Cross-Platform Toxicity Detection via Telemetry Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The video game industry faces challenges in detecting, reporting, and responding to gaming toxicity due to the lack of mechanisms for coordinating user profiles across platforms, inadequate storage of multimedia data, and the inability to filter context-specific toxic behavior in video game environments, leading to underreported incidents and false positives.
Innovation Solution
A system that extracts telemetry data from video game environments, reconstructs game data for manual review, and uses machine learning models to detect and report toxic behavior while masking Personally Identifiable Information (PII) to ensure privacy and context preservation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system stores and processes complete native asset data for toxicity detection, then detection accuracy is improved, but data storage requirements and processing complexity increase significantly
Solution Approach 1:
The system extracts only the necessary telemetry data from complete native asset data for toxicity detection purposes. This extraction process isolates relevant behavioral and interaction data while discarding unnecessary game content, thereby reducing storage requirements while maintaining detection accuracy.
Solution Approach 2:
The system segments the data processing into two distinct stages: first extracting and storing compact telemetry data for analysis, and second reconstructing only the specific portions of native asset data needed for manual review of detected incidents. This segmentation eliminates the need to store complete native asset data while preserving detection capability.
2Quantity of substance
If the system extracts and stores only telemetry data, then data storage efficiency is improved, but the ability to reconstruct context for manual review deteriorates
Solution Approach 1:
The system performs preliminary extraction and storage of telemetry data that contains all necessary information for toxicity detection and reconstruction. By carefully designing the telemetry data structure to include relevant contextual information, the system prepares the data in advance for efficient reconstruction without needing to store complete native asset data.
Solution Approach 2:
The system creates a simplified copy of the native asset data in the form of telemetry data that preserves essential information for toxicity detection and manual review. This telemetry copy contains extracted relevant information while occupying minimal storage space, enabling efficient reconstruction when needed.
3Measurement precision
If the system reconstructs complete native asset data for manual review, then review accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
The system extracts and stores pre-processed telemetry data that contains all essential information needed for manual review. When reconstruction is needed, the system only reconstructs the specific portions of data relevant to the detected incident rather than rebuilding complete native asset data, significantly reducing processing time while maintaining review accuracy.
4Object-affected harmful factors
If the system masks PII in reconstructed data, then user privacy is improved, but the ability to trace incidents to specific users deteriorates
Solution Approach 1:
The system uses anonymized user identifiers as intermediaries in the reconstructed data. These identifiers mask personal information while preserving the ability to trace incidents to specific users through the anonymized reference system. The masking is selective, removing only sensitive PII while maintaining functional traceability for incident investigation and enforcement actions.
Data Source
AI summary
In some embodiments, native asset data of an asset corresponding to a cross-platform profile may be monitored, where the cross-platform profile includes a profile linked to an account, for a user, that is used across multiple assets. An incident of the user may be detected via a machine learning model based on telemetry data extracted from the native asset data, where the machine learning model is trained to detect known incidents in training data including labeled telemetry data. A status of the cross-platform profile may be updated based on the incident, and the status may be generated for presentation in a user interface for the account. In response to receiving a user request to provide reconstructed asset data based on the incident, the reconstructed asset data may be generated based on the incident for presentation in the user interface for the account.