WEBVTT 1 00:00:06.516 --> 00:00:08.217 Welcome to the Be Good podcast, 2 00:00:08.578 --> 00:00:14.622 where we explore the application of behavioral economics for good in order to nudge better business and better lives. 3 00:00:16.203 --> 00:00:16.443 Hi, 4 00:00:16.823 --> 00:00:19.565 and welcome to this episode of Be Good, 5 00:00:20.186 --> 00:00:22.488 brought to you by BVA Nudge Consulting, 6 00:00:22.908 --> 00:00:29.633 a global consultancy specializing in the application of behavioral science for successful behavioral change. 7 00:00:30.233 --> 00:00:30.894 Every month, 8 00:00:30.914 --> 00:00:34.136 we get to speak with a leader in the field of behavioral science. 9 00:00:34.560 --> 00:00:38.784 psychology and neuroscience in order to get to know more about them, 10 00:00:38.884 --> 00:00:39.504 their work, 11 00:00:40.005 --> 00:00:42.146 and its application to emerging issues. 12 00:00:42.787 --> 00:00:44.048 My name is Eric Singler, 13 00:00:44.348 --> 00:00:48.952 Managing Director of the BVA family and CEO of BVA Dutch Consulting, 14 00:00:49.432 --> 00:00:53.075 and with me is my colleague Richard Bordenave, 15 00:00:53.436 --> 00:00:56.458 Chief Behavioral Science Officer at PIRAS in Vivo. 16 00:00:57.139 --> 00:00:57.879 Hi Richard. 17 00:00:58.520 --> 00:00:59.120 Hi Eric, 18 00:00:59.460 --> 00:01:03.644 I'm excited to be joining you for this episode and I'm delighted 19 00:01:03.784 --> 00:01:05.125 to be introducing our guest, 20 00:01:05.205 --> 00:01:06.887 Professor Gerd Gigerenzer. 21 00:01:07.928 --> 00:01:13.553 Professor Gigerenzer is a giant in the world of academic research on human behavior and decision making. 22 00:01:14.554 --> 00:01:20.619 Professor Gigerenzer is director of Harding Center for Risk Literacy at the University of Potsdam, 23 00:01:21.440 --> 00:01:25.564 director emeritus at the Max Planck Institute for Human Development, 24 00:01:25.724 --> 00:01:28.006 and a partner at Simply Rational, 25 00:01:28.706 --> 00:01:30.248 the Institute for Decision and... 26 00:01:30.700 --> 00:01:34.022 He's also vice president of the European Research Council, 27 00:01:35.083 --> 00:01:39.186 and he's a former professor of psychology at the University of Chicago. 28 00:01:40.287 --> 00:01:41.047 So a few of 29 00:01:41.628 --> 00:01:43.669 Professor Gigerentz's awards include 30 00:01:44.170 --> 00:01:46.711 American Associations for the Advancement of Science, 31 00:01:47.672 --> 00:01:50.354 a prize for the best article in the behavioral sciences, 32 00:01:51.395 --> 00:01:56.458 the Association of the American Publisher Prize for the best book in social and behavioral sciences, 33 00:01:57.199 --> 00:01:58.900 the German Psychology Award, 34 00:01:59.280 --> 00:02:02.842 and the Communicator Award of the German Research Foundation. 35 00:02:03.602 --> 00:02:03.723 So 36 00:02:04.643 --> 00:02:09.186 Professor Gigerenzer is author of multiple books on heuristics and decision-making, 37 00:02:10.146 --> 00:02:13.448 which have been translated in many languages, 38 00:02:13.488 --> 00:02:14.228 like more than 20, 39 00:02:14.869 --> 00:02:16.570 including the book of today, 40 00:02:17.290 --> 00:02:18.231 Smart Management, 41 00:02:18.731 --> 00:02:18.991 How 42 00:02:19.792 --> 00:02:24.034 Simple Heuristics Help Leaders Make Good Decisions in an Uncertain 43 00:02:24.394 --> 00:02:27.156 World. I'm very happy to welcome you, 44 00:02:27.216 --> 00:02:27.676 Professor. 45 00:02:28.684 --> 00:02:31.665 And welcome to our Be Good podcast. 46 00:02:32.565 --> 00:02:32.685 Oh, 47 00:02:32.885 --> 00:02:35.346 thank you for having me here again. 48 00:02:36.347 --> 00:02:40.328 And I'm looking forward to have another insightful conversation with you. 49 00:02:41.468 --> 00:02:42.608 Professor Gigerenzer, 50 00:02:43.209 --> 00:02:48.410 thank you so much again for being with us today for this episode of Be Good. 51 00:02:49.871 --> 00:02:51.631 Before talking about your work, 52 00:02:51.951 --> 00:02:53.172 smart management, 53 00:02:53.412 --> 00:02:54.352 we would like to know... 54 00:02:54.848 --> 00:02:55.709 A little more, 55 00:02:55.749 --> 00:02:56.389 if possible, 56 00:02:56.509 --> 00:02:59.392 about you and your amazing career. 57 00:03:00.793 --> 00:03:07.318 Can you tell us first about how you came to be interested in human behavior in general, 58 00:03:07.919 --> 00:03:14.344 and maybe specifically about your interest in decision-making processes? 59 00:03:16.005 --> 00:03:18.487 I had an earlier career as a musician, 60 00:03:20.048 --> 00:03:23.191 and that is how I financed my studies. 61 00:03:24.832 --> 00:03:26.894 when I did my PhD, 62 00:03:26.914 --> 00:03:32.838 I had to make a decision whether to stay on the stage and play. 63 00:03:33.478 --> 00:03:35.119 That was entertainment music, 64 00:03:35.580 --> 00:03:36.821 mostly jazz, 65 00:03:36.881 --> 00:03:37.501 Dixieland, 66 00:03:37.601 --> 00:03:37.921 soul, 67 00:03:38.942 --> 00:03:44.166 and or risk an academic career. 68 00:03:45.607 --> 00:03:47.008 That was a decision. 69 00:03:47.048 --> 00:03:48.469 How do you make such a decision? 70 00:03:48.909 --> 00:03:49.109 Now, 71 00:03:49.750 --> 00:03:50.030 for me, 72 00:03:50.670 --> 00:03:52.251 the way I do it is, 73 00:03:52.231 --> 00:03:52.852 I'm not going to be a 74 00:03:53.472 --> 00:03:54.893 I have been playing music for, 75 00:03:55.293 --> 00:03:55.874 at this time, 76 00:03:55.874 --> 00:03:57.394 for 14 years on the stage. 77 00:03:58.175 --> 00:03:59.516 It was the safe option. 78 00:04:00.456 --> 00:04:05.379 I knew how it works and I earned much more money than as an assistant professor. 79 00:04:06.900 --> 00:04:08.541 And on the other side, 80 00:04:10.643 --> 00:04:11.263 I thought, 81 00:04:11.583 --> 00:04:11.923 okay, 82 00:04:12.684 --> 00:04:15.646 is this what you want to do for the rest of your life? 83 00:04:17.167 --> 00:04:19.008 And at the end... 84 00:04:19.836 --> 00:04:20.817 I took the risk, 85 00:04:21.317 --> 00:04:28.502 but for me it was a risky decision because I couldn't know whether I would ever make it to a professor at a good university. 86 00:04:29.423 --> 00:04:37.069 And that's a kind of important life decision that inspired me to look closer, 87 00:04:37.469 --> 00:04:39.531 how we actually make decisions. 88 00:04:40.771 --> 00:04:46.696 Could you share with us any mentors that had a particularly strong influence on you? 89 00:04:49.120 --> 00:04:57.604 Do you have any researcher or other people who have played an influential role in your professional career? 90 00:04:59.113 --> 00:05:02.415 So the people who intellectually influenced me, 91 00:05:02.535 --> 00:05:04.016 that was certainly Herbert Simon, 92 00:05:04.836 --> 00:05:09.439 with respect to his acknowledge of uncertainty, 93 00:05:09.859 --> 00:05:15.242 as opposed to a world of calculable risk where you can optimize. 94 00:05:15.622 --> 00:05:18.063 That's his famous concept of satisfying. 95 00:05:18.824 --> 00:05:24.327 But also a German-American psychologist, 96 00:05:24.507 --> 00:05:25.407 Egon Brunswick, 97 00:05:25.467 --> 00:05:26.428 who is less known. 