1 00:00:00,370 --> 00:00:05,510 MICHELLE: In 2017, The Australian obtained an internal Facebook report. 2 00:00:05,510 --> 00:00:08,290 Reportedly it explained how advertisers could use their 3 00:00:08,290 --> 00:00:13,710 ad platform together psychological insights based on Facebook activity. 4 00:00:13,710 --> 00:00:18,110 The report which was commissioned by one of Australia's top banks postulated 5 00:00:18,110 --> 00:00:23,230 that advertisers could target millions of teenagers with ads when they feel insecure 6 00:00:23,230 --> 00:00:24,860 and worthless. 7 00:00:24,860 --> 00:00:26,802 This is called predictive analytics. 8 00:00:30,505 --> 00:00:35,214 Predictive analytics uses statistical techniques to analyze current and 9 00:00:35,214 --> 00:00:41,080 historical facts to make predictions about future or otherwise unknown events. 10 00:00:41,080 --> 00:00:44,090 This brings up the complexity of data consent. 11 00:00:44,090 --> 00:00:48,710 Data we're willing to share can imply conclusions we don't want to share. 12 00:00:48,710 --> 00:00:53,860 Yes, users willingly post information about their personal life on Facebook. 13 00:00:53,860 --> 00:00:58,040 But that does not give Facebook explicit consent to interpret that data. 14 00:00:58,040 --> 00:01:00,480 Draw conclusions on their mental health and 15 00:01:00,480 --> 00:01:05,510 then sell that health assumption to advertisers to sell more goods. 16 00:01:05,510 --> 00:01:10,415 Further, an inaccurate diagnosis could harm someone's well-being. 17 00:01:10,415 --> 00:01:13,893 In 2012, a New York Times article revealed Target, 18 00:01:13,893 --> 00:01:19,220 a popular retailer in the United States, predicted a woman's pregnancy. 19 00:01:19,220 --> 00:01:22,500 A man complained about Target sending coupons for baby clothes and 20 00:01:22,500 --> 00:01:24,340 cribs to his daughter. 21 00:01:24,340 --> 00:01:27,224 He thought they had been encouraging her to become pregnant. 22 00:01:27,224 --> 00:01:32,430 Although they actually had accurately predicted she was already pregnant. 23 00:01:32,430 --> 00:01:37,302 Target assigns each shopper a unique number called the guest ID. 24 00:01:37,302 --> 00:01:42,416 Attached to that guest ID is a detailed profile that contains everything 25 00:01:42,416 --> 00:01:47,875 they've ever bought at Target and their behaviors such as using a coupon, 26 00:01:47,875 --> 00:01:51,096 filling out a survey, or opening an email. 27 00:01:51,096 --> 00:01:56,160 HOPE: Per the New York Times, here's a sampling of data Target may have on file. 28 00:01:56,160 --> 00:02:02,502 Target declined to confirm the information it collects or purchases. 29 00:02:02,502 --> 00:02:05,938 Age, whether you are married and have kids, 30 00:02:05,938 --> 00:02:11,554 which part of town you live in, how long it takes you to drive to the store. 31 00:02:11,554 --> 00:02:15,859 Your estimated salary, whether you've moved recently, 32 00:02:15,859 --> 00:02:19,052 what credit cards you carry in your wallet. 33 00:02:19,052 --> 00:02:26,830 What websites you visit, ethnicity, job history, the magazines you read. 34 00:02:26,830 --> 00:02:31,039 If you've ever declared bankruptcy or got divorced, 35 00:02:31,039 --> 00:02:34,065 the year you bought or lost your house. 36 00:02:34,065 --> 00:02:39,330 Where you went to college, what kinds of topics you talk about online. 37 00:02:39,330 --> 00:02:44,259 Whether you prefer certain brands of coffee, paper towels, 38 00:02:44,259 --> 00:02:46,354 cereal, or applesauce. 39 00:02:46,354 --> 00:02:49,387 Your political leanings, reading habits, 40 00:02:49,387 --> 00:02:52,670 charitable giving, the number of cars you own. 41 00:02:54,420 --> 00:02:56,270 I don't mean to pick on Target. 42 00:02:56,270 --> 00:02:58,670 This is typical for retailers of this size. 43 00:02:59,925 --> 00:03:02,740 Target's guest marketing analytics department 44 00:03:02,740 --> 00:03:04,880 reviews this data to make predictions. 