, or disable recommendation features:

• Reset Personalization: Clears algorithmic history and recommender influence.
• Opt-Out of Personalized Recommendations: Switch to non-personalized, chronological, or region-neutral feeds.
• Content Preferences Dashboard: Edit content categories, mute topics, or filter types of recommendations.
• Report Inaccurate or Harmful Suggestions: Reports are reviewed by Trust & Safety teams.
• Block & Filter Tools: Blocking an account removes it from all ranking paths.

Opt-out functionality meets DSA requirements for offering a non-profiling-based recommender system.

5. Safety Layers

Vexor integrates multiple protective layers that automatically restrict or down-rank content that could be harmful, misleading, or inappropriate. These systems prioritize user safety, prevent exposure to dangerous material, and reinforce compliance obligations.

• Misinformation Controls: Down-ranks content flagged by risk models or fact-checking partners.
• Dangerous Acts & Violence: Reduced distribution, age-gating, or removal.
• Sexual Content: Strict adult-content classifiers for all feeds.
• Minor Safety Filters: Prevent adult content from appearing in under-16 feeds.
• Hate Speech & Extremism: Automatically demoted or escalated to human review.
• Spam & Manipulation Detection: Identifies bot-like behavior and prevents ranking exploitation.

High-risk cases are escalated to specialist Trust & Safety teams for manual verification, ensuring human oversight and accountability.

6. Transparency & User Rights

Vexor complies with global transparency obligations by:

• Publishing algorithmic summaries and risk assessments
• Providing explanation disclosures when content visibility is restricted
• Allowing users to appeal algorithmic impact decisions
• Providing machine-readable documentation for auditors and regulators

Users may request further detail about their personalization profile through Vexor’s privacy and data access tools.

7. Contact

For transparency inquiries, regulatory requests, or algorithmic audits, contact:
Transparency Office: transparency@vexor.to
Governance & Compliance: compliance@vexor.to
Privacy Office: privacy@vexor.to