98 00:05:28.789 --> 00:05:31.631 emphasized that to understand behavior, 99 00:05:31.771 --> 00:05:32.652 just like Simon's, 100 00:05:33.352 --> 00:05:34.794 we need to look at the world, 101 00:05:34.934 --> 00:05:36.094 the structure of the world, 102 00:05:36.175 --> 00:05:43.780 not just inside about traits or risk properties and risk diversion and such things. 103 00:05:46.342 --> 00:05:48.564 During my life, 104 00:05:51.606 --> 00:05:55.789 I benefited very much from the research group at the Max Planck Institute. 105 00:05:57.369 --> 00:06:04.633 And that was between around 30 or 35 researchers, 106 00:06:04.673 --> 00:06:10.336 from graduate students to what would have been the US system associate professors. 107 00:06:11.397 --> 00:06:19.361 And where we had for many years an open culture where ideas could be discussed, 108 00:06:19.381 --> 00:06:24.044 where nobody had anxieties about saying, 109 00:06:24.144 --> 00:06:25.945 what's the evidence for that point? 110 00:06:27.041 --> 00:06:28.161 But at the same time, 111 00:06:28.161 --> 00:06:33.104 the group was a big family and we still meet every year. 112 00:06:34.144 --> 00:06:35.945 And I'm just coming back from Barcelona, 113 00:06:36.445 --> 00:06:37.285 the last meeting. 114 00:06:37.805 --> 00:06:40.126 So much of the things 115 00:06:40.787 --> 00:06:42.407 I learned and 116 00:06:43.588 --> 00:06:50.230 I rediscovered were products from many people discussing every time. 117 00:06:51.891 --> 00:06:56.133 And the social aspect of ideas is... 118 00:06:56.553 --> 00:06:57.233 Very important. 119 00:06:58.394 --> 00:06:58.514 Yes. 120 00:06:58.534 --> 00:06:58.654 So, 121 00:06:58.754 --> 00:06:59.214 Professor 122 00:06:59.594 --> 00:07:02.995 Gigerenzer, as I mentioned earlier in your recent book, 123 00:07:03.535 --> 00:07:04.275 Smart Management, 124 00:07:04.355 --> 00:07:07.136 How Leaders Make Decisions in Uncertain Worlds, 125 00:07:07.896 --> 00:07:09.137 which has just been published, 126 00:07:09.177 --> 00:07:12.257 is co-authored with Jochen Reb and Seng Kualuang. 127 00:07:13.398 --> 00:07:15.418 And it will be at the center of the conversation today. 128 00:07:15.818 --> 00:07:18.619 But before we move into the content of it, 129 00:07:19.339 --> 00:07:22.600 can you just tell us what was the inspiration behind it? 130 00:07:23.000 --> 00:07:24.861 Why writing for managers? 131 00:07:25.601 --> 00:07:27.262 What was the idea behind the book? 132 00:07:28.262 --> 00:07:34.464 Management is an ideal topic of how to make decisions under uncertainty. 133 00:07:35.484 --> 00:07:35.624 Now, 134 00:07:35.804 --> 00:07:46.087 uncertainty means that you do not know the complete set of events that might happen in the future, 135 00:07:46.387 --> 00:07:47.467 nor the consequences. 136 00:07:49.208 --> 00:07:50.628 So that's management. 137 00:07:51.588 --> 00:07:52.569 That's most of life. 138 00:07:53.569 --> 00:07:56.231 as opposed to what 139 00:07:57.512 --> 00:07:59.633 Jimmy Savage called a small world, 140 00:08:00.834 --> 00:08:06.638 where you know all the future events and their consequences and their probabilities. 141 00:08:07.439 --> 00:08:09.780 So management is not a lottery. 142 00:08:12.022 --> 00:08:20.327 And so that work came out from the earlier work of the ABC research group, 143 00:08:20.788 --> 00:08:22.469 that is ABC is so adaptive. 144 00:08:22.901 --> 00:08:24.542 behavior and cognition. 145 00:08:25.783 --> 00:08:33.629 And adaptive means that you need to tune your heuristics to the problem at hand. 146 00:08:34.349 --> 00:08:37.131 There is no single hammer that works for everything. 147 00:08:38.252 --> 00:08:43.195 And my two co-authors were part of the research group, 148 00:08:44.356 --> 00:08:46.678 and they are now in Beijing and Singapore. 149 00:08:47.438 --> 00:08:48.979 So we wrote this thing together. 150 00:08:49.019 --> 00:08:50.600 We wanted to do it more personally, 151 00:08:50.680 --> 00:08:51.441 but it was... 152 00:08:51.957 --> 00:08:52.277 COVID-19, 153 00:08:54.138 --> 00:08:58.661 but we managed that to meet often enough. 154 00:08:59.621 --> 00:08:59.841 And 155 00:09:00.862 --> 00:09:08.566 I think this book brings very concrete examples, 156 00:09:08.706 --> 00:09:15.370 very concrete heuristics and discusses when it's a good thing to apply them and when not. 157 00:09:16.811 --> 00:09:20.393 And it provides an alternative to the usual book. 158 00:09:20.969 --> 00:09:27.776 A curricula in business schools where you learn that how should you make good decisions? 159 00:09:28.156 --> 00:09:30.178 Expected utility maximization. 160 00:09:30.759 --> 00:09:32.781 How should you not make them? 161 00:09:33.302 --> 00:09:34.663 Cognitive illusions. 162 00:09:35.304 --> 00:09:35.644 Errors. 163 00:09:36.625 --> 00:09:41.290 But then you don't know what to do because you can't maximize in the real world, 164 00:09:42.131 --> 00:09:43.072 in a VUCA world, 165 00:09:43.232 --> 00:09:44.413 and... 166 00:09:45.217 --> 00:09:47.118 If you just point out what's going wrong, 167 00:09:47.418 --> 00:09:48.938 you don't know what to do right. 168 00:09:49.838 --> 00:09:51.219 And that's what this book, 169 00:09:52.319 --> 00:09:52.439 so 170 00:09:53.139 --> 00:09:53.939 The Blind Spot, 171 00:09:54.199 --> 00:09:54.760 it fills. 172 00:09:55.720 --> 00:09:56.080 Thank you. 173 00:09:57.360 --> 00:09:58.241 So for managers, 174 00:09:58.581 --> 00:10:03.202 what would be the one key learning or benefit in trusting intuition? 175 00:10:04.702 --> 00:10:05.163 So first, 176 00:10:06.763 --> 00:10:09.924 the book teaches a few principles, 177 00:10:09.984 --> 00:10:12.685 which is take uncertainty seriously. 178 00:10:13.725 --> 00:10:14.005 Often. 179 00:10:14.569 --> 00:10:17.310 Uncertainty is reduced to risk, 180 00:10:17.510 --> 00:10:19.971 at least in most economic models, 181 00:10:20.571 --> 00:10:25.052 but also by many behavioral economists who assume that is the right way to do, 182 00:10:26.392 --> 00:10:29.133 that is right in a world of known risk, 183 00:10:30.053 --> 00:10:33.674 and that would be immediately applied to the real world of MOOC. 184 00:10:35.075 --> 00:10:37.315 And then take heuristics seriously, 185 00:10:38.756 --> 00:10:43.617 which means to be a little bit more humble. 186 00:10:44.169 --> 00:10:48.250 and accept that in much of the real world you can't optimize. 187 00:10:49.310 --> 00:10:53.051 And maximizing expected utility and maximum is just a version of that. 188 00:10:53.732 --> 00:10:56.672 So what do you really do? 189 00:10:56.813 --> 00:11:00.734 And then the book shows there are many ways to satisfy us. 190 00:11:01.154 --> 00:11:01.614 For instance, 191 00:11:01.674 --> 00:11:05.275 many ways to hire a person, 192 00:11:07.435 --> 00:11:10.016 to decide between the applicants, 193 00:11:10.756 --> 00:11:12.777 and think about that. 194 00:11:14.117 --> 00:11:20.082 And learn the adaptive toolbox of heuristics and learn wisely to select them. 195 00:11:20.982 --> 00:11:22.423 And here you need experience. 196 00:11:23.284 --> 00:11:26.346 While in the traditional proof of expected utility marks, 197 00:11:26.506 --> 00:11:27.767 you don't need experience. 198 00:11:28.147 --> 00:11:29.308 You don't need to learn. 199 00:11:29.608 --> 00:11:30.829 You just do a calculation. 200 00:11:32.510 --> 00:11:32.990 Professor, 201 00:11:33.351 --> 00:11:36.713 I would like to start with a key concept from your book. 202 00:11:37.093 --> 00:11:38.774 And first of all, 203 00:11:38.834 --> 00:11:42.377 with the concept of simple heuristics. 