Debiasing Recommendation with Personal Popularity

TL;DR

A user-aware version of item popularity named personal popularity (PP), which identifies different popular items for each user by considering the users that share similar interests, is proposed, which helps to produce personalized recommendations and mitigate GP bias.

Abstract

Global popularity (GP) bias is the phenomenon that popular items are recommended much more frequently than they should be, which goes against the goal of providing personalized recommendations and harms user experience and recommendation accuracy. Many methods have been proposed to reduce GP bias but they fail to notice the fundamental problem of GP, i.e., it considers popularity from a \textit{global} perspective of \textit{all users} and uses a single set of popular items, and thus cannot capture the interests of individual users. As such, we propose a user-aware version of item popularity named \textit{personal popularity} (PP), which identifies different popular items for each user by considering the users that share similar interests. As PP models the preferences of individual users, it naturally helps to produce personalized recommendations and mitigate GP bias. To integrate PP into recommendation, we design a general \textit{personal popularity aware counterfactual} (PPAC) framework, which adapts easily to existing recommendation models. In particular, PPAC recognizes that PP and GP have both direct and indirect effects on recommendations and controls direct effects with counterfactual inference techniques for unbiased recommendations. All codes and datasets are available at \url{https://github.com/Stevenn9981/PPAC}.

Figures (8)

• Figure 1: Frequency of head and tail items appeared in the training set and recommendation lists of two well-trained models (i.e., MF and LightGCN) for MovieLens-1M dataset. Head items are the top 10% of items with the most user interactions in the training set, while tail items are the remaining.
• Figure 2: Causal graphs for existing methods (a-b) and our PPAC framework (c-d).
• Figure 3: Analyzing personal popularity on MovieLens-1M.
• Figure 4: Causal graphs of the effect of alcohol consumption on lung cancer, where A, C, and L stand for alcohol, cigarette, and lung cancer. Gray nodes mean that the variables are at reference values (e.g., $A = a^*$).
• Figure 5: Causal graphs of GP and PP aware recommendation in factual and counterfactual worlds. In plot (b-c), X is a proxy variable that models the combination of GP and PP.

Theorems & Definitions (4)

• Definition 1: Global Popularity
• Definition 2: User Similarity
• Definition 3: Similar User Set
• Definition 4: Personal Popularity