1 00:00:15,410 --> 00:00:23,650 Speaker 1: Pushkin. Chris McKinley popped open a second window on his 2 00:00:23,810 --> 00:00:30,450 Speaker 1: desktop computer and checked for messages from ok Cupid. Oh dear, 3 00:00:30,530 --> 00:00:36,010 Speaker 1: Oh dear, Yet again, his favorite online dating website was 4 00:00:36,170 --> 00:00:42,250 Speaker 1: offering tumbleweed. That didn't seem right. Chris was a catch. 5 00:00:42,890 --> 00:00:47,370 Speaker 1: He was smart, a degree in Chinese, a master's in mathematics, 6 00:00:47,890 --> 00:00:50,930 Speaker 1: and he was nearly through a PhD in data science, 7 00:00:51,690 --> 00:00:53,530 Speaker 1: which was why he was sitting in front of a 8 00:00:53,530 --> 00:00:57,010 Speaker 1: computer screen in a cramped cubicle and a deserted open 9 00:00:57,050 --> 00:01:02,410 Speaker 1: plan office at UCLA's Mathematics Sciences Building at three o'clock 10 00:01:02,450 --> 00:01:08,890 Speaker 1: in the morning. He was handsome, blue eyes, six feet tall, slim, rugged, 11 00:01:08,890 --> 00:01:15,290 Speaker 1: good looks, and tousled hair. He was interesting. He liked art, 12 00:01:16,050 --> 00:01:19,450 Speaker 1: loved music, had even spent a few years making a 13 00:01:19,490 --> 00:01:25,010 Speaker 1: living as a professional blackjack player. But now June twenty twelve, 14 00:01:25,570 --> 00:01:29,010 Speaker 1: he was in his early thirties, about to complete a 15 00:01:29,090 --> 00:01:33,450 Speaker 1: highly marketable PhD, and Chris was ready to settle down 16 00:01:33,610 --> 00:01:38,730 Speaker 1: and commit to a serious relationship. He was lovable. He 17 00:01:38,810 --> 00:01:43,330 Speaker 1: knew computers, so why couldn't the computer find him love? 18 00:01:44,890 --> 00:01:49,850 Speaker 1: With a sigh, Chris shut the ok Cupid window and 19 00:01:50,010 --> 00:01:53,570 Speaker 1: switched off the computer. Then he switched off the desk 20 00:01:53,650 --> 00:01:58,050 Speaker 1: lamp too, letting darkness wash over the fifth floor of 21 00:01:58,090 --> 00:02:02,890 Speaker 1: the UCLA building, just as he did every night. He 22 00:02:03,010 --> 00:02:07,530 Speaker 1: rolled out his foam sleeping mat under his desk, stretched 23 00:02:07,570 --> 00:02:14,010 Speaker 1: himself out and waited for sleep to come. I'm Tim Harford, 24 00:02:14,370 --> 00:02:41,530 Speaker 1: and you're listening to cautionary tales. The cover of Science 25 00:02:41,610 --> 00:02:45,930 Speaker 1: and Invention Magazine nineteen twenty four shows what is, at 26 00:02:46,050 --> 00:02:50,250 Speaker 1: first glance, nothing more than a cliched clinch between a 27 00:02:50,290 --> 00:02:56,210 Speaker 1: loving couple. He tall, muscular, clean cut with a striped tie, 28 00:02:56,290 --> 00:03:00,210 Speaker 1: blue waistcoat and crisp white shirt, sleeves rolled above the elbow. 29 00:03:01,050 --> 00:03:05,450 Speaker 1: She rosy cheeks, dark hair in a flapper cut red 30 00:03:05,530 --> 00:03:10,250 Speaker 1: jacket and fashionable red cap, wrapped in his arms and 31 00:03:10,330 --> 00:03:17,410 Speaker 1: gazing helplessly up at his chiseled features. And then you 32 00:03:17,570 --> 00:03:24,530 Speaker 1: notice the equipment. A wristband, armband, vacuum pumps and hoses, 33 00:03:25,290 --> 00:03:29,970 Speaker 1: spool after spool of ticker tape, glass tubes, little brass engines, 34 00:03:30,170 --> 00:03:36,890 Speaker 1: the works, all somehow attached to our dashing hero, presumably 35 00:03:36,970 --> 00:03:41,490 Speaker 1: to see what exactly her adoring gaze and soft embrace 36 00:03:41,690 --> 00:03:48,890 Speaker 1: is doing to his vital signs. Goodness me the promise 37 00:03:48,970 --> 00:03:52,570 Speaker 1: that the latest technology will find you love is not 38 00:03:52,650 --> 00:03:57,290 Speaker 1: a new one. Perhaps that's no surprise. Looking for love 39 00:03:57,370 --> 00:04:02,810 Speaker 1: can be frightening, disheartening, even tortuous. So of course there's 40 00:04:02,890 --> 00:04:06,250 Speaker 1: no shortage of people promising that they can make the 41 00:04:06,410 --> 00:04:11,570 Speaker 1: course of true love run smooth. But can they? Let's 42 00:04:11,570 --> 00:04:15,290 Speaker 1: find out what else Science and Invention Magazine nineteen twenty 43 00:04:15,370 --> 00:04:20,010 Speaker 1: four has to offer us. The magazine's publisher, Hugo Gernsback, 44 00:04:20,090 --> 00:04:23,770 Speaker 1: has written a feature article arguing that it's vital to 45 00:04:23,810 --> 00:04:28,730 Speaker 1: put science to work helping improve the quality of marital matches. 46 00:04:29,610 --> 00:04:34,730 Speaker 1: He describes four tests of compatibility. The first one uses 47 00:04:34,770 --> 00:04:39,730 Speaker 1: the equipment depicted on the cover. Gernsback describes his and 48 00:04:39,850 --> 00:04:45,970 Speaker 1: hers electrodes to allow an electrical shigmagraph to record their pulses. 49 00:04:46,890 --> 00:04:52,850 Speaker 1: A chest mounted chain measures breathing. Gernsback explains around the 50 00:04:52,970 --> 00:04:55,690 Speaker 1: chest of each is a chain which is secured to 51 00:04:55,730 --> 00:04:59,170 Speaker 1: a piece of spring covered by a rubber hose. One 52 00:04:59,250 --> 00:05:02,090 Speaker 1: end of the tube, thus formed, is sealed. The other 53 00:05:02,130 --> 00:05:04,930 Speaker 1: connects to a monometer and also to a tambore equipped 54 00:05:04,930 --> 00:05:07,490 Speaker 1: with a stylus. The stylus leaves a record on a 55 00:05:07,530 --> 00:05:12,410 Speaker 1: moving paper tape. The rate of respiration. It's like a 56 00:05:12,450 --> 00:05:17,410 Speaker 1: seismograph for the earthquake. That is love, because surely there 57 00:05:17,410 --> 00:05:21,570 Speaker 1: can be no better evidence of physical attraction than the 58 00:05:21,690 --> 00:05:26,210 Speaker 1: quickening of a pulse and the heaving of a bosom 59 00:05:26,330 --> 00:05:30,930 Speaker 1: and gurns back content. The single most important element for 60 00:05:30,970 --> 00:05:39,250 Speaker 1: a successful marriage is physical attraction. Okay, what else? There's 61 00:05:39,330 --> 00:05:42,730 Speaker 1: the sympathy test. This uses the same equipment as the 62 00:05:42,730 --> 00:05:46,370 Speaker 1: physical attraction test, but instead of canoodling with her bow, 63 00:05:47,090 --> 00:05:51,250 Speaker 1: the lady watches him go through an unpleasant experience, such 64 00:05:51,290 --> 00:05:56,370 Speaker 1: as having blood drawn again. If she becomes excited, short 65 00:05:56,370 --> 