204 00:11:43.173 --> 00:11:50.718 Could you explain what you mean by a simple heuristic or what you call smart heuristics? 205 00:11:51.819 --> 00:11:52.339 Let's take 206 00:11:53.020 --> 00:11:54.000 Henry Morkowitz, 207 00:11:54.000 --> 00:11:54.661 who won his 208 00:11:55.341 --> 00:11:59.884 Economics Nobel Prize for an optimization method. 209 00:12:00.405 --> 00:12:08.110 So it's known as an optimal portfolio for the question, 210 00:12:08.150 --> 00:12:09.371 you have N assets, 211 00:12:09.591 --> 00:12:12.173 how do you invest in there? 212 00:12:14.157 --> 00:12:16.879 It's known as the mean variance portfolio. 213 00:12:17.759 --> 00:12:17.979 When 214 00:12:18.660 --> 00:12:23.563 Harry Markowitz made his own investments for the time of his retirement, 215 00:12:24.143 --> 00:12:29.286 then we might assume he used his Nobel Prize winning optimization method. 216 00:12:29.906 --> 00:12:30.887 He did not. 217 00:12:31.847 --> 00:12:33.328 He used a simple heuristic. 218 00:12:34.129 --> 00:12:34.909 In that case, 219 00:12:35.650 --> 00:12:36.790 the heuristic is called 220 00:12:37.231 --> 00:12:38.631 1 over n. 221 00:12:39.472 --> 00:12:41.213 n is the number of assets. 222 00:12:42.257 --> 00:12:44.540 And it means divide equally. 223 00:12:44.920 --> 00:12:46.642 So if you have two options, 224 00:12:47.323 --> 00:12:48.364 then 50-50. 225 00:12:48.844 --> 00:12:49.164 Three, 226 00:12:49.365 --> 00:12:49.805 a third, 227 00:12:49.825 --> 00:12:50.226 a third, 228 00:12:50.286 --> 00:12:50.746 and so on. 229 00:12:51.247 --> 00:12:52.148 That's a heuristic. 230 00:12:53.609 --> 00:12:57.233 The optimization portfolio is not a heuristic. 231 00:12:58.294 --> 00:13:00.196 It requires extensive data, 232 00:13:00.757 --> 00:13:02.138 extensive calculation, 233 00:13:03.319 --> 00:13:04.240 and estimation. 234 00:13:05.577 --> 00:13:05.697 So, 235 00:13:06.978 --> 00:13:07.379 studies, 236 00:13:07.459 --> 00:13:07.859 by the way, 237 00:13:07.999 --> 00:13:12.903 show that in many investment situations, 238 00:13:13.023 --> 00:13:15.185 1 over n makes more money, 239 00:13:16.266 --> 00:13:17.667 measured by Sharpe ratio, 240 00:13:17.767 --> 00:13:20.729 by other traditional criteria, 241 00:13:21.310 --> 00:13:23.672 than the Markowitz portfolio. 242 00:13:24.152 --> 00:13:24.573 And also, 243 00:13:25.533 --> 00:13:29.136 many modern versions cannot systematically do better. 244 00:13:30.337 --> 00:13:30.457 So, 245 00:13:31.258 --> 00:13:32.779 this is an example of a heuristic. 246 00:13:33.960 --> 00:13:36.842 And many people use the same heuristic for other things. 247 00:13:36.902 --> 00:13:37.322 For instance, 248 00:13:37.362 --> 00:13:44.247 parents with two or more children try to divide their love and time equally. 249 00:13:44.567 --> 00:13:45.708 That's one over n. 250 00:13:46.688 --> 00:13:48.289 And the heuristics they carry, 251 00:13:48.930 --> 00:13:49.931 as one can see here, 252 00:13:50.451 --> 00:13:52.212 often sense of fairness. 253 00:13:53.193 --> 00:13:53.593 With them, 254 00:13:54.254 --> 00:13:55.334 an equal division. 255 00:13:55.895 --> 00:13:58.937 And the point is that one needs to... 256 00:13:59.297 --> 00:14:03.039 study how well they do in, 257 00:14:03.059 --> 00:14:03.579 for instance, 258 00:14:03.920 --> 00:14:04.440 accuracy. 259 00:14:05.341 --> 00:14:05.461 So, 260 00:14:05.961 --> 00:14:12.585 what's often called naive diversification can do very well under UNSREC. 261 00:14:14.066 --> 00:14:17.468 Could you give us some concrete examples? 262 00:14:19.989 --> 00:14:20.189 Here, 263 00:14:20.209 --> 00:14:23.031 I'll give you an example of three heuristics. 264 00:14:24.692 --> 00:14:26.113 Let's start with Elon Musk. 265 00:14:27.081 --> 00:14:30.062 When Elon Musk was young and Tesla was young, 266 00:14:30.882 --> 00:14:39.365 he reported that he made the hiring procedure and he relied on a single criterion, 267 00:14:41.206 --> 00:14:41.626 which is, 268 00:14:43.007 --> 00:14:45.748 does the person have an exceptional ability? 269 00:14:46.468 --> 00:14:46.848 If no, 270 00:14:47.768 --> 00:14:48.429 not hired. 271 00:14:48.789 --> 00:14:49.249 If yes, 272 00:14:49.629 --> 00:14:50.009 hired. 273 00:14:50.789 --> 00:14:52.470 This is an extreme heuristic, 274 00:14:53.150 --> 00:14:55.291 because it just looks at one. 275 00:14:56.083 --> 00:14:58.444 variable or one feature. 276 00:14:59.685 --> 00:14:59.885 Now, 277 00:15:00.625 --> 00:15:04.246 you might think that it is a bit irrational. 278 00:15:06.487 --> 00:15:13.270 At least it would look from the heuristics and biases program that he might have cognitive limitations or something like that. 279 00:15:13.370 --> 00:15:14.030 But no, 280 00:15:14.931 --> 00:15:16.571 one needs to look at this and study it. 281 00:15:16.912 --> 00:15:17.032 So, 282 00:15:17.612 --> 00:15:18.492 in this case, 283 00:15:19.052 --> 00:15:24.995 there is one variable that is strongly correlated with a number of other variables. 284 00:15:26.055 --> 00:15:26.876 For instance, 285 00:15:27.436 --> 00:15:29.457 if someone has an exceptional ability, 286 00:15:29.597 --> 00:15:33.379 and even if this person is an excellent musician, 287 00:15:33.499 --> 00:15:37.202 which is not necessarily important for Tesla, 288 00:15:38.662 --> 00:15:42.324 but that means the person is likely able to concentrate, 289 00:15:43.025 --> 00:15:44.366 to stay on the problem, 290 00:15:44.886 --> 00:15:45.486 to sweat, 291 00:15:46.087 --> 00:15:46.987 to persevere, 292 00:15:47.968 --> 00:15:51.290 to be able to work with others together, 293 00:15:51.370 --> 00:15:52.991 like in an orchestra. 294 00:15:54.011 --> 00:15:54.131 So, 295 00:15:55.312 --> 00:16:00.475 finding an excellent variable is often the real question. 296 00:16:01.275 --> 00:16:05.217 And that brings along many others that are here. 297 00:16:05.898 --> 00:16:06.018 So, 298 00:16:06.078 --> 00:16:08.399 this is a heuristic from a broader class. 299 00:16:09.120 --> 00:16:12.121 There is one clever Q heuristic. 300 00:16:12.261 --> 00:16:14.422 Find an important aspect. 301 00:16:15.423 --> 00:16:15.723 Now, 302 00:16:16.504 --> 00:16:17.184 let's move on. 303 00:16:17.624 --> 00:16:18.685 A second example. 304 00:16:20.046 --> 00:16:20.166 So, 305 00:16:20.266 --> 00:16:21.106 Jeff Bezos, 306 00:16:21.747 --> 00:16:22.167 Amazon. 307 00:16:22.811 --> 00:16:26.132 He reported that when Amazon was small, 308 00:16:26.252 --> 00:16:27.213 he did the hiring, 309 00:16:28.313 --> 00:16:30.574 but he did it slightly different from Musk. 310 00:16:31.194 --> 00:16:36.997 He used a different heuristic that we call a fast and frugal tree. 311 00:16:37.617 --> 00:16:37.837 Okay, 312 00:16:37.897 --> 00:16:38.657 let me explain. 313 00:16:39.618 --> 00:16:39.738 So, 314 00:16:40.458 --> 00:16:43.099 it looks at most at three variables, 315 00:16:43.219 --> 00:16:43.940 not just one. 316 00:16:44.440 --> 00:16:45.240 And interestingly, 317 00:16:45.280 --> 00:16:48.021 the first one was the same as Musk. 318 00:16:48.401 --> 00:16:48.682 Namely, 319 00:16:49.102 --> 00:16:51.343 does the person have an exception ability? 320 00:16:51.423 --> 00:16:51.823 If no... 321 00:16:52.951 --> 00:16:53.511 No higher. 