45 00:03:06,960 --> 00:03:11,320 The pregnancy prediction score enables target to send coupons and 46 00:03:11,320 --> 00:03:13,810 ads timed to the pregnancy stages. 47 00:03:14,870 --> 00:03:18,740 This is bound to lead to uncomfortable circumstances. 48 00:03:18,740 --> 00:03:23,200 Somebody sent a box of baby formula to my parents' house in my name 49 00:03:23,200 --> 00:03:26,920 when I was younger, and my parents grilled me. 50 00:03:26,920 --> 00:03:30,230 They thought I might be pregnant and hiding it from them. 51 00:03:30,230 --> 00:03:34,854 I wasn't pregnant, but looking back on it now, I must have inadvertently set off 52 00:03:34,854 --> 00:03:37,810 a trigger to lead a marketer to think I was pregnant. 53 00:03:39,340 --> 00:03:43,590 Pervasive social sorting enables companies to identify, 54 00:03:43,590 --> 00:03:45,870 categorize, and rate customers. 55 00:03:47,090 --> 00:03:50,420 This affects a person's available choices and 56 00:03:50,420 --> 00:03:53,980 even results in paying different prices for the same goods. 57 00:03:55,120 --> 00:04:00,850 According to a study of 5 top auto insurers in 15 US cities, 58 00:04:00,850 --> 00:04:05,380 the Consumer Federation of America found that good drivers of lower 59 00:04:05,380 --> 00:04:09,390 economic status consistently pay significantly more for 60 00:04:09,390 --> 00:04:13,010 auto insurance than higher economic status drivers. 61 00:04:14,600 --> 00:04:20,148 Four out of five of the nation's largest auto insurers regularly charged 40% to 62 00:04:20,148 --> 00:04:28,090 92% more to lower income drivers, even when they have perfect driving records. 63 00:04:28,090 --> 00:04:34,479 Further, the insurance company Allstate analyzes consumer shopping habits. 64 00:04:34,479 --> 00:04:39,319 Per CFA, Allstate charges a higher rate to customers not identified 65 00:04:39,319 --> 00:04:41,730 as bargain shoppers. 66 00:04:41,730 --> 00:04:45,040 The consumer's location is another factor. 67 00:04:45,040 --> 00:04:49,272 If other insurers offer competitive rates in the area, 68 00:04:49,272 --> 00:04:51,801 Allstate provides a lower rate. 69 00:04:51,801 --> 00:04:55,175 CFA cautions against unfair price optimization. 70 00:04:56,260 --> 00:05:00,340 Traditional rating factors such as driving safety record, or 71 00:05:00,340 --> 00:05:03,320 the average number of miles driven are more appropriate. 72 00:05:05,590 --> 00:05:10,290 MICHELLE: Advocates of data mining speak about improving the user experience 73 00:05:10,290 --> 00:05:14,913 through personalizing experiences to tailor to a user's needs. 74 00:05:14,913 --> 00:05:20,460 But how much personal information are you willing to give up for that convenience? 75 00:05:20,460 --> 00:05:24,290 And what about when data is given with one consent in mind and 76 00:05:24,290 --> 00:05:28,010 then later it is used in another context? 77 00:05:28,010 --> 00:05:31,820 When it comes to data privacy, the ethical line is difficult to draw. 78 00:05:33,020 --> 00:05:37,950 Increasingly, behavioral data is making automated decisions on groups of people. 79 00:05:39,200 --> 00:05:41,580 It's one thing to see an irrelevant ad. 80 00:05:41,580 --> 00:05:46,831 It's another level to be discriminated against in crucial areas such as finance, 81 00:05:46,831 --> 00:05:48,790 insurance, and healthcare. 82 00:05:48,790 --> 00:05:54,832 And data driven persuasion systematically influences people's behavior, 83 00:05:54,832 --> 00:05:57,218 even in political elections. 84 00:05:57,218 --> 00:06:00,478 Being mindful of how data is used and abused. 85 00:06:00,478 --> 00:06:05,431 I'm interested in more privacy conscious tools coming out, such as Matomo, and 86 00:06:05,431 --> 00:06:07,597 alternative to Google Analytics. 87 00:06:07,597 --> 00:06:12,568 Matomo analytics is an open source tool that keeps data on your server instead 88 00:06:12,568 --> 00:06:15,470 of sharing it with third parties. 89 00:06:15,470 --> 00:06:17,330 For more tool recommendations, 90 00:06:17,330 --> 00:06:20,750 check out the Ethical Design Handbook linked in the teacher's notes.