Is Social Media a Threat to Democracy? ISSUES: Focusing on Six Key Risks Issue Three: Conflation of popularity, legitimacy and user intentionality Over the last two decades, technology companies have instances of algorithmic bias—is that the criteria spent a huge amount of money and effort to develop ways attribute legitimacy to popularity; thereby flooding the to get people to trust each other online, in conversations public with multiple, competing, unverified assertions.21 and transactions, on various platforms and marketplaces. This isn’t just restricted to Facebook and Twitter; a Social media takes this to the next level—doubling down variant of the problem exists at Google, where “auto-fill on the age-old locus of trust, reputation, and belief in search terms” assume user intentionality and conflate one’s networks. Perhaps it is no surprise that today, a this with interest. For instance, in 2013, UN Women majority of global respondents to the Edelman Trust launched a powerful ad campaign revealing Google’s Survey claim to believe individuals over institutions. autocomplete suggestions for “Women shouldn’t…” However, one of the arenas in which this has serious “Women cannot…” and “Women need to…” among consequences is on platforms where all individuals others. The top results included “Women shouldn’t have can publish without meaningful editorial insight, and rights,” “Women cannot drive,” and “Women need to be where polarization has led to echo chambers. Crowd- put in their place.” The algorithms for those phrases have sourced discussion platforms, including ones such as since been updated, and there are certain terms that Wikipedia, Quora, and Reddit, further blur the lines Google will not autocomplete, including “Bisexuals are…” between specialists and the layperson, creating false and “Lesbians are…” Yet plenty of other examples of equivalencies. In the U.S., the crowd-sourced information bigotry, sexism, and racism lurk within other seemingly phenomenon is now tied into part of a larger narrative innocuous searches. Such auto-fill search terms cannot and growing backlash against experts and elites, who only reinforce prejudices; but when used to analyze are viewed having a self-serving agenda. user preferences and behavior, they can also reveal fascinating and hard-to-prove insights. Crucial to how users consume information is the algorithmic logic of certain social media platforms and Finally, attributing legitimacy to popular search queries the way they engineer viral sharing in the interest of is worrying in a world where it can reinforce —or even their business models. The non-neutral algorithms of leverage—assumptions that code is unbiased. For Facebook and Twitter actively use selection criteria to instance, the Edelman Trust Barometer revealed that enhance the visibility of certain information. What’s 59% of global respondents are prone to believe a search highly problematic about this—apart from documented engine over a human editor.22,23 It may be worth questioning whether this is due to the editorial choices of the platforms, or whether it is a result of platform design responding to reader preferences and prejudice. 22 Edelman Trust Barometer, 2017 21 © Copyright The Omidyar Group 2017. All rights reserved. 23  or, conversely, can individuals trust supposedly human bylines N anymore, as automated content increases on the Internet, and raises questions about transparency, transferred trust and ethics. PAGE 9

Select target paragraph3