[Paper Review] A Personality-Guided Preference Aggregator for Ephemeral Group Recommendation

Guangze Ye, Wen Wu+4 more|arXiv (Cornell University)|Apr 18, 2023

Recommender Systems and Techniques4 citations

TL;DR

This paper proposes PEGA, a personality-guided preference aggregator for ephemeral group recommendation that leverages users' personality traits to improve preference aggregation and alleviate data sparsity. By modeling group personality via hyper-rectangles and using a personality attention mechanism, PEGA outperforms state-of-the-art methods on Amazon and Yelp datasets, achieving significant gains in Recall and NDCG.

ABSTRACT

Ephemeral group recommendation (EGR) aims to suggest items for a group of users who come together for the first time. Existing work typically consider individual preferences as the sole factor in aggregating group preferences. However, they neglect to take into account the importance of the individual inherent factors, such as personality, and thus fail to accurately simulate the group decision-making process. Additionally, these methods often struggle due to insufficient interactive records. To tackle these issues, a Personality-Guided Preference Aggregator (PEGA) is proposed, which guides the preference aggregation of group members based on their personalities, rather than relying solely on their preferences. Specifically, implicit personalities are first extracted from user reviews. Hyper-rectangles are then used to aggregate individual personalities to obtain the "Group Personality", which allows for the learning of personality distributions within the group. Subsequently, a personality attention mechanism is employed to aggregate group preferences, and a preference-based fine-tuning module is used to balance the weights of personality and preferences. The role of personality in this approach is twofold: (1) To estimate the importance of individual users in a group and provide explainability; (2) To alleviate the data sparsity issue encountered in ephemeral groups. Experimental results demonstrate that, on four real-world datasets, the PEGA model significantly outperforms related baseline models in terms of classification accuracy and interpretability. Moreover, empirical evidence supports the idea that personality plays a pivotal role in enhancing the performance of EGR tasks.

Motivation & Objective

• To address the challenge of aggregating individual preferences in ephemeral groups with few interactions and high data sparsity.
• To incorporate individual personality traits as a key factor in estimating user influence during group decision-making.
• To improve recommendation accuracy and explainability by modeling group-level personality through hyper-rectangles.
• To reduce reliance on user interaction history by leveraging stable personality features that are less prone to sparsity.

Proposed method

• Using the LIWC lexicon to extract implicit personality traits from user review texts based on linguistic features.
• Defining Group Personality as a hyper-rectangle in the personality space, representing the fusion of individual personalities in a group.
• Employing a personality attention mechanism that assigns dynamic weights to users based on their personality traits, reflecting their influence in group decisions.
• Balancing personality and preference contributions via a learnable hyperparameter λ to optimize aggregation.
• Training the model end-to-end using a joint loss function that combines ranking and personality-aware preference learning.
• Integrating personality embeddings into the preference aggregation module to guide group-level decision-making.

Experimental results

Research questions

• RQ1 How does incorporating personality traits improve preference aggregation in ephemeral groups with sparse interactions?
• RQ2 To what extent can personality-based weighting enhance the explainability and accuracy of group recommendations?
• RQ3 Can a hyper-rectangle-based representation of group personality effectively model collective personality and reduce data sparsity?
• RQ4 How robust is the proposed method across varying group sizes and random group compositions?
• RQ5 What is the relative contribution of personality versus preference in determining user influence within ephemeral groups?

Key findings

• PEGA significantly outperforms state-of-the-art methods on both Amazon and Yelp datasets, achieving higher Recall and NDCG scores.
• The optimal balance between personality and preference is achieved at λ = 0.3, indicating personality has a stronger influence than preference in group decisions.
• The model demonstrates strong robustness across different group sizes, with stable performance even in large groups (≥12 members), where other models degrade.
• Case studies confirm that users with high openness, extraversion, and agreeableness are assigned higher influence weights, aligning with real-world group dynamics.
• Personality-based weighting effectively mitigates preference noise in diverse groups, especially when individual preferences are highly varied.
• The personality attention mechanism provides explainability by visually linking user personality traits to their influence in group decisions.