Building Metrics for Success When Combating Extremism Online
Blog Post
Shutterstock.com
May 24, 2018
In January, representatives from the nation’s most influential tech companies convened on Capitol Hill to discuss methods for combating the publication of violent and extremist content on their respective platforms. Leaders of Companies such as Facebook, Twitter, and Youtube highlighted their work in Countering Violence and Extremism (CVE) - a term that encompasses campaigns meant to counter extremist rhetoric - online. During the hearing aptly titled “#IsBigTechDoingEnough,” high-level business officials were quick to shine a spotlight on various statistics as metrics for success, citing an increase in workers to scan extremist content, as well as the large amount of extremist content their companies have taken down. But do these numbers truly illustrate the effectiveness of Silicon Valley’s campaign to combat the presence of such content online?
During New America’s Millennial Public Policy Symposium, Millennial Fellow with the Open Technology Institute Spandi Singh hosted a discussion with New America International Security Fellow Evanna Hu that explored the role of data in enhancing the field of CVE, as well as the validity of content takedown statistics that are often referred to by tech companies as metrics for success. Both Singh and Hu began the conversation by establishing that it is extremely difficult to equate actions pertaining to CVE as a direct cause for the disruption of extremist activity. For this reason, legislators are eager to consume data - such as click through rates, shares, and likes - that may seem impressive on the surface, but do not necessarily represent progress in countering extremist narratives.
The discussion also drew attention to the disconnect between the threshold for success established by legislators on Capitol Hill and their understanding of the technologies companies use to engage in CVE. Last month, Facebook co-founder and CEO Mark Zuckerberg attended a heavily publicized congressional hearing following the company's data-breach scandal. While the focus of the hearing revolved around Facebook’s capabilities to protect the data of it users, Zuckerberg was also questioned about the company’s practices regarding the censorship of violent and extremist content. Facebook’s CEO proudly illustrated the company’s capabilities to address such content by highlighting the fact that 99% of ISIS and Al Qaida content taken down from the platform is identified by Artificial Intelligence (A.I).
However, research conducted by Omelas - an advisory firm co-founded by Hu that maps the online information environment for security assessments - revealed that large swaths of extremist content can go undetected for a significant span of time, even under the watchful eye of A.I. programs. Unfortunately, legislators have traditionally found a false sense of comfort in surface level metrics provided by tech companies, stifling their ability to truly measure the success of CVE campaigns. What's more, the technical illiteracy of legislators at the highest level of government drives the creation of insufficient indicators of success in the CVE space.
Singh and Hu also explored the consequences of tech companies operating unproven CVE programs. Not only do unproven CVE methods have the potential to further marginalize legitimate voices through censorship, they can also further radicalize target audiences. Additionally, CVE programs tend to emphasize content takedown of larger extremist organizations, such ISIS and Al Qaida, creating an avenue for smaller organizations to increase their presence online.
The discussion closed by identifying steps that can be taken to strengthen the field of CVE. While tech companies are moving in the right direction by providing transparency reporting on actions associated with CVE, a neutral party functioning as an independent auditor has the potential to illustrate the state of online information environments beyond non-descriptive statistics. Lastly, Hu highlighted the potential of civil society organizations in developing strong metrics for success in the CVE space, by educating organizations to become literate in CVE terminology and breaking the current threshold established by legislators lacking sufficient knowledge of the field.
This blog is part of Caffeinated Commentary - a monthly series where the Millennial Fellows create interesting and engaging content around a theme. For May, the fellows are recapping some of the many important conversations from their April Millennial Public Policy Symposium: New Voices and Ideas on Care, Community, Technology, and Civic Engagement.