M-scan: A Multi-Scenario Causal-driven Adaptive Network for Recommendation

TL;DR

M-scan tackles multi-scenario recommendation by explicitly modeling how the scenario S directly influences click behavior and user interests. It introduces two core modules: Scenario-Aware Co-Attention to extract cross-scenario interests aligned with the current scenario, and Scenario Bias Eliminator to mitigate biases from other scenarios via causal counterfactual inference. The model trains with a dual loss capturing M→Y and S→Y effects and performs a counterfactual adjustment during inference to produce unbiased, scenario-specific predictions. Empirical results on two public datasets show superior performance over strong baselines, with ablations confirming the necessity of both modules and analyses highlighting robust hyperparameter behavior. The work advances causal reasoning in multi-scenario recommendations and suggests directions for online deployment and scale-up to larger models.

Abstract

We primarily focus on the field of multi-scenario recommendation, which poses a significant challenge in effectively leveraging data from different scenarios to enhance predictions in scenarios with limited data. Current mainstream efforts mainly center around innovative model network architectures, with the aim of enabling the network to implicitly acquire knowledge from diverse scenarios. However, the uncertainty of implicit learning in networks arises from the absence of explicit modeling, leading to not only difficulty in training but also incomplete user representation and suboptimal performance. Furthermore, through causal graph analysis, we have discovered that the scenario itself directly influences click behavior, yet existing approaches directly incorporate data from other scenarios during the training of the current scenario, leading to prediction biases when they directly utilize click behaviors from other scenarios to train models. To address these problems, we propose the Multi-Scenario Causal-driven Adaptive Network M-scan). This model incorporates a Scenario-Aware Co-Attention mechanism that explicitly extracts user interests from other scenarios that align with the current scenario. Additionally, it employs a Scenario Bias Eliminator module utilizing causal counterfactual inference to mitigate biases introduced by data from other scenarios. Extensive experiments on two public datasets demonstrate the efficacy of our M-scan compared to the existing baseline models.

Figures (6)

• Figure 1: Illustration of single and multi scenario situations. Left: the whole page as a single-scenario in Amazon. Medium: horizontal lists as multi-scenarios in Netflix. Right: vertical and horizontal lists as multi-scenarios in Google Play.
• Figure 2: Causal analysis of multi-scenario recommendation.
• Figure 3: Overall illustration of M-scan.
• Figure 4: Counterfactual causal graph of multi-scenario recommendation.
• Figure 5: Performance of M-scan using different $c$ values of Eq. \ref{['eq:inference']} on two datasets.