A/HRC/44/57 electronic health records.65 Integrated directly into existing electronic health records, Epic’s artificial intelligence tool estimates the likelihood that a patient will no-show by using the patient’s personal information, including ethnicity, class, religion and body mass index, as well as the patient’s record of prior no-shows. In pointing out the obvious potential to discriminate against vulnerable patient populations, the researchers note that “removing sensitive personal characteristics from a model is an incomplete approach to removing bias.”66 Prior no-shows, for example, likely correlate with socioeconomic status mediated by the patient’s inability to cover transportation or childcare costs, or to take time off work for the appointment. They also likely correlate with race and ethnicity because of correlations between socioeconomic status, and race and ethnicity.67 It was revealed in another recent study that black patients were more likely to be scheduled into overbooked appointment slots and thus had to wait longer when they did show up.68 32. In the housing context, studies in the United States have shown ethnic discrimination in Facebook’s targeted advertising. Facebook used to allow advertisers to “narrow audience” by excluding Facebook users with certain “ethnic affinities” under the “demographics” category of its ad-targeting tool.69 This targeted advertising could be used to prevent black people from viewing specific housing advertisements, which is prohibited under United States anti-discrimination law. Facebook controls an estimated 22 per cent of the market share for digital advertisements in the United States, 70 and its targeted advertising, which is the core of the company’s business model,71 has been shown to be racially exclusionary.72 These practices are best understood as a form of digital redlining, which is defined as “the creation and maintenance of technology practices that further entrench discriminatory practices against already marginalized groups”. 73 Facebook uses targeted advertising in the employment context as well, raising similar concerns. 33. In yet other cases, access to technology – and the information available through it – are denied in ways that have disparate impacts, or that target specific racial, ethnic or religious groups, sometimes on a discriminatory basis. In 2019, multiple States, including Bangladesh, the Democratic Republic of the Congo, Egypt, India, Indonesia, Iran (Islamic Republic of), Myanmar, the Sudan and Zimbabwe, completely restricted Internet access to specific regions, with the effect of preventing nearly all communication in or out of those regions.74 Researchers have linked more targeted Internet shutdowns to regions with higher densities of minority groups. 75 34. With respect to the right to a fair trial, multiple courts in Latin America have begun using Prometea, a software that uses voice recognition and machine learning prediction, to streamline judicial proceedings. The district attorney’s office and courts in Buenos Aires use this artificial intelligence system to automate judicial decision-making in simple cases, such as disputes about taxi licences and complaints from teachers about not being compensated for school supplies.76 In such cases, Prometea interprets the facts given to it and suggests a legal outcome based on prior jurisprudence in similar cases. A judge must approve the decision before it is made official, which is the case 96 per cent of the time.77 A 65 66 67 68 69 70 71 72 73 74 75 76 77 10 See www.healthaffairs.org/do/10.1377/hblog20200128.626576/full. Ibid. See https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3467047. Ibid. See www.propublica.org/article/facebook-lets-advertisers-exclude-users-by-race. See www.emarketer.com/content/us-digital-ad-spending-will-surpass-traditional-in-2019. See www.motherjones.com/politics/2019/12/facebook-agreed-not-to-let-its-ads-discriminate-butthey-still-can. See www.propublica.org/article/facebook-advertising-discrimination-housing-race-sex-nationalorigin. See www.congress.gov/116/meeting/house/110251/witnesses/HHRG-116-BA00-Wstate-GillardC20191121.pdf. See www.hrw.org/news/2019/12/19/shutting-down-internet-shut-critics. See www.accessnow.org/cms/assets/uploads/2020/02/KeepItOn-2019-report-1.pdf. See www.bloombergquint.com/businessweek/this-ai-startup-generates-legal-papers-without-lawyersand-suggests-a-ruling. See www.giswatch.org/2019-artificial-intelligence-human-rights-social-justice-and-development.