December 18, 2018

Announcing the winners of the Facebook Computational Social Science Methodology research awards

By: Lada Adamic

The scale and complexity of phenomena that can be observed online present new opportunities and challenges in social science research. Novel computational advances yield interesting insights about social phenomena, testing prior understanding and generating new approaches and methods. This in turn can help improve products that facilitate people’s online and offline interactions, giving people the power to build community and bring the world closer together. Facebook is interested in supporting and engaging with the academic community in these and other areas.

Today we are pleased to announce the winning proposals of our Computational Social Science Methodology RFP:

Detecting Fake News on Social Networks
Michael Bronstein, Universita della Svizzera italiana

Validating Theory for Quantifying Peer Influence in Observational Data
Edward McFowland III and Gordon Burtch, University of Minnesota—Twin Cities

Scaling Unstructured Human Insights to Improve Foster Care in Wisconsin
Shion Guha, Marquette University

These proposals aim to advance, in a variety of ways, computational social science approaches across a range of areas. No Facebook data will be provided to award recipients and research will be conducted independent of Facebook.

The ability of billions of people to connect online has created great potential not only for communication, commerce, community and learning, but also for harmful phenomena, such as the spread of misinformation. An important part of addressing this problem is detecting potential misinformation so that it can be surfaced to fact checkers. “Detecting fake news of social networks” will apply graphical deep learning to detect fake news not by its linguistic content, but rather by the manner in which it spreads across a social network. If successful, it will not only improve the state of the art in fake news detection, but also advance the state of the art of deep learning on graphs.

Another interesting challenge is understanding the causality of observed patterns. How do we know if two people have taken an action — for example, donating to a charity — because of a common interest or because one’s action influenced the other’s? This is a difficult question to answer using observational data alone. However, recent theoretical results have suggested that under certain conditions it should be possible. “Validating Theory for Quantifying Peer Influence in Observational Data” will pair theory, simulation and experiment to get to the bottom of quantifying social influence.

While computational social science is typically associated with large quantities of data on recorded actions, the third award goes to a proposal to develop a novel approach combining quantitative and grounded theory methods applied to a collection of human-recorded observations. “Scaling Unstructured Human Insights to Improve Foster Care in Wisconsin” will address an important opportunity to take advantage of twenty years of caseworker observations, which so far have not been utilized in developing quantitative methods in this space. Topic modeling will be used to derive themes in caseworker evaluations associated with positive placement outcomes. A grounded theory approach will develop features within themes which will be used in an overall model. The work has the potential to improve placement of children within the foster care system. Facebook will not be interfacing with any of the data.

Thank you to all the researchers that submitted proposals, and congratulations to the winners. To view our currently open research awards and to subscribe to our Research Awards email list, visit our Research Awards page.