A/HRC/38/35 increasing pressure to address disinformation spread through links to bogus third-party news articles or websites, fake accounts, deceptive advertisements and the manipulation of search rankings. 96 However, because blunt forms of action, such as website blocking or specific removals, risk serious interference with freedom of expression, companies should carefully craft any policies dealing with disinformation. 97 Companies have adopted a variety of responses, including arrangements with third-party fact checkers, heightened enforcement of advertisement policies, enhanced monitoring of suspicious accounts, changes in content curation and search ranking algorithms, and user trainings on identifying false information. 98 Some measures, particularly those that enhance restrictions on news content, may threaten independent and alternative news sources or satirical content. 99 Government authorities have taken positions that may reflect outsized expectations about technology’s power to solve such problems alone.100 Company moderation processes and tools 32. Automated flagging, removal and pre-publication filtering. The massive scale of user-generated content has led the largest companies to develop automated moderation tools. Automation has been employed primarily to flag content for human review, and sometimes to remove it. Automated tools scanning music and video for copyright infringement at the point of upload have raised concerns of overblocking, and calls to expand upload filtering to terrorist-related and other areas of content threaten to establish comprehensive and disproportionate regimes of pre-publication censorship.101 33. Automation may provide value for companies assessing huge volumes of usergenerated content, with tools ranging from keyword filters and spam detection to hashmatching algorithms and natural language processing.102 Hash matching is widely used to identify child sexual abuse images, but its application to “extremist” content — which typically requires assessment of context — is difficult to accomplish without clear rules regarding “extremism” or human evaluation.103 The same is true with natural language processing.104 34. User and trusted flagging. User flags give individuals the ability to log complaints of inappropriate content with content moderators. Flags typically do not enable nuanced discussions about appropriate boundaries (e.g., why content may be offensive but, on balance, better left up).105 They have also been “gamed” to heighten pressure on platforms to remove content supportive of sexual minorities and Muslims. 106 Many companies have developed specialized rosters of “trusted” flaggers, typically experts, high-impact users and, reportedly, sometimes government flaggers. 107 There is little or no public information 96 97 98 99 100 101 102 103 104 105 106 107 12 Ibid.; Allen Babajanian and Christine Wendel, “#FakeNews: innocuous or intolerable?”, Wilton Park report 1542, April 2017. Joint Declaration 2017. Submissions by Association for Progressive Communications, pp. 4–6 and ARTICLE 19, p. 4. Association for Progressive Communications submission, p. 5. See communication No. OL ITA 1/2018. Cf. European Commission, A Multi-Dimensional Approach to Disinformation: Final Report of the Independent High-level Group on Fake News and Disinformation (Luxembourg, 2018). The United Kingdom of Great Britain and Northern Ireland reportedly developed a tool to automatically detect and remove terrorist content at the point of upload. Home Office, “New technology revealed to help fight terrorist content online”, 13 February 2018. Center for Democracy and Technology, Mixed Messages? The Limits of Automated Media Content Analysis (November 2017), p. 9. Open Technology Institute submission, p. 2. Center for Democracy and Technology, Mixed Messages?, p. 4. On user flags, see generally Kate Crawford and Tarleton Gillespie, “What is a flag for? Social media reporting tools and the vocabulary of complaint”, New Media and Society, vol. 18, No. 3 (March 2016), pp. 410−428. Ibid., p. 421. YouTube Help, YouTube Trusted Flagger Program; YouTube Help, “Get involved with YouTube contributors”.