322 00:16:54.512 --> 00:16:55.192 If yes, 323 00:16:55.832 --> 00:16:56.713 that's not enough. 324 00:16:57.673 --> 00:16:58.313 For Bezos, 325 00:16:58.773 --> 00:17:00.334 a second question is asked, 326 00:17:00.614 --> 00:17:00.894 namely, 327 00:17:01.595 --> 00:17:04.216 can I admire this person, 328 00:17:04.636 --> 00:17:06.537 which is an unusual question. 329 00:17:09.058 --> 00:17:09.278 But 330 00:17:09.718 --> 00:17:10.378 Bezos said, 331 00:17:11.078 --> 00:17:12.559 if I admire a person, 332 00:17:12.939 --> 00:17:14.740 I will learn from that person, 333 00:17:15.020 --> 00:17:16.241 and that's important for him. 334 00:17:17.601 --> 00:17:17.721 So, 335 00:17:18.482 --> 00:17:19.342 if the person... 336 00:17:19.362 --> 00:17:20.242 Eh. 337 00:17:22.543 --> 00:17:25.085 If he thinks he cannot admire the person, 338 00:17:25.506 --> 00:17:26.126 no hire. 339 00:17:27.727 --> 00:17:28.228 But yes, 340 00:17:28.448 --> 00:17:30.830 a third and last question is asked, 341 00:17:30.850 --> 00:17:36.755 will the person improve the average quality in the unit he or she will be in? 342 00:17:37.576 --> 00:17:41.299 And that's a reasonable question because by hiring in this way, 343 00:17:41.859 --> 00:17:44.101 you always improve the group. 344 00:17:44.642 --> 00:17:46.043 And only then it's been hired. 345 00:17:46.744 --> 00:17:50.687 So this can be seen as a decision tree that is not complete. 346 00:17:51.259 --> 00:17:53.120 whether the decision is in every point. 347 00:17:53.620 --> 00:17:55.121 So exception ability, 348 00:17:55.501 --> 00:17:55.981 yes or no? 349 00:17:56.781 --> 00:17:57.382 And if yes, 350 00:17:57.822 --> 00:17:59.643 then admire, 351 00:18:00.443 --> 00:18:00.883 yes or no? 352 00:18:01.123 --> 00:18:01.864 And if no, 353 00:18:02.264 --> 00:18:05.805 then will it raise the proof? 354 00:18:07.146 --> 00:18:12.649 Note that these fast and frugally trees are not like in a full tree, 355 00:18:13.589 --> 00:18:14.849 and there's an order in it. 356 00:18:15.970 --> 00:18:17.531 So if you fail on the first one, 357 00:18:18.367 --> 00:18:20.569 the other two will not compensate. 358 00:18:20.949 --> 00:18:23.831 That's very different for most rational decision models, 359 00:18:24.212 --> 00:18:25.553 where you always can compensate. 360 00:18:27.575 --> 00:18:34.660 A second concept that is fundamental in your thinking is ecological rationality. 361 00:18:35.541 --> 00:18:44.849 Could you define what ecological rationality is and why this concept is fundamental for making good decisions? 362 00:18:45.869 --> 00:18:46.710 Yeah. 363 00:18:47.431 --> 00:18:47.551 The... 364 00:18:48.615 --> 00:18:49.595 As you may see, 365 00:18:50.276 --> 00:18:55.598 my approach to decision-making is to look precisely how a decision is being made. 366 00:18:56.459 --> 00:19:00.060 The standard concept of rationality is about consistency. 367 00:19:01.441 --> 00:19:01.561 So, 368 00:19:02.281 --> 00:19:07.323 whether it's transitive and the consistency axioms. 369 00:19:08.304 --> 00:19:15.107 And the ecological rationality is a term that it is... 370 00:19:16.291 --> 00:19:20.073 basically maps into Simon's concept of bounded rationality, 371 00:19:21.394 --> 00:19:24.576 which is not what Kahneman called bounded rationality, 372 00:19:25.997 --> 00:19:30.199 because that's also the reason why we use the term ecological rationality. 373 00:19:30.199 --> 00:19:31.740 It means it's a functional term. 374 00:19:32.961 --> 00:19:34.241 How good is this heuristic? 375 00:19:34.702 --> 00:19:36.983 To what degree is it adapted to the problem? 376 00:19:38.544 --> 00:19:40.725 So let me illustrate. 377 00:19:41.285 --> 00:19:44.207 In the case of Elon Musk, 378 00:19:45.083 --> 00:19:48.964 where the hiring heuristic is just one variable. 379 00:19:49.504 --> 00:19:51.625 Has this person an acceptability? 380 00:19:52.786 --> 00:19:53.086 That, 381 00:19:53.706 --> 00:20:01.388 now you can do some mathematics and see that whether this Q, 382 00:20:03.069 --> 00:20:03.369 whether, 383 00:20:04.289 --> 00:20:04.889 put it this way, 384 00:20:05.290 --> 00:20:06.170 if you add more, 385 00:20:07.570 --> 00:20:08.730 variables to this queue, 386 00:20:08.891 --> 00:20:09.691 does it improve? 387 00:20:11.891 --> 00:20:15.312 And if you imagine now, 388 00:20:16.012 --> 00:20:16.193 say, 389 00:20:17.733 --> 00:20:22.354 the first variable has this validity in terms of a linear model, 390 00:20:22.714 --> 00:20:25.295 and the second one has only half of that, 391 00:20:25.335 --> 00:20:27.115 and the third additional contribution, 392 00:20:27.176 --> 00:20:27.956 like beta weights, 393 00:20:28.596 --> 00:20:31.397 then you can prove that in this situation, 394 00:20:33.177 --> 00:20:36.278 looking at more variables will improve. 395 00:20:36.498 --> 00:20:40.020 not improve the decision. 396 00:20:40.020 --> 00:20:46.684 It even can make the decision worse because you have getting more and more error. 397 00:20:47.264 --> 00:20:54.488 So ecological rationality studies the condition under which certain heuristics work and do not work. 398 00:20:55.549 --> 00:21:04.174 And this is a new discipline because it has been assumed by at least still it's been assumed by the mainstream mainstream 399 00:21:04.722 --> 00:21:10.906 behavioral economists that a heuristic is always second class and optimization is always better. 400 00:21:11.966 --> 00:21:12.126 Now, 401 00:21:13.147 --> 00:21:14.448 in a local world, 402 00:21:14.628 --> 00:21:15.928 you can't optimize, 403 00:21:16.189 --> 00:21:16.629 period. 404 00:21:17.669 --> 00:21:19.911 That's an illusion to think this way. 405 00:21:20.331 --> 00:21:22.012 You have to ask another question. 406 00:21:22.392 --> 00:21:23.513 There are many heuristics. 407 00:21:24.373 --> 00:21:27.795 Which one of that will work in this situation? 408 00:21:28.996 --> 00:21:32.498 And if you have an exponentially decreasing 409 00:21:33.899 --> 00:21:34.319 weights, 410 00:21:34.920 --> 00:21:40.204 then you can just go with the best reason and ignore everything else. 411 00:21:40.625 --> 00:21:41.806 If that's not the case, 412 00:21:43.087 --> 00:21:46.730 if the weights of the variables are rather flat, 413 00:21:47.711 --> 00:21:50.493 then procedures like 414 00:21:50.973 --> 00:21:57.659 1 over n or tallying that you're giving don't estimate weights, 415 00:21:57.739 --> 00:22:00.482 but just doing unit weights just can't. 416 00:22:01.570 --> 00:22:07.613 avoid error in estimation and can be better than regression models. 417 00:22:09.893 --> 00:22:13.695 And these are effects that we call less is more. 418 00:22:14.035 --> 00:22:16.516 And less is more happens under uncertainty. 419 00:22:17.517 --> 00:22:20.438 It doesn't happen in the world of, 420 00:22:21.398 --> 00:22:23.499 in the small world of risk models. 421 00:22:24.840 --> 00:22:30.762 The third concept that seems fundamentally in your thinking is the adaptive. 422 00:22:30.882 --> 00:22:31.342 toolbox. 423 00:22:32.163 --> 00:22:32.543 Again, 424 00:22:32.763 --> 00:22:36.266 could you explain this idea of the adaptive toolbox? 425 00:22:37.807 --> 00:22:42.009 How do you think humans make decisions using this toolbox? 