00:06:00,930 Speaker 1: of breath, tense muscles, then she's sympathetic to him, and 66 00:06:00,970 --> 00:06:04,530 Speaker 1: this is a sign of a good match. The position 67 00:06:04,690 --> 00:06:08,130 Speaker 1: should then be reversed, of course, to check that when 68 00:06:08,330 --> 00:06:12,410 Speaker 1: she is in this comfort, he is sufficiently sympathetic too. 69 00:06:13,930 --> 00:06:18,450 Speaker 1: It gets better. Let's talk about the body odor test. 70 00:06:19,450 --> 00:06:22,810 Speaker 1: One of the prospective couple is placed into an enormous 71 00:06:22,930 --> 00:06:27,570 Speaker 1: glass capsule with a hose emerging from it. The other person, 72 00:06:28,010 --> 00:06:32,050 Speaker 1: still attached to the electrical shigmograph and the other paraphernalia 73 00:06:32,130 --> 00:06:36,290 Speaker 1: to monitor their physical response, takes the attachment at the 74 00:06:36,330 --> 00:06:40,650 Speaker 1: end of the hose places it over their nose and 75 00:06:40,730 --> 00:06:45,330 Speaker 1: takes a big sniff. Maybe they catch a whiff of 76 00:06:45,650 --> 00:06:49,770 Speaker 1: armpit or bad breath. More likely, there's nothing to smell 77 00:06:49,810 --> 00:06:55,970 Speaker 1: but damp rubber hose. No matter, if the aroma doesn't 78 00:06:56,010 --> 00:07:00,130 Speaker 1: provoke an adverse physical reaction, that is supposed to be 79 00:07:00,330 --> 00:07:04,930 Speaker 1: a good sign. Gernsback a pines that in all probability, 80 00:07:05,370 --> 00:07:09,810 Speaker 1: more marriages are destroyed by body odors than any other reason, 81 00:07:10,410 --> 00:07:17,770 Speaker 1: and who can prove him wrong. The last scientific test 82 00:07:18,090 --> 00:07:23,650 Speaker 1: is simplicity itself. While a couple's vital signs are being recorded, 83 00:07:24,130 --> 00:07:27,530 Speaker 1: one of the experimental team walks into the room and 84 00:07:27,610 --> 00:07:30,890 Speaker 1: fires a gun into the air. If only one of 85 00:07:30,930 --> 00:07:34,690 Speaker 1: the couple shows physiological symptoms of being startled or upset, 86 00:07:34,770 --> 00:07:38,130 Speaker 1: that's fine. But if both of them panic, that's a 87 00:07:38,130 --> 00:07:42,290 Speaker 1: bad result. Someone has to keep cool in a stressful situation. 88 00:07:42,570 --> 00:07:50,690 Speaker 1: Gurns Back explains, so far, so ridiculous. But Gernsback's explanation 89 00:07:50,930 --> 00:07:56,570 Speaker 1: for all this nonsense seems strikingly contemporary. How much would 90 00:07:56,610 --> 00:08:01,010 Speaker 1: the average man or woman give to know beforehand if 91 00:08:01,050 --> 00:08:05,010 Speaker 1: his or her prospective married life is to be success 92 00:08:05,290 --> 00:08:09,690 Speaker 1: or failure, asks Gernsback, who was at the time three 93 00:08:09,770 --> 00:08:14,570 Speaker 1: years into his second of three marriages. At present, marriage 94 00:08:14,610 --> 00:08:19,010 Speaker 1: is a lottery. It seems impossible to predict beforehand how 95 00:08:19,050 --> 00:08:23,210 Speaker 1: your prospective mate will turn out in the future. These 96 00:08:23,290 --> 00:08:27,330 Speaker 1: days we might speak of relationships rather than marriages, but 97 00:08:27,370 --> 00:08:32,410 Speaker 1: the promise is familiar. Through the miracle of the latest technology, 98 00:08:32,730 --> 00:08:38,890 Speaker 1: the labyrinth of love can be easily navigated. And there's 99 00:08:38,890 --> 00:08:42,810 Speaker 1: a clue in the way he phrases the matter. How 100 00:08:42,930 --> 00:08:47,810 Speaker 1: much would the average man or woman give to know? Or, 101 00:08:47,890 --> 00:08:51,210 Speaker 1: to put it another way, how much could we charge 102 00:08:51,290 --> 00:08:57,050 Speaker 1: to tell them? Gurns back ads. Through certain fundamentals which 103 00:08:57,090 --> 00:09:00,730 Speaker 1: can easily be ascertained, one can be reasonably certain as 104 00:09:00,770 --> 00:09:04,770 Speaker 1: to one's choice. There are certain basic tests which can 105 00:09:04,810 --> 00:09:07,930 Speaker 1: be made today and which will give one a reasonable 106 00:09:07,970 --> 00:09:13,570 Speaker 1: assurance of married happiness. You won't find many takers today 107 00:09:13,810 --> 00:09:17,210 Speaker 1: for watching your partner having blood taken, or putting them 108 00:09:17,250 --> 00:09:20,330 Speaker 1: in a capsule to harvest their body odor, or for 109 00:09:20,450 --> 00:09:23,290 Speaker 1: checking their reaction to having a gun go off without warning. 110 00:09:24,210 --> 00:09:30,850 Speaker 1: These century old tests seem ridiculous, but will future generations 111 00:09:30,930 --> 00:09:35,410 Speaker 1: look back at today's multi billion dollar business of online 112 00:09:35,490 --> 00:09:43,610 Speaker 1: dating and conclude that the joke is on us? There 113 00:09:43,610 --> 00:09:47,810 Speaker 1: are two million women in Los Angeles and most of 114 00:09:47,850 --> 00:09:50,970 Speaker 1: them didn't know that Chris McKinley relied on the UCLA 115 00:09:51,130 --> 00:09:54,850 Speaker 1: gym for its shower facilities and slept in a cubicle, 116 00:09:56,050 --> 00:10:01,410 Speaker 1: so why couldn't he get a girlfriend? McKinley found that 117 00:10:01,690 --> 00:10:05,810 Speaker 1: ok cupid was only suggesting a match once every few days, 118 00:10:06,610 --> 00:10:09,050 Speaker 1: and that just wasn't a deep enough pool of people 119 00:10:09,250 --> 00:10:12,770 Speaker 1: to have much success. He'd look at the new matches 120 00:10:12,930 --> 00:10:17,290 Speaker 1: profile and think, hmmm, maybe not for me. Or sometimes 121 00:10:17,330 --> 00:10:20,850 Speaker 1: a girl would come along and he'd think she's perfect. 