Joint Notification and Feed Optimization Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Social networking services face challenges in determining which items to display in user feeds and when to send notifications, as existing algorithms focus on immediate user interaction rather than long-term impacts and viral contributions.
Innovation Solution
A machine learned model is developed to optimize notification and feed object display based on both immediate and long-term user interactions, considering the probability of users engaging with feed objects and encouraging others to contribute content, using a joint optimization approach that combines ranking and notifications models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing algorithms focus on immediate user interaction to determine feed object display, then immediate click-through rate is improved, but long-term user contribution and viral impact deteriorate
Solution Approach 1:
The system performs preliminary analysis of both immediate and long-term impacts before displaying feed objects. The machine learned model pre-calculates downstream impact scores and viral potential metrics in advance, allowing the system to select feed objects that will benefit both immediate engagement and long-term contribution before the user actually interacts with them.
Solution Approach 2:
The invention changes the optimization parameters from solely immediate interaction metrics to a composite parameter set that includes both immediate click-through probability and long-term downstream impact. The machine learned model adjusts weights between short-term engagement metrics and long-term contribution metrics dynamically, transforming the decision-making criteria to balance both time horizons.
2Productivity
If notifications are overused to encourage user interaction with feed objects, then immediate engagement is improved, but user fatigue increases and interaction decreases
Solution Approach 1:
Instead of sending notifications for all feed objects or using excessive notification frequency, the system applies partial action by selectively notifying users only for feed objects with high predicted downstream impact. The machine learned model identifies the subset of feed objects where notifications will be most beneficial, avoiding unnecessary notifications that would contribute to user fatigue while still capturing the essential interactions.
Solution Approach 2:
The system implements feedback loops where user responses to notifications are continuously monitored and fed back into the machine learned model. This feedback mechanism allows the system to learn from actual user behavior patterns, adjusting notification strategies based on what works and what causes fatigue, thereby optimizing the balance between engagement and user comfort over time.
3Device complexity
If a single algorithm is used for feed object ranking, then system complexity is reduced, but the ability to capture both immediate and long-term effects deteriorates
Solution Approach 1:
The invention merges multiple specialized algorithms into a unified machine learned model that simultaneously handles both immediate interaction prediction and long-term impact assessment. Rather than maintaining separate ranking algorithms for short-term and long-term effects, the system combines these functions into a single integrated model that processes both types of signals together, reducing overall system complexity while maintaining comprehensive adaptability.
Solution Approach 2:
The machine learned model is designed with universal functionality to handle multiple objectives simultaneously - it can predict immediate click-through behavior, assess long-term downstream impact, evaluate viral potential, and determine notification effectiveness all within a single model framework. This multi-functional approach eliminates the need for multiple specialized algorithms while capturing the full ecosystem effects.
Data Source
AI summary
In an example embodiment, a machine learned model is used to determine whether to send a notification for a feed object to a user. This machine learned model is optimized not just based on the likelihood that the notification will cause the user to interact with the feed object, but also the likely short-term and long-term impacts of the user interacting with the feed object. This machine learned model factors in not only the viewer's probability of immediate action, such as clicking on a feed object, but also the probability of long-term impact, such as the display causing the viewer to contribute content to the network or the viewer's response encouraging more people to contribute content to the network. As such, the machine learned model is optimized not just on notification interactivity but also on feed objects interactivity.