426 00:22:43.110 --> 00:22:45.272 As an adaptive toolbox of heuristics, 427 00:22:45.392 --> 00:22:47.753 they can be used consciously or unconsciously. 428 00:22:48.374 --> 00:22:49.655 Take the example of hiring. 429 00:22:50.415 --> 00:22:50.915 Now we have 430 00:22:51.356 --> 00:22:54.178 Musk's one reason hiring rule. 431 00:22:54.358 --> 00:22:58.721 We have Jeff Bezos'fast and frugal three with up to three reasons. 432 00:22:59.301 --> 00:23:00.462 Then you can add a few. 433 00:23:01.054 --> 00:23:01.534 For instance, 434 00:23:01.595 --> 00:23:03.456 a number of heuristics are social, 435 00:23:04.417 --> 00:23:10.302 and a social heuristic for hiring would be word of mouth. 436 00:23:10.342 --> 00:23:12.123 Word of mouth is often used, 437 00:23:12.664 --> 00:23:14.585 so you ask your own employees, 438 00:23:15.106 --> 00:23:19.890 do you know someone who would be a good person for that job? 439 00:23:21.031 --> 00:23:25.555 And there are studies that show that in a healthy company, 440 00:23:26.496 --> 00:23:28.838 word of mouth is hard to beat. 441 00:23:30.458 --> 00:23:31.500 And you can see why. 442 00:23:32.040 --> 00:23:41.833 Because the person who recommends another person to the job feels responsible him or herself. 443 00:23:42.954 --> 00:23:47.060 I will not recommend someone who works less than he or she. 444 00:23:51.066 --> 00:23:54.769 So that's an example of the adaptive toolbox. 445 00:23:55.869 --> 00:24:01.613 And then you can classify the heuristics in broader classes we've just seen. 446 00:24:02.354 --> 00:24:04.055 One is social heuristics, 447 00:24:04.195 --> 00:24:05.056 like imitation. 448 00:24:05.236 --> 00:24:08.458 Very important for innovation in businesses. 449 00:24:09.299 --> 00:24:12.781 Much of innovation is copying and editing. 450 00:24:14.882 --> 00:24:18.405 It's not that Facebook invented its Facebook. 451 00:24:19.386 --> 00:24:19.946 It copied. 452 00:24:20.794 --> 00:24:21.715 or Google copy. 453 00:24:23.196 --> 00:24:32.524 And so these type of heuristics can be systematically studied and their ecological rationality investigated. 454 00:24:34.466 --> 00:24:38.349 And that's much harder than saying you just maximize some utility. 455 00:24:40.102 --> 00:24:44.925 This is one of the central elements of the book and more generally of your thinking. 456 00:24:45.566 --> 00:24:51.910 And you highlight particularly the underestimation of the importance of intuition in decision making. 457 00:24:52.390 --> 00:24:53.931 And just for the audience to know, 458 00:24:53.991 --> 00:24:59.435 you've also recently devoted another book called The Intelligence of Intuition. 459 00:25:00.136 --> 00:25:05.119 So one thing I would like to know is what is your definition of intuition? 460 00:25:05.279 --> 00:25:08.301 What do you have in mind when you talk about intuition? 461 00:25:08.301 --> 00:25:09.262 Because I think you have very... 462 00:25:10.042 --> 00:25:11.743 specific definition of it. 463 00:25:12.903 --> 00:25:16.125 So intuition is a form of unconscious intelligence. 464 00:25:16.865 --> 00:25:21.467 So it is a feeling that has three components. 465 00:25:21.667 --> 00:25:24.929 One is it's based on long experience. 466 00:25:25.989 --> 00:25:26.349 Otherwise, 467 00:25:26.369 --> 00:25:27.610 there's no intuition. 468 00:25:28.550 --> 00:25:30.931 So there may be a year-long experience with the subject. 469 00:25:31.211 --> 00:25:31.531 Second, 470 00:25:33.111 --> 00:25:37.953 you sense very quickly what you should do or not do. 471 00:25:39.073 --> 00:25:39.713 And third, 472 00:25:41.113 --> 00:25:48.195 you cannot explain why you sense that you should do that or not. 473 00:25:49.096 --> 00:25:54.557 So intuition is not a sixth sense or not a God's voice. 474 00:25:54.977 --> 00:25:57.018 It's also not something that... 475 00:25:58.018 --> 00:25:59.479 women have, 476 00:25:59.659 --> 00:26:00.699 and may not, 477 00:26:00.899 --> 00:26:01.980 everybody has intuition, 478 00:26:02.400 --> 00:26:09.863 who works for sufficient time in a certain topic where there's also feedback. 479 00:26:11.944 --> 00:26:17.006 And you are right saying that in decision-making theory, 480 00:26:17.146 --> 00:26:21.068 intuition is looked at suspicious, 481 00:26:21.748 --> 00:26:24.069 but not in many other fields. 482 00:26:25.210 --> 00:26:25.330 So, 483 00:26:25.330 --> 00:26:25.710 you're right. 484 00:26:26.670 --> 00:26:29.651 Just think about mathematics or physics, 485 00:26:31.231 --> 00:26:32.832 where intuition is respected. 486 00:26:34.392 --> 00:26:35.172 Einstein said, 487 00:26:36.793 --> 00:26:40.174 intuition is a gift. 488 00:26:40.714 --> 00:26:43.335 So the intuitive spirit, 489 00:26:43.395 --> 00:26:43.815 he said, 490 00:26:44.315 --> 00:26:45.455 is a gift, 491 00:26:45.915 --> 00:26:49.976 and the rational spirit is its servant. 492 00:26:50.857 --> 00:26:55.598 And we have created a society that honors the servant and has forgotten. 493 00:26:55.798 --> 00:26:56.418 the gift. 494 00:26:56.958 --> 00:26:57.979 Just an example. 495 00:26:58.759 --> 00:27:03.300 If you talk with the best chess players in the world, 496 00:27:03.420 --> 00:27:05.741 Judas Polgar or Magnus Carlsen, 497 00:27:06.281 --> 00:27:15.324 they emphasize that their excellent play is a mixer between intuition and deliberate learning. 498 00:27:16.724 --> 00:27:18.765 And this is the first important point. 499 00:27:20.365 --> 00:27:20.665 Namely, 500 00:27:20.965 --> 00:27:22.866 intuition is not a post. 501 00:27:23.350 --> 00:27:25.451 to deliberate thinking, 502 00:27:26.171 --> 00:27:29.532 as it's still assumed in much of behavioral economics, 503 00:27:29.892 --> 00:27:31.872 where you have a system one and a system two. 504 00:27:32.532 --> 00:27:32.653 No, 505 00:27:32.753 --> 00:27:32.873 no, 506 00:27:33.393 --> 00:27:34.853 it goes hand in hand. 507 00:27:36.253 --> 00:27:36.374 So, 508 00:27:37.254 --> 00:27:38.314 another example, 509 00:27:38.674 --> 00:27:42.015 a doctor who sees you often, 510 00:27:42.775 --> 00:27:46.876 and today the doctor thinks something is wrong with you, 511 00:27:48.217 --> 00:27:50.537 but cannot explain what's wrong. 512 00:27:51.658 --> 00:27:52.558 That's an intuition. 513 00:27:53.766 --> 00:27:54.846 Long experience, 514 00:27:55.447 --> 00:27:58.667 it's quickly in consciousness, 515 00:27:59.288 --> 00:28:00.788 and the doctor cannot explain it. 516 00:28:01.928 --> 00:28:05.329 And then the doctor will move on and do diagnostics. 517 00:28:06.229 --> 00:28:07.870 But there's no contradiction there. 518 00:28:08.410 --> 00:28:13.331 It's not an either-or that you should stop all your intuition, 519 00:28:13.411 --> 00:28:14.612 as we are told. 520 00:28:16.152 --> 00:28:20.033 And that is an important insight. 521 00:28:20.333 --> 00:28:22.474 There will be almost no error. 522 00:28:23.206 --> 00:28:27.249 where you will do well without intuition. 523 00:28:28.810 --> 00:28:29.631 As you can add, 524 00:28:29.671 --> 00:28:34.354 there will be almost no error where you will do well without thinking. 525 00:28:35.494 --> 00:28:37.096 And it's not a contradiction. 526 00:28:37.976 --> 00:28:38.356 Interesting, 527 00:28:38.356 --> 00:28:39.157 because in management, 528 00:28:39.697 --> 00:28:47.863 sometimes we have an ambiguous point of view where managers are also invited to overcome their cognitive biases. 