122 00:10:22,410 --> 00:10:27,210 Speaker 1: But then Chris explained, there's these many kind of emotional 123 00:10:27,330 --> 00:10:31,690 Speaker 1: cycles of like, you know, hope and disappointment, like thinking, 124 00:10:32,250 --> 00:10:34,650 Speaker 1: oh wow, this person seems really cool, and then of 125 00:10:34,650 --> 00:10:37,770 Speaker 1: course you know, you write them and they don't write 126 00:10:37,810 --> 00:10:43,530 Speaker 1: you back. We do know, Chris, we do know. But 127 00:10:43,770 --> 00:10:47,130 Speaker 1: Chris was a computer guy, and as a man who 128 00:10:47,170 --> 00:10:51,290 Speaker 1: once made money counting cards at the blackjack tables of Vegas, 129 00:10:51,450 --> 00:10:55,050 Speaker 1: he was perfectly comfortable with looking for a little edge, 130 00:10:56,250 --> 00:11:02,370 Speaker 1: and so he decided to hack okay cupid step one 131 00:11:03,570 --> 00:11:07,770 Speaker 1: figure out why the site wasn't delivering many matches. That 132 00:11:07,930 --> 00:11:11,650 Speaker 1: meant he'd have to revert engineer the algorithm, figure out 133 00:11:11,690 --> 00:11:15,770 Speaker 1: what variables it was using to make its recommendations. This 134 00:11:15,930 --> 00:11:20,210 Speaker 1: wasn't easy. Ok Cupid made its matching recommendations on the 135 00:11:20,250 --> 00:11:24,370 Speaker 1: basis of how people had answered various questions about themselves 136 00:11:24,730 --> 00:11:28,610 Speaker 1: and their preferences. But there were hundreds of questions, maybe 137 00:11:28,690 --> 00:11:32,650 Speaker 1: thousands of questions, anything from do you like the taste 138 00:11:32,690 --> 00:11:36,770 Speaker 1: of beer? To how important is religion or God in 139 00:11:36,810 --> 00:11:40,490 Speaker 1: your life? As well as an answer, people would also 140 00:11:40,610 --> 00:11:45,370 Speaker 1: record what answers they'd find acceptable from a mate and 141 00:11:45,450 --> 00:11:49,770 Speaker 1: how important the question was on a one to five scale. 142 00:11:49,770 --> 00:11:52,930 Speaker 1: Most people only answered a few of these questions, but 143 00:11:53,170 --> 00:11:57,170 Speaker 1: Chris wanted to master okay Cupid, so he had to 144 00:11:57,170 --> 00:12:00,890 Speaker 1: figure out which were the most important questions and which 145 00:12:01,090 --> 00:12:05,250 Speaker 1: were the most popular answers. And okay Cupid wasn't saying 146 00:12:06,730 --> 00:12:11,490 Speaker 1: except there was a little quirk on the site. If 147 00:12:11,530 --> 00:12:15,250 Speaker 1: your questions and your answers were a perfect match for someone, 148 00:12:15,690 --> 00:12:19,410 Speaker 1: you'd be shown their profile when you logged in. So 149 00:12:19,530 --> 00:12:23,850 Speaker 1: all Chris had to do was create thousands and thousands 150 00:12:23,890 --> 00:12:28,250 Speaker 1: of different profiles, each with different answers, each revealing a 151 00:12:28,290 --> 00:12:32,250 Speaker 1: different woman's profile, then look each of them and try 152 00:12:32,290 --> 00:12:35,130 Speaker 1: to figure out what sort of women seem to answer, 153 00:12:35,170 --> 00:12:40,210 Speaker 1: what sort of questions put like that, it sounds impossible 154 00:12:41,410 --> 00:12:44,250 Speaker 1: unless you're relaxed about bending the rules. And in the 155 00:12:44,290 --> 00:12:49,770 Speaker 1: final months of a PhD in computer science, Chris created 156 00:12:49,810 --> 00:12:53,810 Speaker 1: an army of bots, programmed them to simulate the clicking 157 00:12:53,890 --> 00:12:57,570 Speaker 1: and typing behavior of humans it wouldn't get banned, and 158 00:12:57,770 --> 00:13:03,530 Speaker 1: sent them into battle cautionary tales will be back after 159 00:13:03,530 --> 00:13:17,090 Speaker 1: the break. Nearly half a century after the event, Dan 160 00:13:17,370 --> 00:13:21,770 Speaker 1: Slater received a package from his father. It contained a 161 00:13:21,810 --> 00:13:25,370 Speaker 1: stack of old letters and postcards from nineteen sixty six 162 00:13:26,130 --> 00:13:29,410 Speaker 1: written by his mother at Mount Holyoke College to his 163 00:13:29,490 --> 00:13:34,810 Speaker 1: father at Harvard University. Dan had never asked his parents 164 00:13:35,010 --> 00:13:41,370 Speaker 1: how they had met. The letters revealed the secret. In 165 00:13:41,410 --> 00:13:46,810 Speaker 1: one note, his mother wrote, thank you you old mathematically minded, 166 00:13:46,930 --> 00:13:50,490 Speaker 1: can't mind your own business mass production postcard instigating work 167 00:13:50,530 --> 00:13:55,170 Speaker 1: of art in stainles steel computer. Thank you who writes 168 00:13:55,170 --> 00:13:59,010 Speaker 1: a thank you note to a computer then sends it 169 00:13:59,010 --> 00:14:03,410 Speaker 1: to her boyfriend? Further down the pile was a questioned 170 00:14:03,450 --> 00:14:10,890 Speaker 1: aire contact personality Preference inventory. Dan called his father to 171 00:14:10,930 --> 00:14:14,570 Speaker 1: ask what on earth he was looking at. His father 172 00:14:14,730 --> 00:14:20,090 Speaker 1: was slightly puzzled, Oh, yeah, can you know him? Mother 173 00:14:20,130 --> 00:14:23,970 Speaker 1: and I met through a computer dating service. These days 174 00:14:23,970 --> 00:14:28,010 Speaker 1: they're all over the internet. They certainly are, but such 175 00:14:28,170 --> 00:14:32,810 Speaker 1: dating services are older than you might think. Mister and 176 00:14:32,850 --> 00:14:38,090 Speaker 1: missus Slater were indeed introduced in nineteen sixty six by 177 00:14:38,130 --> 00:14:43,250 Speaker 1: an IBM supercomputer on the basis of their answers to questionnaires, 178 00:14:44,090 --> 00:14:48,090 Speaker 1: and something must have gone well, because otherwise Dan Slater 179 00:14:48,370 --> 00:14:56,650 Speaker 1: wouldn't exist. In the mid nineteen sixties, some young Harvard 180 00:14:56,690 --> 00:15:00,410 Speaker 1: students found themselves drinking and keeping each other company on 181 00:15:00,450 --> 00:15:03,970 Speaker 1: a Saturday night, having failed to persuade any women to 182 00:15:04,010 --> 00:15:08,170 Speaker 1: spend the evening with them. The dating scene was tough. 183 00:15:08,250 --> 00:15:12,050 Speaker 1: They agreed there were two ways to meet girls, blind 184 00:15:12,130 --> 00:15:16,570 Speaker 1: dates or parties, but blind dates were a lottery and 185 00:15:16,770 --> 00:15:20,730 Speaker 1: parties were awkward. The friends agreed there had to be 186 00:15:20,770 --> 00:15:23,370 Speaker 1: a better way to get a date, and so they 187 00:15:23,410 --> 00:15:31,210 Speaker 1: set up an impressive sounding organization, Compatibility Research Ink. How 188 00:15:31,290 --> 00:15:35,130 Speaker 1: much Hugo Gernsback had asked back in nineteen twenty four 189 00:15:36,050 --> 00:15:41,770 Speaker 1: would people give to know whether they were compatible? Compatibility 190 00:15:41,810 --> 00:15:47,250 Speaker 1: Research Ink charged three dollars about thirty dollars in today's terms. 