Content Recommendation Vector Adjustment for Habit Bias Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Recommender algorithms often rely on habit-based predictions, which can lead to content recommendations that do not accurately reflect a customer's true preferences, as they tend to favor familiar content and overlook diverse options.

Innovation Solution

The system determines a habit-based content selection vector for a customer device based on a habit profile, adjusts this vector to filter out habitual consumption patterns, and generates recommendations using an adjusted content selection vector that better aligns with actual customer preferences.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If recommender algorithms rely on habit-based predictions, then the system can maintain operational simplicity and processing efficiency, but the recommendation accuracy and customer preference alignment deteriorate

Engineering Contradiction:

Improverecommendation processing efficiencyVSAvoidpreference prediction accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent segments the content selection vector into two distinct components: a habit-based portion derived from historical consumption patterns and a non-habit-based portion derived from diverse content interactions. This segmentation allows the system to process habit-based recommendations efficiently while separately identifying and promoting diverse content options, thereby resolving the contradiction between processing efficiency and preference accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by treating different portions of the content selection vector differently. The habit-based portion is processed using efficient historical pattern matching, while the non-habit-based portion is processed using more computationally intensive methods that analyze diverse content interactions. This localized differentiation allows the system to optimize for both efficiency and accuracy in appropriate contexts.

Inventive Principle:

Principle #3Local quality

2Reliability

If the system recommends familiar content based on past behavior, then user engagement from habitual viewing is maintained, but content diversity and discovery opportunities are reduced

Engineering Contradiction:

Improveuser engagement stabilityVSAvoidcontent diversity

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent extracts the habit-based component from the overall content selection vector and separates it from the diverse content recommendations. By taking out the habit-based portion and processing it separately, the system can maintain reliable engagement through familiar content while simultaneously generating separate diverse content recommendations that promote discovery and content versatility.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies preliminary anti-action by proactively counteracting the habit bias before it fully influences recommendations. The system identifies and neutralizes the habit-based portion of the content selection vector in advance, then compensates by emphasizing diverse content options. This preliminary counter-action ensures that habit bias does not dominate the recommendations, maintaining both engagement stability and content diversity.

Inventive Principle:

Principle #9Preliminary anti-action

3Measurement precision

If the content selection vector is adjusted to filter out habitual patterns, then recommendation accuracy improves, but computational complexity increases

Engineering Contradiction:

Improvepreference prediction accuracyVSAvoidalgorithm complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the content selection vector into habit-based and non-habit-based portions, allowing the system to apply different processing methods to each segment. This segmentation reduces overall algorithmic complexity by handling the habit-based portion through simpler historical pattern matching while applying more complex analysis only to the non-habit-based portion where it is most needed for improving accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by adjusting only the non-habit-based portion of the content selection vector rather than completely reprocessing the entire vector. This partial adjustment approach improves preference prediction accuracy by focusing computational resources on the portions of the vector that benefit most from complex analysis, while leaving the habit-based portion processed through simpler methods.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11589121B2Content recommendation techniques with reduced habit bias effects

Publication Date: 2023.02.21 AT&T INTELLECTUAL PROPERTY I L P

AI summary

Aspects of the subject disclosure may include, for example, identifying content consumption data associated with media content consumption at a customer device, and generating a content selection recommendation for the customer device. Some embodiments can include determining a habit-based content selection vector for the customer device. Various embodiments can include determining the habit-based content selection vector based on a habit profile for the customer device. Some embodiments can include adjusting a content selection vector for the customer device based on the habit-based content selection vector for the customer device. Various embodiments can include generating the content selection recommendation for the customer device based on the adjusted content selection vector. Other embodiments are disclosed.