529 00:28:48.744 --> 00:28:51.465 So what's your thinking? 530 00:28:52.466 --> 00:29:08.530 against these approaches that were into can intuition actually be leveraged while avoiding pitfalls yeah of course so i'm not going here to criticize much the heuristic and biases approach 531 00:29:08.590 --> 00:29:21.894 i've done this many times so just make sure many of the so-called buyers buyers as we know today aren't any biases and they they are 532 00:29:22.390 --> 00:29:23.651 If you ignore information, 533 00:29:24.291 --> 00:29:25.591 then maybe something good or not. 534 00:29:25.631 --> 00:29:27.452 The question is ecological rationality. 535 00:29:28.132 --> 00:29:28.393 So when 536 00:29:28.873 --> 00:29:33.195 Harry Markowitz ignores his own optimizing portfolio, 537 00:29:33.935 --> 00:29:34.655 he may be right. 538 00:29:36.476 --> 00:29:36.616 When 539 00:29:37.857 --> 00:29:40.558 Elon Musk just relies on one variable, 540 00:29:40.838 --> 00:29:41.578 he may be right. 541 00:29:41.958 --> 00:29:42.979 Depends on the situation. 542 00:29:43.859 --> 00:29:48.181 So the juristic and biases tradition needs an ecological perspective. 543 00:29:49.038 --> 00:29:52.701 It's not true that being consistent would be always right. 544 00:29:54.041 --> 00:29:57.224 And using a heuristic loophole was such a first thing. 545 00:29:58.985 --> 00:30:00.606 Intuition is, 546 00:30:01.487 --> 00:30:03.688 I have worked with many large companies. 547 00:30:04.789 --> 00:30:06.130 And the 548 00:30:06.530 --> 00:30:09.612 CEOs and the top leaders of large companies, 549 00:30:11.653 --> 00:30:15.216 when in the studies with these companies, 550 00:30:17.097 --> 00:30:18.238 you will find that 551 00:30:18.746 --> 00:30:19.706 They say that 552 00:30:21.067 --> 00:30:27.568 50% of the important professional decisions is at the end a gut decision, 553 00:30:27.848 --> 00:30:28.809 so an intuition. 554 00:30:31.429 --> 00:30:44.613 You get these if you do an anonymous service with them or in some cases we use a person high up in the company who 555 00:30:44.653 --> 00:30:48.154 has the trust of everyone and can just openly talk. 556 00:30:49.298 --> 00:30:58.422 The same executives who make about every other question at the end intuitive, 557 00:30:58.862 --> 00:31:00.543 and the emphasis is on the end. 558 00:31:00.843 --> 00:31:02.563 It's not arbitrary. 559 00:31:03.364 --> 00:31:04.724 So they sit on data, 560 00:31:05.825 --> 00:31:06.945 and there's too much data, 561 00:31:07.185 --> 00:31:09.006 and they don't know how reliable it is. 562 00:31:09.346 --> 00:31:10.387 And in many situations, 563 00:31:10.387 --> 00:31:11.527 the data gives you an answer, 564 00:31:11.607 --> 00:31:12.827 and many others, 565 00:31:13.088 --> 00:31:13.208 no. 566 00:31:14.228 --> 00:31:18.010 And then what's an intuitive decision is that you then, 567 00:31:18.430 --> 00:31:18.710 after, 568 00:31:19.282 --> 00:31:21.464 the data can tell you, 569 00:31:23.065 --> 00:31:24.626 feel that you shouldn't do that. 570 00:31:26.908 --> 00:31:27.028 So, 571 00:31:27.828 --> 00:31:28.469 in the studies, 572 00:31:28.509 --> 00:31:29.029 as I said, 573 00:31:30.831 --> 00:31:33.793 in each of the large companies I've worked with, 574 00:31:33.813 --> 00:31:35.114 and these are worldwide companies, 575 00:31:35.154 --> 00:31:37.555 about 50% of all decisions are, 576 00:31:37.615 --> 00:31:38.116 at the end, 577 00:31:38.496 --> 00:31:39.437 an intuitive one. 578 00:31:40.297 --> 00:31:41.538 And now the interesting point, 579 00:31:42.899 --> 00:31:46.002 the same executives would never admit that in public. 580 00:31:48.410 --> 00:31:49.070 There's fear. 581 00:31:50.411 --> 00:31:52.892 For an intuitive decision, 582 00:31:53.512 --> 00:31:55.233 you have to take responsibility. 583 00:31:56.373 --> 00:32:03.076 We live in a society where fewer and fewer executives are willing to take responsibility. 584 00:32:03.757 --> 00:32:04.517 So what do they do? 585 00:32:05.737 --> 00:32:07.938 They just made an intuitive decision, 586 00:32:08.759 --> 00:32:12.700 but they can't announce it as one. 587 00:32:13.181 --> 00:32:13.301 So, 588 00:32:15.362 --> 00:32:16.022 one version is... 589 00:32:16.590 --> 00:32:22.775 They take a middle manager and then would ask him, 590 00:32:23.296 --> 00:32:24.456 you have a week or two, 591 00:32:25.337 --> 00:32:26.178 find me the reasons. 592 00:32:29.460 --> 00:32:30.561 And that's a loss of time, 593 00:32:31.742 --> 00:32:33.844 intelligence and resources. 594 00:32:34.805 --> 00:32:45.393 The more expensive version of that is the company or the top management hires a consulting firm. 595 00:32:46.754 --> 00:32:54.976 That then will waste months on finding reasons for the already made decision. 596 00:32:57.657 --> 00:32:58.818 How often does it happen? 597 00:32:59.678 --> 00:33:08.020 I have worked with one of the largest worldwide consulting firms and asked the principal over lunch, 598 00:33:08.420 --> 00:33:13.622 would you be willing and tell me how often customer contacts involve? 599 00:33:14.462 --> 00:33:18.746 Justifying decisions that already have been made. 600 00:33:19.867 --> 00:33:20.288 He said, 601 00:33:20.388 --> 00:33:21.409 Professor Gigerenzer, 602 00:33:21.409 --> 00:33:22.169 if you don't tell me, 603 00:33:23.751 --> 00:33:24.772 if you don't tell my name, 604 00:33:24.792 --> 00:33:25.212 I will. 605 00:33:26.053 --> 00:33:27.514 It's over 50%. 606 00:33:28.635 --> 00:33:28.755 So, 607 00:33:29.756 --> 00:33:35.642 I tell you this story to illustrate the anxiety of admitting gut decisions. 608 00:33:36.022 --> 00:33:37.223 They are made, 609 00:33:38.104 --> 00:33:39.285 but there is an anxiety. 610 00:33:40.586 --> 00:33:44.450 And together with the anxiety of taking over responsibility, 611 00:33:45.851 --> 00:33:50.015 and then the waste of time and resources, 612 00:33:51.616 --> 00:33:58.623 only to pretend that decision has been made with only, 613 00:33:58.943 --> 00:34:00.244 with no intuition. 614 00:34:02.266 --> 00:34:03.447 And this is the world we live. 615 00:34:03.988 --> 00:34:05.129 Sounds familiar somehow. 616 00:34:07.090 --> 00:34:11.211 We do research and we can hear that very often. 617 00:34:11.852 --> 00:34:11.972 Now, 618 00:34:11.972 --> 00:34:16.573 there are still some major advantages in using intuition, 619 00:34:16.653 --> 00:34:20.034 sometimes against algorithm or complex systems. 620 00:34:21.234 --> 00:34:29.036 Would you highlight to managers why intuition can actually be a very good way to make decisions? 621 00:34:29.576 --> 00:34:32.477 What are the main advantages you would highlight? 622 00:34:33.678 --> 00:34:34.358 Yeah. 623 00:34:34.358 --> 00:34:34.878 So first... 624 00:34:35.298 --> 00:34:36.379 You don't waste money. 625 00:34:37.200 --> 00:34:38.662 You don't waste time. 626 00:34:41.645 --> 00:34:45.008 But you need courage to stand up and say, 627 00:34:45.649 --> 00:34:48.371 as a manager or as a politician, 628 00:34:48.992 --> 00:34:49.232 look, 629 00:34:50.574 --> 00:34:50.694 so... 630 00:34:53.002 --> 00:34:54.683 The honest version would be, 631 00:34:55.383 --> 00:34:56.743 in the cases I've written, 632 00:34:57.004 --> 00:35:01.965 not to hire a consulting firm and deceive yourself and everyone else, 633 00:35:02.465 --> 00:35:04.546 but to stand up and say, 634 00:35:04.786 --> 00:35:04.966 look, 635 00:35:06.166 --> 00:35:06.286 we, 636 00:35:06.966 --> 00:35:07.466 the board, 637 00:35:07.927 --> 00:35:11.928 have now spent a week over this decision. 