191 00:15:48,050 --> 00:15:52,290 Speaker 1: Hopeful singles would then fill out a questionnaire. There were 192 00:15:52,450 --> 00:15:58,490 Speaker 1: simple numerical questions age, height, grade point average. This was Harvard, 193 00:15:58,490 --> 00:16:03,930 Speaker 1: After all, there were less quantifiable queries. Do you believe 194 00:16:03,970 --> 00:16:09,090 Speaker 1: in a god who answers? Prayer? Is extensive? Sexual activity 195 00:16:09,250 --> 00:16:13,810 Speaker 1: and preparation for marriage part of growing up? There were 196 00:16:13,850 --> 00:16:18,370 Speaker 1: even little scenarios with multiple choice responses. If you were 197 00:16:18,650 --> 00:16:21,610 Speaker 1: set up with a blind date for a dance, but 198 00:16:21,690 --> 00:16:28,050 Speaker 1: the person was quote embarrassingly unattractive, would you one suggest 199 00:16:28,050 --> 00:16:32,130 Speaker 1: a movie instead, two move in on your roommate's date, 200 00:16:32,890 --> 00:16:36,570 Speaker 1: Three go through with a date that make excuses and 201 00:16:36,650 --> 00:16:41,210 Speaker 1: leave early, or four be very friendly at the risk 202 00:16:41,250 --> 00:16:45,610 Speaker 1: of being backed into a second date. What exactly the 203 00:16:45,690 --> 00:16:48,810 Speaker 1: computer was supposed to make of all this Is unclear, 204 00:16:49,450 --> 00:16:54,130 Speaker 1: but the founders of Compatibility Research, Inc. Were clearly having fun. 205 00:16:55,570 --> 00:16:58,610 Speaker 1: The answers would be converted into a punch card, and 206 00:16:58,690 --> 00:17:03,010 Speaker 1: the all knowing computer, an IBM mainframe the size of 207 00:17:03,050 --> 00:17:08,370 Speaker 1: a bus, would do the rest. If you think that 208 00:17:08,530 --> 00:17:12,250 Speaker 1: all sound a bit like Okay Cupid, you might be right. 209 00:17:13,050 --> 00:17:16,330 Speaker 1: Oka Cupid, which was founded in two thousand and four, 210 00:17:16,650 --> 00:17:20,970 Speaker 1: also by Harvard students, also promised to use questionnaires to 211 00:17:20,970 --> 00:17:25,890 Speaker 1: find the perfectly compatible couple. You can see why people 212 00:17:26,010 --> 00:17:31,610 Speaker 1: found this plausible. One of Compatibility Inc's founders later recalled 213 00:17:31,650 --> 00:17:34,810 Speaker 1: that the computer just gave the whole exercise a sense 214 00:17:34,850 --> 00:17:41,090 Speaker 1: of scientific legitimacy, but the truth was much simpler. Jeff Tar, 215 00:17:41,490 --> 00:17:45,650 Speaker 1: one of the founders, admitted much later, the first thing 216 00:17:45,690 --> 00:17:47,490 Speaker 1: we did was to make sure they were in the 217 00:17:47,530 --> 00:17:50,890 Speaker 1: same area. Mostly girls wanted to go out with boys 218 00:17:50,890 --> 00:17:54,370 Speaker 1: who were the same age or older. They're high, tor taller, 219 00:17:54,490 --> 00:17:58,370 Speaker 1: the same religion. So after we had these cuts, then 220 00:17:58,410 --> 00:18:04,290 Speaker 1: we just kind of randomly matched them. That's it. So 221 00:18:04,410 --> 00:18:09,810 Speaker 1: much for compatibility Research. The IBM computer did what computers 222 00:18:09,850 --> 00:18:14,290 Speaker 1: do very easily, found a match for ZIP code, religion, age, 223 00:18:14,330 --> 00:18:20,930 Speaker 1: and height. Further sorting was usually unnecessary, and the implication 224 00:18:21,010 --> 00:18:24,170 Speaker 1: of Jeff Tar's account is that most of the questions 225 00:18:24,210 --> 00:18:30,010 Speaker 1: were there purely for effect. According to Dan Slater's book 226 00:18:30,610 --> 00:18:35,130 Speaker 1: Love in the Time of Algorithms, the founders of Compatibility 227 00:18:35,170 --> 00:18:38,810 Speaker 1: Research were planning to take first pick of the ladies 228 00:18:38,810 --> 00:18:43,530 Speaker 1: who signed up. Ok Cupid had many more customers and 229 00:18:43,610 --> 00:18:48,050 Speaker 1: many more questions to work with than Compatibility Research, but 230 00:18:48,090 --> 00:18:52,170 Speaker 1: it's not clear that their matching algorithm really worked any better. 231 00:18:53,810 --> 00:18:57,610 Speaker 1: In the summer of twenty fourteen, ok Cupid published the 232 00:18:57,650 --> 00:19:00,170 Speaker 1: results of a few experiments that had been running on 233 00:19:00,210 --> 00:19:04,050 Speaker 1: the site's users. One of these experiments deployed a kind 234 00:19:04,090 --> 00:19:08,890 Speaker 1: of placebo matching algorithm. Users were told that they were 235 00:19:08,930 --> 00:19:12,810 Speaker 1: not ziety percent compatible, whatever that means, even though the 236 00:19:12,850 --> 00:19:17,730 Speaker 1: computer estimated they were barely compatible at all. Others were 237 00:19:17,770 --> 00:19:21,290 Speaker 1: told the same story about their high level of compatibility, 238 00:19:21,690 --> 00:19:25,730 Speaker 1: except that the computer actually believed it. The result of 239 00:19:25,770 --> 00:19:30,370 Speaker 1: the experiment, what the computer actually thought didn't matter much. 240 00:19:31,010 --> 00:19:34,050 Speaker 1: What mattered was that users were told they were a 241 00:19:34,050 --> 00:19:39,490 Speaker 1: good match. That assurance was perfectly effective at promoting an 242 00:19:39,530 --> 00:19:46,970 Speaker 1: extended online chat. Compatibility was a placebo. Believing that you 243 00:19:47,050 --> 00:19:52,850 Speaker 1: were compatible was extremely helpful. Actual compatibility, according to the algorithm, 244 00:19:53,330 --> 00:20:02,170 Speaker 1: didn't add much extra. Did computer dating via punch Card's work. Sure, 245 00:20:02,570 --> 00:20:06,570 Speaker 1: up to a point. Dan Slater exists, After all, a 246 00:20:06,610 --> 00:20:10,930 Speaker 1: computer dating service succeeded in introducing his parents to each 247 00:20:10,970 --> 00:20:14,090 Speaker 1: other in the nineteen sixties, and they succeeded in getting 248 00:20:14,130 --> 00:20:21,050 Speaker 1: married and producing Dan. Then again, Dan's parents divorced when 249 00:20:21,090 --> 00:20:24,930 Speaker 1: he was three years old. So if you're going to 250 00:20:24,930 --> 00:20:28,210 Speaker 1: give the old mathematically minded can't mind your own business 251 00:20:28,210 --> 00:20:31,810 Speaker 1: mass production postcard instigating work of art in stainless steel 252 00:20:31,810 --> 00:20:35,890 Speaker 1: computer A grade, it's hard to go much higher than 253 00:20:35,930 --> 00:20:43,250 Speaker 1: a B minus. Back in the twenty first century, computer 254 00:20:43,290 --> 00:20:47,410 Speaker 1: whiz and former card counter Chris McKinley had many more 255 00:20:47,490 --> 00:20:51,570 Speaker 1: potential dates, a much more powerful algorithm to work with, 256 00:20:51,970 --> 00:20:56,930 Speaker 1: and an army of dating bots on his side. Surely 257 00:20:57,330 --> 00:21:03,290 Speaker 1: he couldn't fail. McKinley's bots harvested data on six million 258 00:21:03,490 --> 00:21:09,330 Speaker 1: answers from twenty thousand women. Then he deployed a machine 259 00:21:09,410 --> 00:21:15,170 Speaker 1: learning algorithm called k modes, designed to group data into clusters. 