Long-Term User Interaction Prediction for Content Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional online systems limit user interaction by presenting content mainly from connected users, resulting in users being shown content they are not interested in, which decreases engagement and interaction with the system.
Innovation Solution
The online system selects content for presentation based on user interaction likelihoods by maintaining models that predict user actions over time, using information about user interactions with applications, such as time spent and frequency of use, to determine long-term engagement and visibility, and generates metrics for future engagement using supervised learning methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the online system presents content items from users who are not connected to the viewing user, then the variety of content presented to the user is increased, but the user may be presented with content items in which the user has minimal interest, leading to decreased user interaction
Solution Approach 1:
The system changes the parameters of content selection by incorporating long-term engagement metrics (duration, frequency, recency) alongside traditional affinity scores. This allows the system to present content from non-connected users that aligns with demonstrated user interests, balancing content variety with user engagement by dynamically adjusting selection criteria based on multiple temporal dimensions of user behavior
Solution Approach 2:
The system implements feedback loops where user interactions with presented content (views, likes, shares, time spent) are continuously monitored and fed back into the content selection model. This feedback mechanism allows the system to learn from user responses and refine future content recommendations, ensuring that increased content variety does not compromise user engagement by constantly adapting to demonstrated preferences
2Ease of operation
If the online system merely presents content received from other users connected to the user, then the user is presented with content from their network, but this limits the content presented to the user and reduces the user's interaction with content provided by the online system
Solution Approach 1:
The content selection system serves multiple functions simultaneously: it maintains traditional connected-user content delivery while also incorporating non-connected user content, long-term engagement analysis, and multi-dimensional scoring. This multi-functional approach allows the system to preserve simple connected-user content delivery mechanisms while expanding content sources and improving user interaction with system-provided content through sophisticated selection criteria
3Loss of time
If the online system uses conventional models to predict user actions, then the prediction is based on limited timeframes, but the system is unable to leverage information to determine a likelihood of the user's long-term interaction or engagement with the application
Solution Approach 1:
The system performs preliminary actions by continuously collecting and storing user interaction data (duration, frequency, recency) in advance of content selection needs. This pre-computation and storage of long-term engagement metrics allows the system to quickly leverage this information when predicting user actions, eliminating the need to analyze raw interaction histories in real-time and enabling long-term prediction capabilities without computational delays
Data Source
AI summary
An online system generates one or more models that determine a likelihood of a user interacting with an application over a particular time interval after installing the application. To generate the one or more models, the online system obtains information describing a user's interaction with the application that occurred greater than a threshold time period prior to a time for which user interaction with the application is to be determined. Example user interactions with the application include: usage of the application, numbers of particular interactions with the application, an amount of compensation the application receives from the user, interactions with other users of the application via the application, and any other suitable interactions. Various engagement metrics may be predicted by the one or more models such as an amount of time spent using the application, particular actions taken in the application, and revenue generated by the user in the application.