Feed Actor Optimization via Machine Learning Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Social networking services face challenges in determining which items to display in user feeds and their order, as existing algorithms focus on immediate user interactions rather than long-term impacts, leading to an imbalance where content from prominent contributors is prioritized over newer ones, known as the 'rich getting richer' problem.
Innovation Solution
A machine learned model is employed to assess both short-term and long-term impacts of feed items, considering the probability of immediate actions and long-term contributions, to distribute user attention evenly and motivate all contributors by incorporating factors like expected downstream interactions and the value of responses from actors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing algorithms prioritize items based on immediate user interaction likelihood, then click-through rates improve, but long-term contributor engagement deteriorates
Solution Approach 1:
The system performs preliminary actions by showing feed items to downstream viewers before the actor can see them, creating a preview mechanism that allows users to engage with content in advance. This preliminary exposure generates early interactions that can then be amplified to the actor, addressing both immediate engagement metrics and long-term contributor motivation by ensuring content gets initial traction regardless of the actor's existing prominence
Solution Approach 2:
The system implements feedback loops where downstream viewer interactions (clicks, shares, comments) are captured and fed back to influence the actor's feed visibility. This feedback mechanism creates a virtuous cycle where initial engagement from downstream users amplifies the actor's content reach, motivating continued contribution while maintaining balanced feed distribution that doesn't solely depend on the actor's pre-existing prominence
2Productivity
If feed algorithms focus on prominent contributors, then immediate engagement metrics improve, but content diversity deteriorates
Solution Approach 1:
The system applies local quality by treating different users in the feed ecosystem differently based on their role and position. Downstream viewers receive customized feed items tailored to their preferences and engagement patterns, while actors receive amplified visibility when their content gains downstream traction. This localized customization allows prominent contributors to maintain high engagement metrics while simultaneously providing diverse content opportunities to lesser-known contributors through the downstream amplification mechanism
Solution Approach 2:
By showing feed items to downstream viewers first, the system enables content from diverse sources to gain preliminary engagement before reaching actors. This preliminary action mechanism allows content from lesser-known contributors to accumulate initial interactions that can then trigger amplification, thereby diversifying the content ecosystem while maintaining engagement metrics through the structured preview-and-amplify approach
Data Source
AI summary
A plurality of potential feed objects and corresponding identifications of actors who performed a user interface action that caused a corresponding potential feed object to be generated are obtained. The plurality of potential feed objects and corresponding actor identifications are then fed into a machine learned feed object ranking model, with the machine learned feed object ranking model having been trained via a machine learning algorithm to calculate a score for each of the potential feed objects. The score is based on a combination of a likelihood that the user will perform an interaction, via the user interface, on the potential feed object, likelihood that the user's interaction will cause one or more downstream events by other users, and likelihood that a response from a viewer will cause the actor corresponding to the potential feed object to perform an additional user interface action to generate another potential feed object.