260 00:21:16,410 --> 00:21:20,650 Speaker 1: Kevin Paulson, who wrote about McKinley for Wired magazine, described 261 00:21:20,690 --> 00:21:24,730 Speaker 1: the algorithm as clumping data like the colored wax in 262 00:21:24,810 --> 00:21:28,370 Speaker 1: a lava lamp. With fine tuning, he could adjust the 263 00:21:28,490 --> 00:21:32,610 Speaker 1: viscosity of the results, thinning it into a slick or 264 00:21:32,650 --> 00:21:39,050 Speaker 1: coagulating it into a single glob. Soon enough, McKinley had 265 00:21:39,090 --> 00:21:44,090 Speaker 1: identified some potential globs. There were seven distinct types of 266 00:21:44,130 --> 00:21:47,770 Speaker 1: women who the computer reckoned had some potential for matching 267 00:21:47,810 --> 00:21:52,930 Speaker 1: with him. Sifting through each cluster of women, McKinley disagreed. 268 00:21:53,570 --> 00:21:57,130 Speaker 1: One group of women were highly religious, not a good match. 269 00:21:57,730 --> 00:22:00,370 Speaker 1: Another group was full of women who were new to 270 00:22:00,570 --> 00:22:05,290 Speaker 1: online dating and looking for a fling. Ah. McKinley wanted 271 00:22:05,290 --> 00:22:08,690 Speaker 1: a soul mate. These women over here were too old. 272 00:22:09,210 --> 00:22:13,050 Speaker 1: Those women over there were too young. But there were 273 00:22:13,250 --> 00:22:18,970 Speaker 1: two globs full of women with appealing qualities. One group 274 00:22:19,050 --> 00:22:23,730 Speaker 1: were professional creatives in their thirties. There were designers and editors. 275 00:22:24,450 --> 00:22:29,810 Speaker 1: The other were a little more punk and a little younger, tattooed, musicians, 276 00:22:30,490 --> 00:22:34,410 Speaker 1: free spirited. He thought it'd take a chance on both. 277 00:22:35,890 --> 00:22:39,570 Speaker 1: But Chris had barely begun to solve his problem, because 278 00:22:39,610 --> 00:22:43,650 Speaker 1: now he had to design the perfect profile, or actually 279 00:22:43,810 --> 00:22:49,930 Speaker 1: two different perfect profiles, one for each cluster. His dream 280 00:22:50,090 --> 00:22:53,010 Speaker 1: was that when one of those women logged on, no 281 00:22:53,010 --> 00:22:57,850 Speaker 1: matter what exactly she searched for or how exactly she 282 00:22:57,970 --> 00:23:03,690 Speaker 1: answered the questions, the computer would have a single top recommendation, 283 00:23:04,970 --> 00:23:14,010 Speaker 1: have you considered a date with Chris McKinley. Of course, 284 00:23:14,410 --> 00:23:17,370 Speaker 1: Chris had exactly the data he needed to construct his 285 00:23:17,650 --> 00:23:23,170 Speaker 1: Adonis profile, and he gave himself a little extra help. Okay, 286 00:23:23,250 --> 00:23:27,730 Speaker 1: Cupid members get a notification when someone views their profile page, 287 00:23:28,210 --> 00:23:33,010 Speaker 1: So Chris simply programmed another bot to patiently visit every 288 00:23:33,130 --> 00:23:37,330 Speaker 1: woman who matched. On Monday, the bot would swing past 289 00:23:37,330 --> 00:23:40,970 Speaker 1: one thousand, forty one year olds. On Tuesday, it would 290 00:23:41,050 --> 00:23:44,370 Speaker 1: visit one thousand and forty year olds. Two weeks later 291 00:23:44,490 --> 00:23:46,890 Speaker 1: it would get to the twenty seven year olds, and 292 00:23:46,930 --> 00:23:52,650 Speaker 1: then it would begin the cycle again. But would all 293 00:23:52,730 --> 00:23:57,810 Speaker 1: this really work? His PhD had taken a back seat 294 00:23:57,930 --> 00:24:02,130 Speaker 1: while he deployed all his computational chops to the problem 295 00:24:02,170 --> 00:24:07,570 Speaker 1: of getting a girlfriend. Was he just fooling himself? Feeling 296 00:24:07,810 --> 00:24:13,010 Speaker 1: slightly silly, Chris loaded up his superhero profiles, let his 297 00:24:13,210 --> 00:24:17,570 Speaker 1: bot off the leash, unrolled his sleeping mat, and went 298 00:24:17,970 --> 00:24:26,850 Speaker 1: to sleep. And he woke up. He recalls, that's the 299 00:24:27,010 --> 00:24:30,530 Speaker 1: point at which events started to take their own course, 300 00:24:31,410 --> 00:24:34,450 Speaker 1: and I think I stopped really being in control of this. 301 00:24:36,170 --> 00:24:43,810 Speaker 1: Chris found himself drowning in flirtatious messages. I haven't until 302 00:24:43,890 --> 00:24:48,690 Speaker 1: now come across anyone with such winning numbers, writes one woman. 303 00:24:49,410 --> 00:24:53,250 Speaker 1: Also something about a rugged man who's really good with numbers. 304 00:24:53,770 --> 00:24:58,290 Speaker 1: Thought I'd say hi. All of these messages came from 305 00:24:58,330 --> 00:25:01,810 Speaker 1: exactly the kind of woman Chris had already decided he 306 00:25:01,930 --> 00:25:05,930 Speaker 1: was keen to meet. His hero profiles were a ninety 307 00:25:06,050 --> 00:25:09,810 Speaker 1: nine percent match with dozens and dozens of local women, 308 00:25:10,090 --> 00:25:13,930 Speaker 1: and ninety percent compatible with more than ten thousand of 309 00:25:13,970 --> 00:25:19,690 Speaker 1: his fellow Los Angelinos. Every few minutes, a new message 310 00:25:19,730 --> 00:25:24,010 Speaker 1: came in. I wasn't really prepared to sort through all 311 00:25:24,010 --> 00:25:30,170 Speaker 1: the kind of human consequences, he recalled, and again Carpai 312 00:25:30,290 --> 00:25:48,690 Speaker 1: dm man cautionary tales will returned after a break. Here 313 00:25:48,770 --> 00:25:54,370 Speaker 1: is a puzzle. Sex between consenting adults has never been 314 00:25:54,450 --> 00:25:58,970 Speaker 1: less taboo, and yet somehow people are having less and 315 00:25:59,130 --> 00:26:04,290 Speaker 1: less sex. Americans are having their first sexual experiences later 316 00:26:04,370 --> 00:26:08,170 Speaker 1: than they used to, Millennials born in the nineteen nineties 317 00:26:09,090 --> 00:26:12,730 Speaker 1: which less likely to have had any sexual partners between 318 00:26:12,730 --> 00:26:16,250 Speaker 1: their eighteenth birthday and their early twenties than were jen 319 00:26:16,610 --> 00:26:20,890 Speaker 1: xers born in the nineteen sixties, and American adults in 320 00:26:21,010 --> 00:26:26,210 Speaker 1: general were having sex