638 00:35:12.168 --> 00:35:12.408 Maybe, 639 00:35:13.948 --> 00:35:15.149 should we buy this company, 640 00:35:15.269 --> 00:35:15.949 another company, 641 00:35:16.009 --> 00:35:16.429 or not? 642 00:35:17.309 --> 00:35:20.170 Or should we move to Vietnam? 643 00:35:21.158 --> 00:35:21.718 Something big. 644 00:35:23.759 --> 00:35:24.820 We have looked at data, 645 00:35:25.460 --> 00:35:26.661 a previous experience, 646 00:35:27.181 --> 00:35:28.422 and the data isn't clear. 647 00:35:30.603 --> 00:35:33.405 And there's no point now to go on with that. 648 00:35:34.105 --> 00:35:37.347 But someone has to take the responsibility, 649 00:35:37.667 --> 00:35:38.127 and I, 650 00:35:38.147 --> 00:35:38.808 as the CEO, 651 00:35:39.508 --> 00:35:39.988 have to do. 652 00:35:41.189 --> 00:35:41.309 So, 653 00:35:42.670 --> 00:35:44.711 based on my own experience, 654 00:35:46.232 --> 00:35:48.713 and on my intuition, 655 00:35:49.253 --> 00:35:50.654 I think we should not 656 00:35:50.914 --> 00:35:54.856 pursue this any first. 657 00:35:54.856 --> 00:35:56.317 So that would be the honest version. 658 00:35:58.018 --> 00:36:00.159 And it would save the company lots of time. 659 00:36:02.060 --> 00:36:06.623 And many of the problems being too slow inside the company, 660 00:36:06.803 --> 00:36:07.844 always the customers, 661 00:36:08.164 --> 00:36:12.866 comes from the anxiety of making decisions and prolonging. 662 00:36:12.866 --> 00:36:14.868 You can measure the, 663 00:36:15.528 --> 00:36:19.030 it's the technical term is defensive decision making. 664 00:36:20.282 --> 00:36:28.268 So you protect yourself and may even go with a decision that's not the best for the company. 665 00:36:32.452 --> 00:36:33.032 So what do you, 666 00:36:34.173 --> 00:36:38.156 defensive decision making and the lack of a good error culture, 667 00:36:39.457 --> 00:36:42.720 and that is what hurts many companies. 668 00:36:43.180 --> 00:36:45.882 You do not find it as much in family companies. 669 00:36:47.938 --> 00:36:49.119 And in a family, 670 00:36:49.599 --> 00:36:50.419 it's your own money. 671 00:36:51.760 --> 00:36:52.481 For the CEO, 672 00:36:52.681 --> 00:36:54.442 it's not his or her own money. 673 00:36:55.682 --> 00:36:56.663 It's a company's money. 674 00:36:59.404 --> 00:37:01.966 And family businesses plan long, 675 00:37:02.046 --> 00:37:03.067 the next generation, 676 00:37:03.087 --> 00:37:05.968 and not to the next three months report. 677 00:37:08.329 --> 00:37:09.850 So intuition, 678 00:37:10.251 --> 00:37:13.312 not taking intuition seriously, 679 00:37:13.953 --> 00:37:16.394 costs enormous amounts of time. 680 00:37:17.082 --> 00:37:18.683 and money to companies. 681 00:37:19.964 --> 00:37:20.885 Two very good reasons. 682 00:37:20.925 --> 00:37:22.426 Maybe I'm going to hand over to 683 00:37:23.027 --> 00:37:24.949 Eric, just conscious of time. 684 00:37:27.451 --> 00:37:27.711 Yes, 685 00:37:27.751 --> 00:37:28.311 professor. 686 00:37:30.013 --> 00:37:34.797 How would it be possible to change this situation? 687 00:37:35.277 --> 00:37:45.385 Meaning to make intuition at the center of our decision-making process and the intuition as you define it. 688 00:37:45.545 --> 00:37:46.126 So, 689 00:37:46.962 --> 00:37:56.246 There are a number of very concrete procedures I have in my work with the simple rational done. 690 00:37:57.126 --> 00:37:57.907 So one is, 691 00:37:58.127 --> 00:37:59.207 give you just two examples. 692 00:37:59.347 --> 00:38:05.730 Almost all big companies have problems in defensive decision making, 693 00:38:06.731 --> 00:38:08.211 in wasting too much time. 694 00:38:08.611 --> 00:38:09.492 There are exceptions. 695 00:38:11.093 --> 00:38:11.713 For instance, 696 00:38:12.553 --> 00:38:12.673 the 697 00:38:13.173 --> 00:38:14.234 International Air 698 00:38:16.018 --> 00:38:16.138 So, 699 00:38:17.499 --> 00:38:18.699 companies like Lufthansa, 700 00:38:18.759 --> 00:38:19.659 with whom I have worked, 701 00:38:20.660 --> 00:38:22.900 they have an excellent cockpit culture. 702 00:38:24.281 --> 00:38:25.101 Cockpit culture, 703 00:38:25.121 --> 00:38:30.522 that doesn't mean that in the rest of these companies everything is okay. 704 00:38:31.663 --> 00:38:33.523 And that's why it's so safe to fly. 705 00:38:34.783 --> 00:38:35.504 On the other hand, 706 00:38:35.864 --> 00:38:38.805 many hospitals don't have a good air culture. 707 00:38:40.465 --> 00:38:43.226 And that's why so many people die in hospitals. 708 00:38:44.126 --> 00:38:45.127 That's not the only reason. 709 00:38:46.368 --> 00:38:50.811 So this just illustrates that we live in a society where we have, 710 00:38:51.251 --> 00:38:56.054 here's a pocket of excellent decision culture and error culture, 711 00:38:56.115 --> 00:38:56.875 and there is not. 712 00:38:58.016 --> 00:39:00.538 So what to do if you have a company where it's not? 713 00:39:01.218 --> 00:39:01.959 That's the aggression. 714 00:39:03.420 --> 00:39:06.722 I'll give you just two examples. 715 00:39:07.122 --> 00:39:09.904 One is a juristic that's called, 716 00:39:10.245 --> 00:39:10.365 yeah. 717 00:39:12.566 --> 00:39:19.221 Find a good model and that the best model would be the CEO or people on the very, 718 00:39:19.301 --> 00:39:19.802 very top. 719 00:39:20.283 --> 00:39:20.784 Example. 720 00:39:22.619 --> 00:39:27.863 But with a large health provider that had the same problem, 721 00:39:28.104 --> 00:39:29.485 that decisions are too slow, 722 00:39:30.045 --> 00:39:32.327 that they are too defensive, 723 00:39:33.168 --> 00:39:36.131 and nobody wants to take responsibility. 724 00:39:37.512 --> 00:39:38.913 So there was a new CEO. 725 00:39:40.334 --> 00:39:40.935 It was a she. 726 00:39:41.615 --> 00:39:42.756 And in my experience, 727 00:39:42.836 --> 00:39:47.260 often women have less fear to make decisions. 728 00:39:48.581 --> 00:39:49.702 And she... 729 00:39:51.779 --> 00:39:56.582 She assembled all the top management and said, 730 00:39:56.822 --> 00:39:57.002 look, 731 00:39:57.943 --> 00:40:00.204 last year we made this decision. 732 00:40:00.444 --> 00:40:01.665 We all were for it. 733 00:40:01.945 --> 00:40:02.905 And it was wrong, 734 00:40:03.066 --> 00:40:03.786 as we know now. 735 00:40:04.847 --> 00:40:05.007 Now, 736 00:40:05.327 --> 00:40:06.768 let's discuss what 737 00:40:07.208 --> 00:40:10.030 I, who voted for the decision, 738 00:40:10.830 --> 00:40:10.970 what 739 00:40:11.490 --> 00:40:12.651 I did wrong. 740 00:40:14.152 --> 00:40:18.835 And that sets a totally different sign for a different culture. 741 00:40:19.471 --> 00:40:19.591 So, 742 00:40:19.651 --> 00:40:21.773 from discussing our culture. 743 00:40:21.773 --> 00:40:22.133 And then, 744 00:40:22.153 --> 00:40:22.553 of course, 745 00:40:22.933 --> 00:40:23.854 the manager says, 746 00:40:24.074 --> 00:40:24.194 oh, 747 00:40:24.234 --> 00:40:25.355 we can talk about that. 748 00:40:25.775 --> 00:40:26.756 We don't have to hide that. 749 00:40:28.157 --> 00:40:28.938 It's one example. 