Biases in Information Ecosystem

Countering COVID-19 Misinformation
As the COVID-19 pandemic spreads, countries and cities around the globe have taken stringent measures including quarantine and regional lockdown. The increasing isolation, along with the panic and anxiety, creates challenges for countering misinformation -- people are increasingly tapping into online information sources already familiar to them with declining chances of accessing alternative stories. This project will develop mechanisms based on text and image analysis, social psychology, and crowd-sourcing that can be used to counter misinformation. We aim to contribute to the scientific understanding of misinformation in terms of which false information is most influential, who is most affected by it and how to "debunk" the problematic information automatically in social media. The work is supported by an NSF award.

Understanding Group Biases
The recent proliferation of digital human trace data and machine learning techniques together provide opportunities for revolutionizing the ways of understanding group biases. There have been works that apply techniques, ranging from text and network mining to deep learning, to study group biases; however, these existing works are mostly limited in the capacity of adding new qualitative understanding about a group, or lacking a rigorous, reproducible analysis procedure to verify and interpret the newly revealed patterns. In other words, the trade-off between qualitative, "thick" data and quantitative, "big" data has not been addressed. The goal of this project is to develop capabilities to capture cultural models that reflect different group biases at new speeds and scales. To this end, we propose to develop an automatic group bias analytic framework with two specific aims: (1) purposefully incorporate data and machine biases to characterize group biases, (2) allow for scalable and reproducible comparative analysis across groups. This project will have implications into building trust, and avoiding conflicts and misunderstandings within and across groups. The work is supported by the DARPA Understanding Group Biases (UGB) program

Disaster Analytics

Collective Sense Making of Terrorist Attacks
Social media has become central to the public's response to terrorism. From the transmission of breaking news, to the offering of social support, to the dissemination of radical, hateful messages, people increasingly turn to social media to both share and gain understandings of terrorist events. This project utilizes the social media data to investigate the reactions of individuals located in Paris during the November 2015 attacks. This research outcome will both improve responses to specific terrorist attacks as well as enhance public understanding of the specific means through which terrorism wields social influence. See the project page for more details. The work is in part supported by an NSF award.

Threat Perception Following Mass Violence Events
How does experiencing mass violence, terrorism, or other traumatic events shape individuals' perceive and respond to their social world? Anecdotally, following extensive media coverage of mass violence events, many report perceiving objects, people, and situations as particularly threatening; and, as media coverage shifts to emphasize resilience and community cohesion. In this project, we empirically study how emotionally potent media coverage of a real-world threat alters threat perception. This work could reveal potential harmful real-world consequences of emotionally potent media reporting of a terrorism event, and will also help characterize the types of individuals who are at greatest risk of altered threat perception after a mass violence or terrorism event. See the project page for more details. The work is in part supported by an NSF award.

Other Projects

Diffusion Analytics for Public Policy Research
Stability in social, technical, and financial systems, as well as the capacity of organizations to work across borders, requires consistency in public policy across jurisdictions. Policy diffusion has been a topic of focus across the social sciences for several decades, but due to limitations of data and computational technology, researchers have not taken a comprehensive look at patterns of diffusion across many different policies. In this project, we combine cutting-edge methods of text and network analysis to understand how policies, as represented in digitized text, spread through networks connecting the American states. The project results will help researchers, public officials, advocacy groups, and other relevant stakeholders understand how to balance the competing goals of innovation and consistency in public policy. See the project page for more details. The work is in part supported by an NSF award.

Pitt Smart Living: Building a Smart City Ecosystem
Urban planners have begun to realize that a truly sustainable transportation and urban environment in general, requires a shift to multimodal transportation. This project is a pilot study to evaluate the benefits of making real­time transportation information available to city­dwellers and also the potential impact of incentives as a way to encourage pro-­social behavior. In this project, we will develop, deploy, and evaluate techniques that will integrate information for increasing the utilization and quality of public transportation. See the project website for more details. The work is in part supported by an NSF award.

Collective Behavior During Emergencies and Other Exogenous Events
Large-scale emergencies, including natural disasters, and terrorist attacks, are now among the largest threats to national security. There is an indisputably increasing need for new tools to strengthen disaster resilience at all levels of society. We have worked on issues and challenges relating to emergency understanding using large-scale data generated by hundreds of millions of users. Related resesarch venue: 2014 KDD Workshop on Learning about Emergencies from Social Information (KDD-LESI 2014). Related projects: rising stars (PLoS ONE), big bird (ICWSM), voice of victory (WWW), ripple (EPJ Data Science).

Scholarly Big Data
Academics and researchers worldwide continue to produce large numbers of scholarly documents including papers, books, technical reports, etc. and associated data such as tutorials, proposals, lab note books, and course materials. The abundance of data sources enables researchers to study scholarly collaboration at a very large scale. We have taken the lead on developing research methods and techniques for scholarly knowledge discovery. Related research venues: International Workshop on Collaborative Big Data (C-Big 2014), International Workshop on Challenges & Issues on Scholarly Big Data Discovery and Collaboration (SBD 2014). Related projects: ContexTour (SDM), progressive group mining (forthcoming), Twitter in academia (HyperText).

Urban and Local Community Analytics
In everyday urban life, individuals interact with each other not only through face-to-face meetings, but increasingly through mobile communication devices and Internet-based activities like emails and social media (e.g., Facebook and Twitter). Those types of technologically mediated communications leave fine-grained "digital traces" about when, where, and what one talks about and to whom, offering an unprecedented opportunity to examine the structure and dynamics of social and information behavior. We have taken the initiative to examine the social-spatial phenomena in a data-driven manner, and to connect the socio-spatial big data with the profound neighborhood effect observed in the social science literature. Related projects: habitat (PLoS ONE), sentiment segregation (SocialCom), sustainable local forums (forthcoming: HICSS). The work is in part supported by University of Pittsburgh Central Research Development Fund (CRDF): "Assessing the 'Neighborhood Effect' With Social Media Traces" (PI: Yu-Ru Lin).

Visual Data Mining
Data visualization has been increasingly important in this "big data" world. In our school, we have worked on new visualization techniques that enable different stakeholders to synthesize information and derive insight from massive, dynamic, multi-sourced, multi-faced, interrelated and sometimes conflicting data, and provide timely assessment for making decisions. With the explosion of different types of data generated from multiple sources, ranging from research groups, government agencies to participatory sensing data collected through social media and mobile devices, it is important to provide intelligent interface for all stakeholders in order to support actionable information seeking, particularly for supporting collective sense-making, situational awareness, and dissemination of credible crisis-relevant information. Related projects: Whisper (TVCG), FacetAtlas (TVCG), ContexTour (SDM), FluxFlow (TVCG), UnTangle (ICDM), SocialHelix (JoV).

Health Informatics via Crowdsourcing and Social Media
With the recent advances in Internet-based communication technologies, online social sites have become a prevailing cyberspace for people to seek health information or related supports. Our research uniquely leverages crowdsourced content analysis and machine learning techniques in developing new tools for users to express complex health-related questions or needs by including more personalized or contextual informations beyond simple queries. Related projects: health-related question intent characterization (forthcoming), health message diffusion (forthcoming).

Areas
Data Science, Network Science, Computational Social Science, Data Visualization

Methods
network metrics, graph analysis, data mining (esp. graph mining), visualization and visual analytics

Application domains
Collective Behavior, Scholarly Big Data, Urban and Local Community Analytics, Visual Data Mining, Health Informatics, Emergency Response

» Related publications

» More on earlier research highlights.

» Research opportunity:  
for Ph.D. applicants, and for master and undergraduate students.