A General Neural Causal Model for Interactive Recommendation

TL;DR

This paper introduces a neural causal model that addresses survivor bias in recommender systems by enabling counterfactual inference, improving long-term recommendation quality through causal reasoning and reinforcement learning.

Contribution

It proposes a novel neural causal framework with structural causal modeling and counterfactual consistency for interactive recommendation systems.

Findings

The model effectively mitigates survivor bias in recommendation scenarios.

The approach demonstrates superior performance in empirical evaluations.

Theoretical analysis confirms the validity of the causal inference method.

Abstract

Survivor bias in observational data leads the optimization of recommender systems towards local optima. Currently most solutions re-mines existing human-system collaboration patterns to maximize longer-term satisfaction by reinforcement learning. However, from the causal perspective, mitigating survivor effects requires answering a counterfactual problem, which is generally unidentifiable and inestimable. In this work, we propose a neural causal model to achieve counterfactual inference. Specifically, we first build a learnable structural causal model based on its available graphical representations which qualitatively characterizes the preference transitions. Mitigation of the survivor bias is achieved though counterfactual consistency. To identify the consistency, we use the Gumbel-max function as structural constrains. To estimate the consistency, we apply reinforcement optimizations, and use Gumbel-Softmax as a trade-off to get a differentiable function. Both theoretical and empirical studies demonstrate the effectiveness of our solution.