Social Movements & Media

Studies of social movements have traditionally relied on ethnographic methods to learn about the motives and attitudes of participants. These methods are powerful but they also have certain limitations. For instance, they generally sample only active participants and are therefore unable to draw objective comparisons with non-participants. Also, by construction participants can only be observed or interviewed after the movement has begun; hence objective comparisons with preexisting beliefs and attitudes are difficult. So then here is the question: Can we use social media data to fill this gap? Can we use these non-representative noisy data to perform causal inference in the context of social movements? We claim the answer is yes. And here is a short explanation as to why this is the case: 1) the "always on" nature of Twitter allows researchers to construct ex-post panels even years after the events of interest have taken place, and this is of particular importance in the case of political uprisings that are hard to foresee. 2) They allow for easy sampling of the "right" set of non-participants to performs comparisons to. 3) The amount of information available for the participants (and non-participants) is impressive and goes well beyond what can be extracted with a short survey or interview. 4) Information extracted goes beyond individuals and captures organizational structure and network of participants.

These advantages motivate a number of research questions we work on. For instance, in our recent work we introduced a new methodology to use social media data to understand how a movement affects participants that take part in them (in the context of 2013 Gezi uprising). We are currently working on identifying stickiness of activism (Does participating in one protest increase your chances of participating in future protests? Does that vary across different movements?) and the most (and least) successful recruitment methods. We are also working on new collective action methods that incorporate the new findings from our data-driven studies.

Social Networks

The power of online social networks lies in their ability to enable the diffusion of information, ideas, and innovations. This significance motivates our work that focuses on understanding the diffusion process, and using this understanding to build technological solutions to facilitate and guide such spread. Our efforts that resulted in advances of the state-of-the-art in various areas listed below:

Understanding Diffusion: While the number of studies on modeling information diffusion is ever increasing, most research focuses on the influence of friends and assumes that users are either unaware of or not interested in activity happening outside their immediate neighborhood. Our work challenges this unrealistic assumption. Our approach produced a cross-models, cross-networks and cross-metrics evaluation framework which takes a strong first step towards attaining reproducibility and testing generalizability in modeling social behavior.

Managing Diffusion: The open nature of online social networks leaves their users vulnerable to the spread of misinformation with possibly devastating implications. Clearly, it is vital to investigate effective methods to limit the spread of misinformation to avoid such societal effects. Our work was the first study to focus on optimally managing the spread of misinformation in online social networks.

Reporting Diffusion: Information trends in online social networks reveal societal needs, fears or interests. They can also help users stay on top of news without having to sift through vast amounts of shared information. We introduced a number of methods ([1], [2]) that improve the state-of-the-art in this area.

Group Dynamics: In Bowling Alone: The Collapse and Revival of American Community, Robert D. Putnam discusses how we are fast becoming disconnected from family, friends, and our democratic structures. Online groups present a possible remedy here. But most online groups are short-lived given their inability to sustain members. So, what makes people stick? Can we predict it? For more, go here.

News Media Production & Consumption

News media are a key source of information for society. How is this source of information produced and how is it consumed? Are news outlets biased? Do readers prefer ideologically biased content?

In a recent work, we examined issue filtering and ideological framing in U.S. news media. Through the use of supervised learning and crowdsourcing techniques, we constructed a representative sample of politically relevant news coverage with ideological position broken down by issue. Our technique not only provides an unbiased ordering of outlets according to their political slant, but also discovers how ideological filtering and framing contribute to the overall ideological position. We are currently extending this work to examine how ideological bias on news outlets change over time in comparison to public opinion polls.

We are also working on another project that focuses on news forums to determine their value and shortcomings. When news forums first came to being, scholars hoped that they would create an era of reader engagement and deliberation. Those hopes have only been partially fulfilled. Many news forums today are known for incivility and an aggressively adversarial tone. That drives out participants who would otherwise engage in constructive discussion, and can adversely affect readers' understanding of the news. This has led a number of news organizations in the US to heavily regulate their commenting sites or shut them down entirely. But before we go there, we need to understand the value of these forums and if/how we can improve them. In this ongoing project, we are aiming to do exactly that.

Charitable Giving

How do people make charitable giving decisions? Scholars had to rely on small surveys or experiments to answer this question. Here, we ask the question: can we do better by making use of data that is publicly available online (such as charity evaluation sites and donor reviews) as well as web browsing data that reveals how donors navigate the online charity space?

Take for instance, the following question: Do donors take the efficiency of organizations they contribute to into consideration when making charitable giving decisions? The answer to this question is mostly unknown. A simple analysis of tax return forms of charitable organizations reveal that there is a non-negligible amount of money being donated to charities with extremely high overhead (more than 50% of donations being "wasted" on expenses unrelated to charity cause). This suggests that donors do not consider efficiency of charities as much as one would hope and that there is a sizable amount of waste in the charity marketplace. But to what extend is this true? Are people really not taking the efficiency of charities into consideration or are they stuck with the high-overhead charities because there are no reasonable alternatives for that particular cause with a lower overhead? And how much money could donors collectively save if charities were to go through a recommendation system that identified and presented reasonable alternatives in an algorithmic way? For more, read here.

Let's think of another question: Why do people care about the causes that they do? Does exposure increase interest? We are currently working on this problem in the context of poverty and homelessness. Stay tuned for our results...