Bridging Expertise: Doctor Recommendations for Cross-Disciplinary Collaborations in Online Medical ConsultationsOnline Medical Consultations (OMCs) are increasingly prevalent in Online Healthcare Communities (OHCs), with a considerable focus on doctor recommendations. However, existing research has predominantly centered on recommending individual specialized doctors, neglecting the exploration of more reliable interdisciplinary options for collaborative consultations. This paper introduces an innovative framework for interdisciplinary doctor recommendations for collaborative work in OMC scenarios, comprising two modules: one for calculating expertise and another for collaborative computing. Leveraging the medical knowledge graph, the former employs an expertise encoder to derive specialty embeddings for patient queries, doctor profiles, and historical consultations. The latter builds a collaborative circle for each doctor based on three dimensions: scientific citation, academic collaboration, and spatiotemporal proximity. This circle analysis delves into the potential for cooperation between doctors, aiming to identify optimal combinations. Experimental results on two real datasets demonstrate the superior performance of our proposed model compared to state-of-the-art methods. Ablation studies highlight the significant contribution of the three social network dimensions and their interactions to collaborative computing outcomes. Our model exhibits robustness across different parameter settings, as demonstrated through experiments. The study's results empower practitioners to develop a portfolio of recommendations, enhancing the effectiveness of cooperative consultations for complex diseases requiring multidisciplinary collaboration.2025XWXiaonan Wu et al.HealthCSCW
Understanding Voter Fraud Misinformation Videos during the 2024 Taiwan Election on YouTubeAlthough voter fraud misinformation in the U.S. has been extensively studied, fewer research studies have investigated voter fraud misinformation in Taiwan’s elections. In this study, we present a mixed-method analysis of video-based voter fraud misinformation during the 2024 Taiwan election on YouTube. We develop a computational pipeline to identify video accounts with different political leanings that may produce election-related misinformation, collect the videos uploaded by these accounts, and automatically determine which videos are related to ballots with the assistance of Large Language Models. Subsequently, we manually identify voter fraud discussion videos using a codebook we developed. Following this, we conduct a comprehensive analysis of the identified videos. We find that video accounts aligned with the Democratic Progressive Party (DPP) produce the highest number of videos discussing voter fraud misinformation. Additionally, videos discussing such misinformation tend to receive more comments but fewer likes compared to videos without this content. We also observe that videos associated with the DPP are quite distinct from those linked to the Kuomintang (KMT) or Taiwan People’s Party (TPP), with unique characteristics that may be further revealed through their audio and video features. Finally, we conduct case studies to examine different patterns in videos either supporting or refuting voter fraud misinformation. Among accounts promoting voter fraud claims, traditional media outlets often include misinformation in their news programs, implying tacit endorsement, while grassroots media may present suspicious vote-counting scenes as evidence to spread misinformation. In contrast, traditional and grassroots media accounts associated with the DPP tend to refute misinformation through news programs or influencer commentary. Our work sheds light on the discourse surrounding voter fraud in Taiwan and offers valuable insights into strategies for mitigating the spread of voter fraud misinformation videos globally.2025YLYanheng Li et al.Combating Misinformation in Elections and Around the WorldCSCW
Libra: An Interaction Model for Data VisualizationWhile existing visualization libraries enable the reuse, extension, and combination of static visualizations, achieving the same for interactions remains nearly impossible. We contribute an interaction model and its implementation to achieve this goal. Our model enables the creation of interactions that support direct manipulation, enforce software modularity by clearly separating visualizations from interactions, and ensure compatibility with existing visualization systems. Interaction management is achieved through an instrument that receives events from the view, dispatches these events to graphical layers containing objects, and then triggers actions. We present a JavaScript prototype implementation of our model called Libra.js, enabling the specification of interactions for visualizations created by different libraries. We demonstrate the effectiveness of Libra by describing and generating a wide range of existing interaction techniques. We evaluate Libra.js through diverse examples, a metric-based notation comparison, and a performance benchmark analysis.2025YZYue Zhao et al.School of Computer Science and Technology, Shandong UniversityInteractive Data VisualizationTime-Series & Network Graph VisualizationCHI
Seeing Through the Overlap: The Impact of Color and Opacity on Depth Order Perception in VisualizationSemi-transparent visualizations are commonly used to reveal information in overlapped regions by applying colors and opacity. While a few studies made recommendations on how to choose colors and opacity levels to maintain depth perception, they often conflict and overlook the interaction effect between these factors. In this paper, we systematically explore the impact of color and opacity on depth order perception across eight colors, three opacity levels, and various layer orders and arrangements. Our inferential analysis shows that both color hue and opacity significantly influence depth order perception, with the effectiveness depending on their interaction. We also derived 12 features for predictive analysis, achieving the best mean accuracy of 80.72% and mean F1 score of 87.75%, with opacity assigned to the front layer as the top feature for most models. Finally, we provide a small design tool and four guidelines to better align the design rules of colors and opacity in semi-transparent visualizations.2025ZMZhiyuan Meng et al.Shandong UniversityInteractive Data VisualizationUncertainty VisualizationVisualization Perception & CognitionCHI