nine fewer times per year in 321 00:26:26,250 --> 00:26:30,410 Speaker 1: the early twenty tens than the late nineteen nineties, partly 322 00:26:30,450 --> 00:26:33,530 Speaker 1: for the very obvious reason that people had become less 323 00:26:33,690 --> 00:26:38,650 Speaker 1: likely to have a long term sexual partner. Let's allow 324 00:26:38,690 --> 00:26:42,970 Speaker 1: that to sink in for a moment. We're in a 325 00:26:43,170 --> 00:26:47,290 Speaker 1: sex recession, partly because people are just less likely to 326 00:26:47,330 --> 00:26:50,930 Speaker 1: have a spouse or a steady girlfriend or boyfriend, and 327 00:26:51,050 --> 00:26:57,210 Speaker 1: yet the most powerful dating technologies ever designed are sitting 328 00:26:57,210 --> 00:27:03,330 Speaker 1: in our pockets. What is going on? No doubt, there 329 00:27:03,330 --> 00:27:07,290 Speaker 1: are many factors that help to explain the sex recession, 330 00:27:07,930 --> 00:27:11,770 Speaker 1: but let's focus on one. Maybe these dating apps just 331 00:27:12,770 --> 00:27:17,730 Speaker 1: aren't very good. Tinder has been losing users, but as 332 00:27:17,810 --> 00:27:21,970 Speaker 1: Valentine's Day twenty twenty six approaches, it's still much the 333 00:27:22,010 --> 00:27:26,970 Speaker 1: most popular dating app. And maybe Tinder in twenty twenty 334 00:27:27,090 --> 00:27:30,930 Speaker 1: six feels a bit like Chris McKinley after hacking ok 335 00:27:31,210 --> 00:27:34,610 Speaker 1: Cupid in twenty twelve. There's just no end to the 336 00:27:34,690 --> 00:27:41,330 Speaker 1: number of possible matches, which should help, right. Maybe not. 337 00:27:44,410 --> 00:27:48,770 Speaker 1: About twenty five years ago, the psychologists Daniel Gilbert and 338 00:27:48,890 --> 00:27:52,970 Speaker 1: Jane Ebert conducted an experiment that seems to be about 339 00:27:53,010 --> 00:27:57,730 Speaker 1: photography but actually points at a deep contradiction between what 340 00:27:57,890 --> 00:28:03,370 Speaker 1: love lawn singles want and what online dating companies give them. 341 00:28:04,010 --> 00:28:08,730 Speaker 1: Gilbert and Ebert recruited Harvard students for a photography course, 342 00:28:09,290 --> 00:28:13,010 Speaker 1: pretending to be conducting a study of teaching methods. The 343 00:28:13,050 --> 00:28:16,290 Speaker 1: students were given hours of training and an analog camera 344 00:28:16,530 --> 00:28:20,410 Speaker 1: and told to make a dozen photographs that meaningfully captured 345 00:28:20,450 --> 00:28:24,290 Speaker 1: their time at Harvard. Then they spent supervised time in 346 00:28:24,330 --> 00:28:27,530 Speaker 1: the darkroom making a contact sheet and were asked to 347 00:28:27,570 --> 00:28:32,610 Speaker 1: evaluate each photograph. How much did they like each composition 348 00:28:32,890 --> 00:28:36,010 Speaker 1: on a scale of one to nine. They made beautiful 349 00:28:36,170 --> 00:28:40,010 Speaker 1: eight by ten inch glossy prints of two of the negatives. 350 00:28:40,810 --> 00:28:45,050 Speaker 1: Then came the experiment. All the students were told that 351 00:28:45,090 --> 00:28:48,970 Speaker 1: the Teaching Project, which had sponsored all this training, needed 352 00:28:49,010 --> 00:28:51,210 Speaker 1: to keep one of the prints on file for their 353 00:28:51,250 --> 00:28:55,490 Speaker 1: records back at headquarters in England. The student could keep 354 00:28:55,530 --> 00:28:59,850 Speaker 1: the other print as a memento. So which one would 355 00:28:59,850 --> 00:29:02,570 Speaker 1: they like to keep and which would be shipped to 356 00:29:02,610 --> 00:29:08,690 Speaker 1: the archive in England. An agonizing choice two handmade prints 357 00:29:09,290 --> 00:29:14,130 Speaker 1: labors of love, originally designed to be meaningful. Which one 358 00:29:14,170 --> 00:29:18,850 Speaker 1: would go and which would stay. Half of the students, 359 00:29:19,050 --> 00:29:22,410 Speaker 1: chosen at random, were told that the prince they relinquished 360 00:29:22,970 --> 00:29:26,130 Speaker 1: were going to be mailed out to England that afternoon. 361 00:29:27,050 --> 00:29:29,130 Speaker 1: The other half were told that the prince wouldn't be 362 00:29:29,170 --> 00:29:32,090 Speaker 1: sent off for a few days, and the experiment who 363 00:29:32,090 --> 00:29:34,890 Speaker 1: would call them beforehand to double check. They didn't want 364 00:29:34,930 --> 00:29:39,450 Speaker 1: to change their mind. Instinctively, they'd think it would be 365 00:29:39,570 --> 00:29:42,730 Speaker 1: nice to have the option to switch, or at the 366 00:29:42,810 --> 00:29:47,370 Speaker 1: very least, that it would do no harm wrong. It's 367 00:29:47,450 --> 00:29:50,450 Speaker 1: much better to have a now or never choice and 368 00:29:50,650 --> 00:29:54,650 Speaker 1: try to make the best of it. Dan Gilbert summarized 369 00:29:54,690 --> 00:29:58,930 Speaker 1: the findings this way. People who are stuck with that picture, 370 00:29:59,090 --> 00:30:02,610 Speaker 1: who have no choice, who can never change their mind, 371 00:30:02,690 --> 00:30:06,130 Speaker 1: like it a lot. And people who are deliberating should 372 00:30:06,170 --> 00:30:08,570 Speaker 1: I return it? Have I gotten the right one? Maybe 373 00:30:08,570 --> 00:30:11,730 Speaker 1: this isn't the good one? Maybe I left the good one? Well, 374 00:30:11,850 --> 00:30:17,210 Speaker 1: says Gilbert. Those people have destroyed their joy by pondering 375 00:30:17,250 --> 00:30:22,250 Speaker 1: their choices over and over again. They don't like their picture. 376 00:30:25,250 --> 00:30:30,050 Speaker 1: The implications of this study for online dating are devastating. 377 00:30:31,050 --> 00:30:35,850 Speaker 1: Dating apps seem to offer infinite romantic riches, endless scrolling, 378 00:30:36,090 --> 00:30:41,530 Speaker 1: endless swipes, endless choice, endless opportunities to ghost somebody and 379 00:30:41,610 --> 00:30:45,210 Speaker 1: dive back into the dating pool. But if Gilbert and 380 00:30:45,330 --> 00:30:49,530 Speaker 1: Ebert are right, there is no better way to ensure 381 00:30:49,570 --> 00:30:53,130 Speaker 1: that you're unsatisfied with the romantic prospect in front of 382 00:30:53,170 --> 00:30:56,730 Speaker 1: you than to know that another date is only a 383 00:30:56,810 --> 00:31:03,890 Speaker 1: click away and one more thing. Hugo gernsback rhetorically asked 384 00:31:04,610 --> 00:31:07,810 Speaker 1: how much would people give to know that they were compatible? 