Over the past few months I’ve had the good fortune to represent The Sentinel Project at a few conferences in the San Francisco Bay area, where I’m based. The conferences were a diverse array of learning and networking opportunities, as each focused on a different theme relevant to the Sentinel Project’s work. I attended the Strata conference in Santa Clara (big data visualization and analysis), the Amnesty International-USA Annual General Meeting in San Francisco (human rights and activism) and the Advancing the New Machine conference in Berkeley (human rights and technology).
Taken as a whole, a few themes and lessons emerged from the conferences that have been beneficial to the Sentinel Project as we continue developing our approach to early warning. These varied from lessons about research methodology to data visualization, and from technologies that we can use internally for data aggregation and collaboration to those that could be deployed in the field for information sharing and data collection. I’ll elaborate on some of these themes below.
- The rules of dataviz: The Sentinel Project’s Threatwiki platform looks to marry new design techniques with existing visualization technologies and apply them to a unique conceptual framework focused on the various processes that indicate a coming or existing genocide. Several speakers at the Strata conference, including Kim Rees of Periscopic, Simon Rogers of The Guardian’s Data blog and Jacques MacKinlay of Tableau, emphasized a few general rules of data visualization that we hope to incorporate into our platform. Among these are that the visualizations should be dynamic and interactive, they should be engaging but simple and parsimonious, and that users should have access to all of the data, with the ability to use filters to drill down and see only the information retaining the characteristics they specify. These components are already present in Threatwiki, and we have ideas in the works for its next release to make the platform even more robust. (For more on the Strata conference, see my guest posts on the Information Aesthetics blog.)
- Analysts AND algorithms > analysts OR algorithms: When Bob McGrew of Palantir Technologies spoke at Strata, he praised the utility of automated analysis but emphasized the superiority of a human’s ability to contextualize information. And while humans are typically more adept at data analysis, computers can be put to work to perform the more rote, tedious tasks, like massive data collection. After a presentation he gave at the human rights and tech conference in Berkeley, we contacted Dr. Michael Best of Georgia Tech to talk with him about some software he mentioned that aggregates relevant event data from various sources, including social media and traditional news, and funnels it into a single stream for easy extraction, which can be automated or done manually. The software is called Aggie, developed by Tom Smyth, a Georgia Tech PhD student, and we are happy to say that the Sentinel Project will be beta testing some of this software in support of our situation monitoring efforts. We hope Aggie will help us to streamline the research and analysis process, and will be giving Tom feedback on what is sure to be a very useful and important tool as he continues its development.
Similarly, Scott Edwards of Amnesty International, speaking at the Berkeley conference on human rights and tech, emphasized a balance between people and machines when he identified three key elements to a comprehensive approach to social prediction: structural (macro-level data for macro-level prediction), single-process time series (finer data for shorter time scales) and game theoretic (as he put it, human beings are agents, not billiard balls). Each of these levels is a process the Sentinel Project is either currently employing or actively developing, as our risk assessments represent structural-level analysis, our process monitoring tools (e.g. Threatwiki) utilize time-series event data, and the methodology we are developing for vulnerability assessments and short-term conflict forecasting will likely incorporate game theoretic approaches.
- Do not reinvent the wheel: As we have seen with the uprisings in North Africa, the Middle East and elsewhere, people are very resourceful when it comes to finding ways to gather and distribute information. There is no need to create new technologies when existing ones will do. At the Berkeley conference, both Olga Werby and Sasha Costanza-Chock, among others, emphasized the benefits of technological appropriation; that is, using technology to achieve your ends, even if that end is not what the product’s designers had in mind. We see this particularly with the use of mobile phones and SMS in documenting abuses, mobilizing otherwise disparate movements and planning mass gatherings. Patrick Meier spoke at both the tech conference and the Amnesty AGM about the ways in which he and his colleagues at Ushahidi collected information via SMS and used crowdsourcing to plot incidents of interpersonal attacks in Kenya, and locations of aid stations and shelters in Haiti and Japan. Much of the information that Ushahidi collected and published would not otherwise have been documented anywhere. Their approach to information collection and presentation was an early inspiration to the Sentinel Project and will continue to be as we work to establish information-sharing networks in the countries we’re monitoring.
There is far more that deserves coverage from these fantastic meetings of the minds, so please don’t mistake lack of inclusion here for lack of importance. These conferences were full of dedicated people with very novel, interesting and inspiring ideas about the uses of technology. We look forward to running into them again at future events, but for now we’re hard at work finding ways to synthesize those ideas and apply them to genocide prevention. Many thanks to all of the participants, and do keep up the good work. We can’t wait to see what you come up with next–and can’t wait to share our own innovations as well.