Accurate Insights, Trustworthy Interactions: Designing a Collaborative AI-Human Multi-Agent System with Knowledge Graph for Diagnosis PredictionHealthcare question-answering (QA) systems can assist physicians in making medical decisions. However, traditional medical QA systems struggle with multi-agents interaction and domain-specific knowledge processing, thereby reducing the accuracy and credibility of clinical decision-making. We thus develop a multi-agent decision-making system by combining a fine-tuned medical model, biomedical knowledge graphs, and PubMed data. By summarizing the symptoms described by users, our system can automatically convene clinical experts from various fields, retrieve domain knowledge, and provide clinical decision support for users. We have validated the system performance using both technical and user-centric approaches in terms of information accuracy, user satisfaction, user trust, ect. We thus provide an effective tool for healthcare professionals to make accurate and timely decisions. Furthermore, this study also reveals new design and research opportunities, including (1) optimizing multi-agent collaboration mechanisms for more complex medical decision-making, (2) improving interaction design to enhance system transparency and explainability, and (3) expanding the system to support a broader range of medical issues and multimodal data.2025HLHaoran Li et al.Renmin University of ChinaBrain-Computer Interface (BCI) & NeurofeedbackExplainable AI (XAI)AI-Assisted Decision-Making & AutomationCHI
I lose vs. I earn: Consumer perceived price fairness toward algorithmic (vs. human) price discriminationMany companies are turning to algorithms to determine prices. However, little research has been done to investigate consumers’ perceived price fairness when price discrimination is implemented by either a human or an algorithm. The results of two experiments with 2 (price-setting agent: algorithm vs. human) × 2 (price discrimination: advantaged vs. disadvantaged) between-subjects design reveal that consumers perceive disadvantaged price discrimination as being more unfair when it is implemented by a human (vs. algorithm). Conversely, they perceive advantaged price discrimination as being more unfair when it is implemented by an algorithm (vs. human). This difference is caused by distinct attribution processes. Consumers are more likely to externalize disadvantaged price discrimination implemented by a human than an algorithm (i.e., attributing it to the unintentionality of price-setting agents), while they are more likely to internalize advantaged price discrimination implemented by a human than an algorithm (i.e., attributing it to perceived personal luck). Based on these findings, we discuss how designers and managers can design and utilize algorithms to implement price discrimination that reduces consumer perception of price unfairness. We believe that reasonable disclosure of algorithmic clues to consumers can maximize the benefits of price discrimination strategies.2024XZXiaoping Zhang et al.Renmin University of ChinaAI Ethics, Fairness & AccountabilityPrivacy by Design & User ControlAlgorithmic Fairness & BiasCHI
Understanding How Low Vision People Read using Eye TrackingWhile being able to read with screen magnifiers, low vision people have slow and unpleasant reading experiences. Eye tracking has the potential to improve their experience by recognizing fine-grained gaze behaviors and providing more targeted enhancements. To inspire gaze-based low vision technology, we investigate the suitable method to collect low vision users' gaze data via commercial eye trackers and thoroughly explore their challenges in reading based on their gaze behaviors. With an improved calibration interface, we collected the gaze data of 20 low vision participants and 20 sighted controls who performed reading tasks on a computer screen; low vision participants were also asked to read with different screen magnifiers. We found that, with an accessible calibration interface and data collection method, commercial eye trackers can collect gaze data of comparable quality from low vision and sighted people. Our study identified low vision people’s unique gaze patterns during reading, building upon which, we propose design implications for gaze-based low vision technology.2023RWRu Wang et al.University of Wisconsin-MadisonEye Tracking & Gaze InteractionVisual Impairment Technologies (Screen Readers, Tactile Graphics, Braille)CHI
Being a Solo Endeavor or Team Worker in Crowdsourcing Contests? It is a Long-term Decision You Need to MakeWorkers in crowdsourcing are evolving from one-off, independent micro-workers to on-demand collaborators with a long-term orientation. They were expected to collaborate as transient teams to solve more complex, non-trivial tasks. However, collaboration as a team may not be as prevalent as possible, given the lack of support for synchronous collaboration and the "competition, collaboration but transient" nature of crowdsourcing. Aiming at unfolding how individuals collaborate as a transient team and how such teamwork can affect an individual's long-term success, this study investigates the individuals' collaborations on Kaggle, a crowdsourcing contest platform for data analysis. The analysis reveals a growing trend of collaborating as a transient team, which is influenced by contest designs like complexity and reward. However, compared with working independently, the surplus of teamwork in a contest varies over time. Furthermore, the teamwork experience is beneficial for individuals in the short term and long term. Our study distinguishes the team-related intellectual capital and solo-related intellectual capital, and finds a path dependency effect for the individual to work solely or collectively. These findings allow us to contribute insights into the collaborative strategies for crowd workers, contest designers, and platform operators like Kaggle.2022KHKeman Huang et al.Crowdsourcing; CrowdsourcingCSCW
Evaluating the Effects of Saccade Types and Directions on Eye Pointing TasksWith the portable and affordable gaze input devices being marketed for end users, gaze-based interactions were getting increasingly popular. Unfortunately, the understanding about the dominant task of gaze input, i.e. eye pointing task, was still not sufficient although a performance model had been specifically proposed in previous study because of that 1) the original model was based on a specific circular target condition, without the ability to predict the performance of acquiring conventional rectangular targets and that 2) there was a lack of explanation from the perspective of the anatomical structure of the eyes. In this paper, we proposed a 2D extension to take account of more general target conditions. Carrying out two experiments, we evaluated the effectiveness of the new model and furthermore we found that the index of difficulty that we redefined for 2D eye pointing (IDeye) was able to properly reflect the asymmetrical impacts of target width and height, and consequently the IDeye model could more accurately and properly predict the performance when acquiring 2D targets than Fitts' law, no matter what kind of saccades or eye orientations (i.e. saccadic eye movement directions) was employed to acquire the desired targets. According to the results, we provided more useful implications and recommendations for gaze-based applications.2021XZXinyong ZhangEye Tracking & Gaze InteractionUIST