Peers

This page introduces some other groups that I think are doing interesting and important work.

Notable Projects:

• Computational Propaganda: The Computational Propaganda project is dedicated to understanding how technology is being used to influence people, with a focus on global politics. A lot of this has been studying how "fake news" and misinformation have spread and what is being done to limit their impact.

• Illegal Wildlife Trade: This project aims to understand how the illegal wildlife trade works and how to reduce demand for parts of endangered animals.

• Economic Geographies of the Darknet: The Internet is used for a variety of illegal and dangerous purposes and this project aims to visualize and understand them.

Notable Projects:

• Mobile Data For Sustainable Development & Humanitarian Action: This group partnered with MIT to determine how mobile data can be used to inform policy decisions and plan responses to disasters without compromising the privacy of users. This is a complex issue because on one hand the data needs to be clear enough to provide useful information, but on the other overly specific data can usually be "de-anonymized" using statistical methods or supplementary databases. If this data can be used properly, it can save lives and proper risk analysis minimizes potential downsides.

• Estimating Migration Flows Using Online Search Data: This study used search data on specific keywords to attempt to predict migration in Australia. By comparing international searches for jobs in specific regions to averages, it may be possible to estimate the flow of immigrants so that countries and economies are better able to prepare for these changes.

• Measuring Poverty with Machine Roof Counting: In Uganda, the type of roofing material a home has serves as a good proxy for economic well-being. Using satellite and UAV data, this group is working on a tool to provide more accurate data to the Ugandan government

Notable Projects:

• Adversarial Machine Learning: Studying how machine learning systems can be tricked by malicious input. Here is an example of the type of problem this project is trying to solve. Relying on systems vulnerable to these sorts of attacks can be worse than having no such system at all. The most effective of these attacks can change the classification to a new class with high certainty, potentially making the attack very difficult to detect.

• AI For Precision Mental Health: Mental health is obviously a complex issue and AI is often applied to healthcare. This project differs from most approaches because it focuses on monitoring changes in an individual to predict mental illness rather than classifying healthy vs control patients. Early prediction of these problems could really help individual quality of life as well as increase understanding of how these illnesses develop and change patients over time.

• Vote Trading Networks: This project uses Monte Carlo simulations and other statistical methods to estimate "vote trading" between different politicians.

Other Projects:

• ORBIS: A project run by Stanford which maps the road network of the Roman Empire, providing available routes with estimates of time and price. This can really help understanding of what it meant to travel in the Roman Empire and understand logistically what communication of ideas and movement of armies must have looked like.

• The Lazarus Project: In my opinion, one of the most exciting topics in the digital humanities is recovering damaged manuscripts and one of the most important is digitizing them. This project does both, using advanced forensic and computational techniques to restore and analyze manuscripts without damaging them. Similar work is done by this group at Oxford on papyrus, although I wonder if machine learning could be used to do things like translate papyri or determine authorship of texts.