Online shoppers often face decision overload when confronted with a vast number of similar-looking products, especially in high-stakes categories like electronics. To make informed choices, users typically shortlist a few items and examine product specifications and user reviews to identify key trade-offs. This process is time-consuming and cognitively demanding. Product listings typically highlight basic attributes like price and ratings, but often fail to disclose potential drawbacks or trade-offs early in the decision process. In this paper, we propose to integrate balanced explanations, brief summaries of pros and cons, directly into the product listing page. By providing balanced explanations upfront, this design aims to help users more efficiently narrow down options and make confident, informed decisions without needing to explore individual product pages. We conducted a within-subject user study (N = 352) comparing baseline listings with a version that included pros and cons. Results show that users preferred the balanced explanation and reported greater confidence in their selections. NFC predicted lower reported influence from pros-and-cons notes, while CISB predicted higher preference for their availability.. Our findings offer practical insights for designing transparent, user-centered recommender systems that support effective early-stage decision-making in online shopping.
E-commerce platforms increasingly use persuasive design elements, such as urgency cues and editorial badges, to steer users toward specific products. While effective at driving engagement, these tactics often emphasize only the positive aspects of items, limiting users’ ability to make well-informed decisions. This paper investigates whether embedding concise warning messages about product drawbacks can counterbalance persuasive cues and support more transparent decision-making. We conducted a controlled user study (N = 352) comparing product selection behavior in a baseline condition (persuasive cues only) versus a warning condition (where some items also displayed critical disclosures). Our results show that warnings significantly influence product choices, even in the presence of persuasive messages. Moreover, individual differences, such as consumer information search behavior and general decision-making style, moderate sensitivity to warnings. Qualitative feedback reveals that users value the added transparency and find warnings helpful, even when they choose to ignore them. We argue that integrating warnings at the point of decision can foster more ethical and trustworthy experiences with e-commerce recommender systems.
This paper focuses on the inter-domain recommendation task in cross-domain recommendation, which aims to address the cold-start problem prevalent in traditional recommender systems. The primary concern is to map a source-domain user’s preference representation to the same latent space as target-domain items. Existing methods fail to strike a balance between generalization and customization during preference mapping, significantly limiting recommendation performance. In this paper, we propose a memory-augmented cluster-level preference mapping framework that captures users’ preference relationships between domains at the cluster level. We learn mapping functions for clusters to transform users’ representations from the source domain to the target domain. To store representations of clusters and mapping function parameters, we employ a key-value memory network. Through manual updates via Write operations after local adaptation on training overlapping users, both key vectors and value memory units of the memory network retain useful information throughout training. By attentively combining projected representations from all clusters using Read operations, we generate source-domain users’ representations in the target domain. Extensive experiments conducted on real-world datasets demonstrate our proposed model’s effectiveness compared with state-of-the-art methods.
This paper models the three-dimensional adjacent connections of brain regions based on resting-state functional magnetic resonance imaging, mainly using Brodmann’s Interactive Atlas to construct the brain network. From each fMRI brain slice, 80 independent Regions of Interest (ROIs) were identified and clustered into 8 functional groups. Train a graph attention network to classify Parkinson’s disease (PD) states. Reconstruct the brain map based on the similarity of regional relationships, and extract primitives with 4 and 5 motifs. The top 4
The widespread dissemination of false information can lead to serious harm, such as misleading public perception and severely undermining the credibility of mainstream media. Considering that traditional multimodal false information detection methods struggle to handle high-risk trending information and highly consistent text-image false information, this paper establishes a multimodal detection model based on a news corpus for verifying trending information, enabling early detection of text-image false information. Compared with traditional false information detection methods, this model introduces the concepts of a hot news time window and a high similarity spatial window, analyzing the authenticity of information from both popularity and novelty perspectives. It designs a data preprocessing pipeline including OCR text recognition and completion, image masking, etc., to improve dataset quality, and integrates text-image features to overcome the limitations of single-modal approaches. Experiments on the constructed dataset demonstrate that the proposed model outperforms existing baseline models in terms of performance metrics such as accuracy and precision.
