InteractionData

Represents a single user interaction event captured from an application.

InteractionData serves as the fundamental unit of information that flows from applications into the KARL learning system. Each instance captures a discrete user action, behavior, or context change that can be learned from to improve future predictions and recommendations.

Data structure principles:

• Standardized format: Consistent structure across different interaction types

• Rich context: Sufficient detail for meaningful pattern recognition

• Privacy protection: No personally identifiable information in raw data

• Extensible design: Details map allows for domain-specific information

• Temporal tracking: Precise timestamps enable time-series analysis

Common interaction types and examples:

UI Interactions:

• "button_click": User selects specific UI elements or controls

• "menu_selection": Navigation through application menus and options

• "form_submission": Completion of data entry or configuration forms

• "scroll_behavior": Reading patterns and content consumption habits

Content Interactions:

• "document_view": Access to files, articles, or media content

• "search_query": Information seeking and discovery behaviors

• "filter_application": Data filtering and refinement preferences

• "content_creation": User-generated content and editing activities

Navigation Patterns:

• "page_transition": Movement between different application sections

• "workflow_step": Progress through multi-step processes or tasks

• "session_start": Beginning of user interaction sessions

• "session_end": Conclusion of user activities with outcome tracking

Context Events:

• "error_occurrence": Error conditions and user recovery actions

• "performance_metric": System responsiveness and user satisfaction indicators

• "preference_change": User customization and setting modifications

• "feature_discovery": First-time use of application features or capabilities

Details map structure and content guidelines:

Contextual information:

• "session_id": Unique identifier for grouping related interactions

• "screen_name": Current application screen or context

• "feature_area": Functional domain or module where interaction occurred

• "user_goal": Inferred or declared user objective for the interaction

Interaction specifics:

• "element_id": Specific UI element or content identifier

• "action_result": Success, failure, or outcome of the interaction

• "duration": Time spent on the interaction or associated task

• "effort_level": Complexity or difficulty of the interaction

Environmental factors:

• "device_type": Platform or device characteristics

• "network_status": Connectivity and performance conditions

• "time_context": Time of day, day of week, or seasonal factors

• "location_context": Geographic or spatial context when relevant

Performance and quality metrics:

• "response_time": System responsiveness for the interaction

• "accuracy": Correctness or precision of user actions

• "completion_status": Whether the interaction achieved its intended goal

• "user_satisfaction": Explicit or implicit satisfaction indicators

Privacy and security considerations:

• Never include passwords, personal identifiers, or sensitive content

• Use anonymized or hashed identifiers for external references

• Apply data minimization principles to include only learning-relevant information

• Implement automatic expiration for temporary or session-specific data

Data quality and validation:

• Ensure consistent timestamp formats and timezone handling

• Validate interaction types against known vocabularies

• Implement schema validation for details map structure

• Monitor data volume and implement sampling for high-frequency events