Filter Bubble or Homogenization? Disentangling the Long-Term Effects of Recommendations on User Consumption Patterns

TL;DR

A more refined definition of homogenization and the filter bubble effect is developed by decomposing them into two key metrics: how different the average consumption is between users (inter-user diversity) and how varied an individual's consumption is (intra-user diversity).

Abstract

Recommendation algorithms play a pivotal role in shaping our media choices, which makes it crucial to comprehend their long-term impact on user behavior. These algorithms are often linked to two critical outcomes: homogenization, wherein users consume similar content despite disparate underlying preferences, and the filter bubble effect, wherein individuals with differing preferences only consume content aligned with their preferences (without much overlap with other users). Prior research assumes a trade-off between homogenization and filter bubble effects and then shows that personalized recommendations mitigate filter bubbles by fostering homogenization. However, because of this assumption of a tradeoff between these two effects, prior work cannot develop a more nuanced view of how recommendation systems may independently impact homogenization and filter bubble effects. We develop a more refined definition of homogenization and the filter bubble effect by decomposing them into two key metrics: how different the average consumption is between users (inter-user diversity) and how varied an individual's consumption is (intra-user diversity). We then use a novel agent-based simulation framework that enables a holistic view of the impact of recommendation systems on homogenization and filter bubble effects. Our simulations show that traditional recommendation algorithms (based on past behavior) mainly reduce filter bubbles by affecting inter-user diversity without significantly impacting intra-user diversity. Building on these findings, we introduce two new recommendation algorithms that take a more nuanced approach by accounting for both types of diversity.

Figures (12)

• Figure 1: Identifying the strength of homogeneity and filter bubble effect for different levels of inter-user and intra-user diversity. The dynamics seen here motivate our novel definitions of homogeneity and the filter bubble effect.
• Figure 2: Our definition of homogeneity, formally given by $1 / \sqrt{\text{inter-user diversity}^2 + \text{intra-user diversity}^2}$, vs. the inverse of the standard deviation of all consumption for each of the seven recommendation algorithms. The Pearson correlation coefficient between the two values is $0.93$, i.e. they are highly correlated.
• Figure 3: Inter-user diversity vs. intra-user diversity for the different recommendation algorithms. As shown, the baseline algorithms induce a direct trade-off between the two types of diversity. Past consumption-based recommendation algorithms deviate from this trade-off line primarily by reducing inter-user diversity — they do not significantly affect intra-user diversity.
• Figure 4: Deviation between mean consumed genre (see section \ref{['measures']}) and preference vs. preferences for the recommendation algorithms. Compared to no recommendation, past consumption-based recommendations cause large deviations in mean consumed genre towards $0$, by pushing all users towards items with near-mode genres.
• Figure 5: Consumed genre variance (see section \ref{['measures']}) against user preferences for the recommendation algorithms in section \ref{['recommendation']}. Compared to no recommendation, past consumption-based recommendations decrease variance for near-mode users and increase variance for niche users by pushing everyone towards blockbuster items.

Theorems & Definitions (10)

• Definition
• Definition
• Definition
• Definition
• Definition
• Definition
• Definition
• Definition
• Definition : binned consumption-based recommendation
• Definition : Skewed top pick recommendation