385 00:31:09,090 --> 00:31:13,970 Speaker 1: Had it backwards. This is a monthly subscription we're talking about. 386 00:31:14,770 --> 00:31:17,450 Speaker 1: People don't pay to find the love of their life. 387 00:31:17,530 --> 00:31:21,730 Speaker 1: They pay to keep looking. If they succeed, that's when 388 00:31:21,770 --> 00:31:26,410 Speaker 1: the subscription revenues dry up. Thankfully for the dating companies, 389 00:31:27,010 --> 00:31:35,930 Speaker 1: success doesn't seem to be all that likely. Chris McKinley 390 00:31:36,210 --> 00:31:39,930 Speaker 1: took a shower in the UCLA gym and drove across 391 00:31:39,970 --> 00:31:45,930 Speaker 1: town for date number one, lunch with Sheila. He was scared. 392 00:31:47,730 --> 00:31:52,010 Speaker 1: Lunch went nowhere. Neither he nor Sheila were feeling it. 393 00:31:52,690 --> 00:31:55,690 Speaker 1: His second date was a walk with a blog editor. 394 00:31:56,770 --> 00:31:59,650 Speaker 1: She was feeling low by the end of the walk, 395 00:32:00,410 --> 00:32:05,610 Speaker 1: so was Chris. Date three was Allison. She was cool. 396 00:32:06,410 --> 00:32:09,650 Speaker 1: They met in a bar in Korea Town. He drank 397 00:32:09,890 --> 00:32:13,050 Speaker 1: too much Korean beer and woke up in his cubicle 398 00:32:13,130 --> 00:32:20,090 Speaker 1: with a terrible hangover. He messaged her. She didn't right back. 399 00:32:21,810 --> 00:32:25,610 Speaker 1: Chris didn't not have a good time. He got invited 400 00:32:25,650 --> 00:32:29,490 Speaker 1: to some crazy parties. He had romantic walks along the 401 00:32:29,530 --> 00:32:34,330 Speaker 1: canals of the Venice District. But as the dates piled up, 402 00:32:34,890 --> 00:32:40,250 Speaker 1: Chris started to seek out shortcuts. No elaborate plans, no 403 00:32:40,370 --> 00:32:45,730 Speaker 1: concerts or movies, no drinking, grab lunch or coffee. Focus 404 00:32:45,770 --> 00:32:48,290 Speaker 1: on her and get to know her a bit. Cut 405 00:32:48,330 --> 00:32:51,770 Speaker 1: it off fast when it's not working. No driving across town, 406 00:32:52,250 --> 00:32:58,930 Speaker 1: keep it local. He was exhausted, and he was also 407 00:32:59,290 --> 00:33:03,330 Speaker 1: struggling to keep track all these girls had been picked 408 00:33:03,410 --> 00:33:07,570 Speaker 1: by the algorithm from the same cluster. He might meet 409 00:33:07,730 --> 00:33:12,170 Speaker 1: three or four on a single weekend. Sometimes he'd take 410 00:33:12,250 --> 00:33:15,010 Speaker 1: two different women to the same place on the same day. 411 00:33:15,810 --> 00:33:18,410 Speaker 1: It was hard to keep them distinct in his mind. 412 00:33:18,690 --> 00:33:22,730 Speaker 1: They were all similar, at least superficially, And if all 413 00:33:22,730 --> 00:33:25,890 Speaker 1: you're doing is getting coffee, then you're never going to 414 00:33:25,890 --> 00:33:35,810 Speaker 1: get past those surface impressions. Thirty first dates past, fifty 415 00:33:35,850 --> 00:33:41,330 Speaker 1: first dates, seventy five. I'm not sure how many of 416 00:33:41,450 --> 00:33:45,570 Speaker 1: us have managed seventy five first dates in our entire lives, 417 00:33:45,810 --> 00:33:50,770 Speaker 1: and how I haven't. But Chris's cleverness with okay Cupid 418 00:33:51,570 --> 00:33:57,650 Speaker 1: had produced only three second dates and just one third date. 419 00:33:58,970 --> 00:34:03,050 Speaker 1: Chris had optimized the heck out of online dating and 420 00:34:03,090 --> 00:34:06,930 Speaker 1: then out of dating in real life. He'd had quite 421 00:34:06,970 --> 00:34:14,290 Speaker 1: a ride, but one thing he didn't have was a girlfriend. 422 00:34:18,570 --> 00:34:23,930 Speaker 1: The late Daniel Carneman, Nobel Prize winning psychologist and author 423 00:34:23,970 --> 00:34:28,770 Speaker 1: of Thinking Fast and Slow, offered many pearls of wisdom 424 00:34:28,810 --> 00:34:32,530 Speaker 1: in his life, but one that's particularly stuck with me 425 00:34:32,970 --> 00:34:37,970 Speaker 1: is this. When faced with a difficult question, we often 426 00:34:38,170 --> 00:34:44,330 Speaker 1: answer an easier one instead, usually without noticing the substitution 427 00:34:46,170 --> 00:34:50,210 Speaker 1: that may be the reason that algorithmic matching appealed to 428 00:34:50,290 --> 00:34:55,210 Speaker 1: Chris and appeals to many other people. The question Chris 429 00:34:55,330 --> 00:34:58,930 Speaker 1: was facing was could I be happy with this woman 430 00:34:59,330 --> 00:35:03,170 Speaker 1: for years, maybe for the rest of my life, and 431 00:35:03,210 --> 00:35:08,170 Speaker 1: could she be happy with me? That's a hard question, 432 00:35:09,130 --> 00:35:11,610 Speaker 1: and it's not a question a computer armed with some 433 00:35:11,730 --> 00:35:16,690 Speaker 1: questionnaire results could reasonably help with. But here's an easier question. 434 00:35:17,850 --> 00:35:19,690 Speaker 1: Do we say we like the same bands and the 435 00:35:19,730 --> 00:35:22,770 Speaker 1: same books and have similar attitudes to religion and to 436 00:35:22,850 --> 00:35:25,770 Speaker 1: sex and to beer. The computer can help you with that, 437 00:35:25,930 --> 00:35:33,770 Speaker 1: for sure, and perhaps you won't notice the substitution. Chris 438 00:35:33,850 --> 00:35:36,930 Speaker 1: was just about ready to shut down the bots and 439 00:35:37,130 --> 00:35:41,050 Speaker 1: cancel his account, and then he received a message from 440 00:35:41,170 --> 00:35:48,450 Speaker 1: Christine Tien Wang. She was an artist and an activist, ambitious, confident, 441 00:35:49,010 --> 00:35:53,850 Speaker 1: direct like him. She was a grad student at UCLA. 442 00:35:54,130 --> 00:35:58,010 Speaker 1: They met at the campus sculpture Garden, then went for sushi. 443 00:35:59,050 --> 00:36:03,890 Speaker 1: There was a spark right from the start. Chris confessed 444 00:36:04,130 --> 00:36:09,690 Speaker 1: to his love hacking, dark and cynical, thought Christine, I 445 00:36:09,850 --> 00:36:16,210 Speaker 1: like it. Two weeks later, they both closed their okay 446 00:36:16,410 --> 00:36:26,050 Speaker 1: Cupid accounts. When Wired magazine reported on the story a 447 00:36:26,170 --> 00:36:30,170 Speaker 1: year after that first date, they were able to share 448 00:36:30,250 --> 00:36:35,290 Speaker 1: the happy news that Chris had proposed and Christine had 449 00:36:35,290 --> 00:36:40,250 Speaker 1: said yes. A decade after that, the BBC returned to 450 00:36:40,290 --> 00:36:47,330 Speaker 1: the story Chris and Christine were still together, still blissfully happy. 