Abstract
Social recommendation models weave social interactions into their design to provide uniquely personalized recommendation results for users. However, social networks not only amplify the popularity bias in recommendation models, resulting in more frequent recommendation of hot items and fewer long-tail items, but also include a substantial amount of redundant information that is essentially meaningless for the model's performance. Existing social recommendation models fail to address the issues of popularity bias and the redundancy of social information, as they directly characterize social influence across the entire social network without making targeted adjustments. In this paper, we propose a Condition-Guided Social Recommendation Model (named CGSoRec) to mitigate the model's popularity bias by denoising the social network and adjusting the weights of user's social preferences. More specifically, CGSoRec first includes a Condition-Guided Social Denoising Model (CSD) to remove redundant social relations in the social network for capturing users' social preferences with items more precisely. Then, CGSoRec calculates users' social preferences based on denoised social network and adjusts the weights in users' social preferences to make them can counteract the popularity bias present in the recommendation model. At last, CGSoRec includes a Condition-Guided Diffusion Recommendation Model (CGD) to introduce the adjusted social preferences as conditions to control the recommendation results for a debiased direction. Comprehensive experiments on three real-world datasets demonstrate the effectiveness of our proposed method.
Introduction: The Popularity Bias Challenge
In recommendation systems, "social recommendation" leverages the theory of social homogeneity—the observation that friends often share similar tastes. While this helps capture collaborative signals, it introduces two significant challenges: popularity bias amplification and information redundancy.
Social networks often act as echo chambers for "hot" items. When models incorporate social data without adjustment, they tend to recommend popular items even more frequently, effectively burying niche or "long-tail" items. Furthermore, not every social connection is a meaningful indicator of shared taste; many links are "noisy" or redundant, providing little value for accurate modeling. As shown in Figure 1, standard social recommendation models like DiffNet significantly increase the recommendation frequency of popular items compared to versions without social data, despite the majority of items belonging to the long-tail.
The CGSoRec Framework
To address these issues, we propose the Condition-Guided Social Recommendation Model (CGSoRec). The framework consists of two primary stages designed to clean social signals and steer recommendations toward a debiased distribution.
Condition-Guided Social Denoising (CSD)
Instead of accepting the social graph as ground truth, CSD treats social relations as potentially noisy data. It employs a diffusion-based denoising process to filter out redundant links. Crucially, this process is guided by "co-purchase" relationships of long-tail items. The underlying logic is that if two users share an interest in the same obscure item, their social link is a much stronger signal of shared taste than if they both interacted with a mainstream blockbuster.
Condition-Guided Diffusion Recommendation (CGD)
This stage performs item recommendation using a Diffusion Model applied to user-item interaction vectors. To combat popularity bias, we introduce a weight adjustment function. We calculate a social preference matrix and then invert the weights: we decrease the influence of popular items favored by a user's social circle and increase the influence of long-tail items. This "nudges" the recommendation toward a debiased direction during the iterative denoising process.
Experimental Evaluation
We evaluated CGSoRec on three real-world datasets: LastFM, DBook, and Ciao. To ensure a fair evaluation of bias, we utilized a debiased testing set where interactions are sampled to form a uniform distribution over items.
Key Results
- Superior Performance: CGSoRec consistently outperformed state-of-the-art baselines like LightGCN and MHCN. On LastFM, CGSoRec improved Recall@5 by over 20% compared to non-social variants.
- Long-Tail Success: While other models often see performance on long-tail items drop when social data is added, CGSoRec saw massive gains. For example, on LastFM's "tail" item group, it achieved a 25.75% improvement in Recall@5.
- Ablation Insights: Removing either the Social Denoising (CSD) or the Weight Adjustment (WA) component led to performance drops, confirming that both cleaning the network and steering preferences are essential.
Conclusion
CGSoRec demonstrates that social networks do not have to be a source of bias. By actively denoising social links based on niche interests and strategically reweighting preferences during the diffusion process, we can achieve highly personalized recommendations that respect the long-tail. Future work will explore optimizing the computational efficiency of these diffusion models for real-time production environments and incorporating multi-modal data to further refine the denoising process.
Summary updated