Social Network Analysis for International Relations
Network analysis has brought new perspectives to studying emergent structures in different contexts, e.g., international relations. Prior work on international relations networks has mostly relied on data derived from formal alliances and trade flows, which barely captures the rapidly evolving state of international relations due to globalization and recent advances in information technology. In this work , I proposed a novel research design that aims at capturing “real-time” international relations through news reporting data (~1 million). I collected worldwide news on a daily basis, and characterized the relationship between any two countries by analyzing the similarity to their news content. I presented the empirical results based on news about the APEC (Asia-Pacific Economic Cooperation) CEO Summit in 2013, revealed interesting and meaningful international relations among member countries.
Figure 1: the networks of country-to-theme, most countries mentioned the trade and growth issues.
Social Recommender Systems for Local Businesses
Figure 2: Location-based recommendation in Pittsburgh.
User Intelligent Interfaces and Social Recommenders
Put the User in Control
A social recommender system aims to provide useful suggestions to the user and prevent the social overload problem. Most of the research efforts address pushing high relevant items to the top of the ranked list, using a weight ensemble approach. However, I argued that the “learned” static fusion is insufficient for different specific contexts. In this project [3, 4, 8], I developed a series of visual recommendation components and a control panel for the user to interact with the recommendation results at an academic conference. The system (shown in Figure 3) offers better recommendation transparency and user-driven fusion of recommended sources. The experiment result shows that the user did fuse the different recommended sources and exploration patterns among tasks. The post-study survey is positively associated with the effectiveness of the system and quality of the explanation function. This finding shed light on future research questions about the design of a recommender system with human intervention and the interface beyond the static ranked list.
Figure 3: Relevance Tune - a visual interface with user-driven control function and meaningful visual encoding.
Break the Filter Bubble
Increasing diversity in the output of a recommender system is an active research question for solving a long-tail issue. Most of the current approaches have focused on ranked list optimization to improve recommendation diversity. However, little is known about the effect that a visual interface can have on this issue. This project [3, 4, 7] shows that a multidimensional visualization (shown in Figure 4) promotes diversity of social exploration in the context of an academic conference. My study showed a significant difference in the exploration pattern between the ranked list and visual interfaces. The results showed that a visual interface could help the user explore a more diverse set of recommended items.
Figure 4: Scatter Viz - a dual social recommender interface, which includes a ranked list and visual scatter plot components.
Controllable and Explainable Artificial Intelligence
Hybrid social recommender systems use social relevance from multiple sources to recommend relevant items or people to a user. To make hybrid recommendations more transparent and controllable, several researchers have explored interactive hybrid recommender interfaces, which allow for a user-driven fusion of recommendation sources. In this field of work [2, 5], advanced visualization has been explored as an approach to increase transparency and enhance the user experience. In this project, I attempted to further increase the transparency of a recommendation process by augmenting an interactive hybrid recommend interface with several types of explanations.
I evaluated the behavior patterns and subjective feedback by a within-subject study of three academic conferences. Results from the evaluation show the effectiveness of the proposed explanation models. The post-treatment survey indicates a significant improvement in the perception of explainability, but such improvement comes with a lower degree of controllability. This finding shed light on future research questions about how to design human-scale explanations in social recommenders.
Figure 5: A subset of visualizations for explaining the social recommendations.
The beyond-relevance objectives of recommender systems have been drawing more and more attention. For example, a diversity-enhanced interface has been shown to associate positively with overall levels of user satisfaction. However, little is known about how users adopt diversity-enhanced interfaces to accomplish various real-world tasks. In this project [3, 4], I presented two attempts at creating a visual diversity-enhanced interface (Figure 3 & 4) that presents recommendations beyond a simple ranked list. My goal was to design a recommender system interface to help users explore the different relevance prospects of recommended items in parallel and to stress their diversity. Two within-subject user studies in the context of social recommendation at academic conferences were conducted to compare our visual interfaces. Results from our user study show that the visual interfaces significantly reduced the exploration efforts required for given tasks and helped users to perceive the recommendation diversity. I showed that the users examined a more diverse set of recommended items while experiencing an improvement in overall user satisfaction. Also, the users' subjective evaluations show significant improvement in many user-centric metrics. Experiences will be discussed that shed light on avenues for future interface designs.
Figure 6: the user experience effects on proposed interfaces through a structural equation model (SEM) analysis.