451 00:36:48,890 --> 00:36:53,170 Speaker 1: Love really is a numbers game, equipped the BBC host 452 00:36:53,290 --> 00:36:58,970 Speaker 1: Hannah Frye, while Wired's headline was how a math genius 453 00:36:59,370 --> 00:37:05,650 Speaker 1: hacked okay Cupid to find true love. It seems that 454 00:37:05,730 --> 00:37:09,010 Speaker 1: we can't quite shake off the idea that science and 455 00:37:09,170 --> 00:37:14,370 Speaker 1: technology they're going to solve romance for us. But I'm 456 00:37:14,410 --> 00:37:17,730 Speaker 1: not sure that's the lesson I draw from this cautionary tale. 457 00:37:19,170 --> 00:37:25,170 Speaker 1: Before meeting Christine, Chris's brilliant hack had delivered eighty seven 458 00:37:25,410 --> 00:37:30,810 Speaker 1: first dates, none of which went anywhere significant, and most 459 00:37:30,810 --> 00:37:36,090 Speaker 1: of which went nowhere at all. That's an astonishingly low 460 00:37:36,170 --> 00:37:39,530 Speaker 1: hit rate. If he had had eighty seven dates with 461 00:37:39,730 --> 00:37:44,090 Speaker 1: eighty seven people randomly chosen from the Los Angeles telephone directory, 462 00:37:44,490 --> 00:37:49,090 Speaker 1: I'm not sure he could really have done any worse. Ah, 463 00:37:49,530 --> 00:37:53,850 Speaker 1: but you might say the eighty eighth date proved that 464 00:37:54,050 --> 00:38:01,010 Speaker 1: his system worked. In the end, not at all. According 465 00:38:01,010 --> 00:38:06,490 Speaker 1: to the algorithm, Christine wasn't in his top ten thousand matches. 466 00:38:07,650 --> 00:38:10,930 Speaker 1: They only matched because she searched on pretty much the 467 00:38:11,010 --> 00:38:17,050 Speaker 1: only information he hadn't tweaked one hundred and eighty centimeters tall, 468 00:38:17,810 --> 00:38:24,810 Speaker 1: blue eyes near UCLA. You didn't find me, she says, 469 00:38:25,810 --> 00:38:33,210 Speaker 1: I found you. The artist, not the maths genius made 470 00:38:33,250 --> 00:38:39,650 Speaker 1: the match. Shig McGrath be damned. Maybe romance will always 471 00:38:39,690 --> 00:38:53,970 Speaker 1: be an art, not a science if you haven't had 472 00:38:54,050 --> 00:38:57,410 Speaker 1: enough of amorous accidents to really get you in the 473 00:38:57,530 --> 00:39:01,690 Speaker 1: Valentine spirit. There is a whole additional episode over on 474 00:39:01,730 --> 00:39:05,810 Speaker 1: the Patreon feed for members of our Cautionary Club, where 475 00:39:05,810 --> 00:39:08,730 Speaker 1: I tell a collection of these shorter stories I found 476 00:39:08,810 --> 00:39:11,770 Speaker 1: over the years. Here's a quick clip, and if you 477 00:39:11,770 --> 00:39:14,810 Speaker 1: want to hear the rest, just head over to patreon 478 00:39:14,930 --> 00:39:22,570 Speaker 1: dot com slash Cautionary Club enjoy. Nigel Blundell's book The 479 00:39:22,610 --> 00:39:27,570 Speaker 1: World's Greatest Mistakes describes another case from the late nineteen seventies. 480 00:39:28,290 --> 00:39:31,610 Speaker 1: An airline pilot had a wife in one city and 481 00:39:31,650 --> 00:39:35,770 Speaker 1: a girlfriend in another. The girlfriend was ensconced in his 482 00:39:35,810 --> 00:39:39,010 Speaker 1: flat in London. As it happens. We'd try not to 483 00:39:39,050 --> 00:39:42,210 Speaker 1: be too judgmental here on Cautionary Tales, but I am 484 00:39:42,290 --> 00:39:45,170 Speaker 1: sorry to report that I hold this gentleman in low 485 00:39:45,250 --> 00:39:48,930 Speaker 1: regard since after a few months, he decided that his 486 00:39:49,010 --> 00:39:53,770 Speaker 1: mistress was surplus to requirements, and he evicted her, giving 487 00:39:53,770 --> 00:39:56,370 Speaker 1: her a couple of days to move out while he 488 00:39:56,450 --> 00:40:00,490 Speaker 1: departed on a series of long distance flights. When the 489 00:40:00,690 --> 00:40:05,090 Speaker 1: cad returned, he found that his girlfriend ex girlfriend was 490 00:40:05,130 --> 00:40:11,170 Speaker 1: gone and the apartment was spotless. There was just one 491 00:40:11,250 --> 00:40:17,690 Speaker 1: thing out of place, the telephone receiver. When he picked 492 00:40:17,730 --> 00:40:21,210 Speaker 1: it up and placed it to his ear, he heard 493 00:40:21,250 --> 00:40:25,730 Speaker 1: only the sound of an American voice, endlessly announcing the 494 00:40:25,850 --> 00:40:30,850 Speaker 1: time she'd dialed the speaking clock in Washington, d C. 495 00:40:31,930 --> 00:40:38,530 Speaker 1: Before she made her exit. Fifty years ago, transatlantic telephone 496 00:40:38,570 --> 00:40:43,290 Speaker 1: calls weren't cheap. The phone bill was, in today's terms, 497 00:40:43,970 --> 00:40:57,090 Speaker 1: about ten thousand dollars. For a full list of sources, 498 00:40:57,370 --> 00:41:03,170 Speaker 1: see Timharford dot com. Cautionary Tales is written by me 499 00:41:03,410 --> 00:41:07,410 Speaker 1: Tim Harford, with Andrew Wright, Alice Fines, and Ryan Dilly. 500 00:41:08,250 --> 00:41:12,530 Speaker 1: It's produced by Georgia Mills and Marilyn Rust. The sound 501 00:41:12,610 --> 00:41:17,130 Speaker 1: design and original music are the work of Pascal wise Bender. 502 00:41:17,170 --> 00:41:21,170 Speaker 1: Dafh Haffrey edited the scripts. It features the voice talents 503 00:41:21,210 --> 00:41:26,810 Speaker 1: of Melanie Guttridge, Genevieve Gaunt, Stella Harford, Massa Munroe, Jamal 504 00:41:26,970 --> 00:41:31,490 Speaker 1: Westman and rufus Wright. The show also wouldn't have been 505 00:41:31,530 --> 00:41:35,890 Speaker 1: possible without the work of Jacob Weisberg, Greta Cohne, Eric Sandler, 506 00:41:36,410 --> 00:41:43,210 Speaker 1: Carrie Brody, Christina Sullivan, Kira Posey, and Owen Miller. Cautionary 507 00:41:43,210 --> 00:41:46,690 Speaker 1: Tales is a production of Pushkin Industries. If you like 508 00:41:46,770 --> 00:41:51,010 Speaker 1: the show, please remember to share, rate, and review. It 509 00:41:51,050 --> 00:41:54,210 Speaker 1: really does make a difference to us and if you 510 00:41:54,290 --> 00:41:57,570 Speaker 1: want to hear it, add free and receive a bonus 511 00:41:57,690 --> 00:42:02,650 Speaker 1: audio episode, video episode, and members only newsletter every month. 512 00:42:03,450 --> 00:42:07,570 Speaker 1: Why not join the Cautionary Club. To sign up, head 513 00:42:07,570 --> 00:42:12,490 Speaker 1: to patreon dot com slash Cautionary Club. That's Patreon, p A, 514 00:42:12,650 --> 00:42:17,770 Speaker 1: t R e o N dot com Slash Cautionary Club.