Algorithmic Transparency Summary

1. Overview

Vibble uses machine-learning models, heuristic rules, and safety filters to determine which posts, media, and accounts users see in their timelines, search results, trends, and Explore recommendations. Because Vibble operates in real-time with high-velocity posting patterns, the recommendation system is built for:

• Fast ranking and resurfacing of new content
• Integrity checks to mitigate manipulation, spam, and coordinated influence
• Safety protections against harmful or misleading content
• Personalization based on user interests and platform interactions
• Fair distribution of content across diverse voices and communities

This document provides a high-level but detailed overview. Additional detail is available in the Algorithmic Accountability Policy and technical documentation.

2. Core Ranking Systems

Vibble uses multiple ranking models depending on the surface:

• Home Timeline Ranking: Personalized ranking of posts from followed accounts, recommended users, and algorithmically surfaced content.
• Explore / Discovery Ranking: Topic, interest, and behavior-based prediction models.
• Search Ranking: Query relevance combined with integrity and safety scoring.
• Trends Ranking: Volume + velocity + geographic interest with safety filters applied.
• Replies & Quote Post Ranking: Prioritized display based on relationship, relevance, safety signals, and toxicity reduction.

3. Key Ranking Signals

Vibble may consider the following signals when generating personalized or non-personalized recommendations:

• User Interactions: Likes, reposts, replies, dwell time, profile clicks, language preferences.
• Content Quality Signals: Freshness, readability, media type, content structure, and engagement velocity.
• Account-Level Signals: Age of account, posting patterns, authenticity checks, user trust score.
• Semantic & Topic Features: NLP-based classification of post topics, sentiment, and themes.
• Safety & Integrity Signals: Toxicity detection, misinformation indicators, policy risk scoring, content sensitivity labels.
• Regional & Language Context: Local trends, cultural signals, time-zone behavior patterns.

4. Signals Vibble Does Not Use for Ranking

To protect user rights and avoid discrimination, Vibble does NOT use the following signals:

• Race, ethnicity, or national origin
• Religion or belief systems
• Gender identity or sexual orientation
• Health status or disability
• Financial status, income, or credit data
• Political ideology or party alignment
• Contacts, private messages, or off-platform behavior

Vibble does not sell user data or provide ranking advantages based on paid relationships.

5. Safety Layers & Integrity Protections

Before any recommendation is delivered, Vibble applies a multi-layer integrity pipeline that removes or suppresses content violating policy or posing systemic risks:

• Zero-Tolerance Filters: CSAM, terrorism, explicit illegal activity, major violations.
• Toxicity & Harassment Filters: Down-rank or hide abusive replies and brigading patterns.
• Misinformation Classifiers: Detect harmful political or crisis misinformation.
• Graphic & Sensitive Media Review: Hide behind sensitive media warnings.
• Spam & Automation Controls: Remove content from bots, link farms, or manipulation rings.
• Election Integrity Rules: Special protections during election cycles and civic events.
• Child Safety Controls: Filters that prevent minors from encountering harmful content.

6. User Controls for Algorithmic Personalization

Vibble gives users the ability to control how personalization affects their experience:

• Switch Between “For You” and “Following” Feeds: Direct choice of algorithmic or chronological content.
• Reset Personalization: Clears stored behavioral signals and restarts ranking from defaults.
• Disable Interest-Based Recommendations: Removes personalization features across Explore and Home.
• Mute, Block & Restrict Tools: Removes unwanted users, topics, or content types.
• Label Dismissal Feedback: Allows users to indicate inaccurate or unwanted recommendations.
• Topic Opt-Out: Opt out of topics such as politics, finance, or sensitive categories.

7. Trending, Explore & Search Transparency

Vibble publicly discloses the factors determining whether a topic or post appears in Trends or Explore:

• Conversation volume and velocity
• Geographic momentum
• Authenticity of engagement (bot filters applied)
• Safety checks (removal of harmful or sensitive trends)
• Misinformation filters for elections and crises

Vibble may remove or suppress trends that are:

• Manipulated by coordinated networks
• Harmful to minors or vulnerable groups
• Linked to illegal activity or incitement
• Based on false claims that could cause real-world harm

8. Explainability & User Education

Vibble provides users with transparency features that explain why content appears in feeds:

• “Why am I seeing this?” explanations
• Labels for sensitive or contextualized content
• Public documentation of major ranking factors
• Clarification when commercial relationships influence placement (Ads)

9. Manipulation, Bots & Abuse Prevention

Algorithmic integrity systems identify harmful or coordinated behavior:

• Bot detection using behavioral patterns, velocity signals, and device fingerprinting
• Spam filters for repeated posting, link farms, and automated replies
• Down-ranking of accounts with repeated policy violations
• Network analysis for identifying abuse rings, misinformation cells, or astroturfing

These systems feed directly into moderation and visibility controls.

10. Data Protection & Privacy

Vibble’s algorithmic systems comply with:

• GDPR (EU)
• DSA (EU)
• Online Safety Act (UK)
• CCPA (California)
• Global data-protection standards

Personalization does not rely on:

• Third-party data brokers
• Sensitive personal attributes
• Private messages or off-platform behavior

11. Appeals & Corrections

Users may appeal or challenge algorithmic decisions that significantly impact content reach:

• Appeal removal of posts or sensitive media labels
• Request explanation of ranking or visibility limitations
• Challenge suspected shadowbans or integrity restrictions

Appeals are handled by a separate, non-automated review team.

12. Oversight, Audits & Accountability

Vibble performs continuous audits to detect bias, fairness issues, and safety gaps:

• Internal algorithmic audits every quarter
• External audits where required by law (e.g., DSA VLOP obligations)
• Bias reduction training and model calibration
• Cross-functional oversight by Nexa Group governance and safety teams

13. Contact Information

For transparency, algorithmic concerns, or regulatory questions:

Transparency Office: transparency@vibble.com
Algorithmic Integrity: algorithms@vibble.com
Nexa Group Compliance: compliance@nexa-group.org

14. Updates to This Summary

Vibble may revise this document to reflect new ranking models, transparency requirements, regulatory changes, or user controls. Major updates will be published with a revised date and linked in the Transparency Center.