750 00:40:29.838 --> 00:40:40.486 A second example is set some signs that signal that the culture has changed. 751 00:40:41.046 --> 00:40:42.687 You can just talk about it. 752 00:40:42.687 --> 00:40:43.548 It doesn't help much. 753 00:40:43.748 --> 00:40:44.108 Science. 754 00:40:44.588 --> 00:40:45.129 Here's one. 755 00:40:46.690 --> 00:40:47.971 Do you know the game? 756 00:40:48.567 --> 00:41:03.898 monopoly so there is a there is you can get into jail there and there is a card called get out of jail card so we instructed the ceo to 757 00:41:04.438 --> 00:41:17.367 give a get out of jail card to every manager with the instruction if you take risks for the company and it goes wrong hand in your card 758 00:41:17.847 --> 00:41:18.548 No questions. 759 00:41:19.028 --> 00:41:19.689 And also, 760 00:41:20.349 --> 00:41:24.392 put the card on your desk so that you see it every day. 761 00:41:25.453 --> 00:41:27.715 And that changes the entire situation. 762 00:41:28.055 --> 00:41:28.616 For instance, 763 00:41:29.256 --> 00:41:33.160 if there's someone who has the card lying on the desk for years, 764 00:41:34.100 --> 00:41:36.722 you ask another question, 765 00:41:36.742 --> 00:41:37.703 a different one. 766 00:41:38.944 --> 00:41:46.931 So these are science models that can help to change the culture in one that helps the organization. 767 00:41:47.583 --> 00:41:51.365 and where the decision makers don't have to hide themselves. 768 00:41:53.306 --> 00:41:53.827 There is, 769 00:41:54.407 --> 00:41:54.827 I think, 770 00:41:54.827 --> 00:41:58.989 a fundamental topic in your book, 771 00:41:59.470 --> 00:42:06.374 which is about helping leaders to create a smart decision culture, 772 00:42:07.334 --> 00:42:09.716 what you call a positive Volcker culture, 773 00:42:09.896 --> 00:42:11.016 heuristic culture, 774 00:42:11.096 --> 00:42:12.417 and error culture. 775 00:42:12.857 --> 00:42:15.759 Could you summarize what... 776 00:42:16.019 --> 00:42:23.122 How are your recommendations to create this smart decision-culturing organization? 777 00:42:23.182 --> 00:42:32.826 I'll give you an example from my own experience as a director at a Max Planck Institute with a research group of maybe 35 researchers, 778 00:42:33.787 --> 00:42:35.308 a dozen technical people, 779 00:42:35.448 --> 00:42:40.390 and any number of research assistants and secretaries and so on. 780 00:42:41.250 --> 00:42:42.431 How to deal with such a group? 781 00:42:43.483 --> 00:42:44.523 The biggest problem is, 782 00:42:46.904 --> 00:42:48.325 it's an interdisciplinary group, 783 00:42:48.365 --> 00:42:49.325 how to get them together. 784 00:42:50.285 --> 00:42:50.405 So, 785 00:42:51.586 --> 00:42:54.947 one heuristic for setting up a research group is, 786 00:42:57.147 --> 00:42:58.028 have them begin, 787 00:42:58.588 --> 00:42:59.928 all of them at the same time, 788 00:43:00.948 --> 00:43:01.529 within a week. 789 00:43:03.069 --> 00:43:03.789 That's important, 790 00:43:04.309 --> 00:43:05.450 because in my experience, 791 00:43:05.850 --> 00:43:08.411 those who are hired a few months earlier, 792 00:43:09.399 --> 00:43:14.721 I tend to think about those who come later always as younger siblings. 793 00:43:17.221 --> 00:43:29.405 And having everyone together and putting the administration into chaos helps everyone to join and to face this new situation. 794 00:43:30.285 --> 00:43:31.905 Second heuristic is, 795 00:43:32.526 --> 00:43:35.726 so every day at four o'clock, 796 00:43:36.167 --> 00:43:36.827 coffee and tea. 797 00:43:38.743 --> 00:43:39.824 And no obligations. 798 00:43:42.366 --> 00:43:43.727 And it's not a waste of time. 799 00:43:45.729 --> 00:43:46.490 This half an hour, 800 00:43:46.630 --> 00:43:47.851 or even if it's more, 801 00:43:48.832 --> 00:43:49.132 people, 802 00:43:49.853 --> 00:43:50.733 there will be trust. 803 00:43:50.773 --> 00:43:53.456 If you talk about something that has nothing to do with research, 804 00:43:54.056 --> 00:43:54.437 trust. 805 00:43:56.574 --> 00:43:57.595 If they talk of research, 806 00:43:57.715 --> 00:44:01.397 this is one of the most important places where new ideas are generated. 807 00:44:03.498 --> 00:44:03.638 And 808 00:44:04.719 --> 00:44:05.199 I always, 809 00:44:05.259 --> 00:44:05.659 when I was, 810 00:44:06.260 --> 00:44:07.340 I always went there. 811 00:44:07.961 --> 00:44:09.461 I never asked people to go there. 812 00:44:09.782 --> 00:44:11.943 It's just signal that. 813 00:44:12.283 --> 00:44:12.863 It's no point. 814 00:44:13.544 --> 00:44:18.046 I know some would try to imitate the model and forced everyone to go there. 815 00:44:18.146 --> 00:44:18.266 No, 816 00:44:18.507 --> 00:44:20.227 that's not important. 817 00:44:20.227 --> 00:44:24.390 Or what's also very important for any manager. 818 00:44:25.002 --> 00:44:26.883 as well for a research director, 819 00:44:27.424 --> 00:44:31.166 is to have at least one contrarian in your group. 820 00:44:31.907 --> 00:44:35.429 A person who speaks up against the director, 821 00:44:35.890 --> 00:44:37.971 against the group consensus, 822 00:44:38.392 --> 00:44:39.913 but on a factual basis, 823 00:44:40.513 --> 00:44:41.254 with respect. 824 00:44:43.675 --> 00:44:48.218 Many politicians fail to see that. 825 00:44:48.699 --> 00:44:51.901 We know about famous politicians like Putin, 826 00:44:52.902 --> 00:44:53.022 who... 827 00:44:53.794 --> 00:44:55.035 As far as I know, 828 00:44:56.677 --> 00:44:57.938 want to have people who clap. 829 00:44:59.499 --> 00:45:00.420 You don't want that. 830 00:45:01.681 --> 00:45:08.627 You want to have people who are willing to stand up for the facts and inform you and criticize you. 831 00:45:09.548 --> 00:45:10.769 I've written an entire paper, 832 00:45:10.809 --> 00:45:12.550 and it's in the book, 833 00:45:12.590 --> 00:45:13.752 The Intelligence of Intuition, 834 00:45:13.752 --> 00:45:15.213 you find more of these rules. 835 00:45:15.693 --> 00:45:16.374 And important, 836 00:45:16.954 --> 00:45:17.595 the rules, 837 00:45:18.115 --> 00:45:19.517 you need always to adapt them. 838 00:45:19.897 --> 00:45:20.317 For instance, 839 00:45:20.317 --> 00:45:21.078 the first rules. 840 00:45:21.490 --> 00:45:26.595 Have everyone start at the same time is not a good rule six years later. 841 00:45:27.796 --> 00:45:33.941 You would do very badly if you would replace your old group at the same time, 842 00:45:34.382 --> 00:45:37.505 because the culture that has evolved will die out. 843 00:45:39.847 --> 00:45:44.771 And it's always going back and forth between facts, 844 00:45:45.192 --> 00:45:46.213 between experience, 845 00:45:46.653 --> 00:45:47.534 between intuition. 846 00:45:49.518 --> 00:45:50.919 That's what leadership does. 847 00:45:50.979 --> 00:45:52.160 There is no one recipe. 848 00:45:53.361 --> 00:45:53.742 Thank you, 849 00:45:54.582 --> 00:45:55.023 Professor. 850 00:45:55.443 --> 00:45:55.603 Well, 851 00:45:56.024 --> 00:45:58.165 maybe a bit of opening, 852 00:45:58.165 --> 00:46:01.929 because there's a hot topic currently with artificial intelligence. 853 00:46:02.629 --> 00:46:07.914 And I've read that you make a link between artificial intelligence and psychological intelligence. 854 00:46:09.735 --> 00:46:11.657 And one can actually help each other. 855 00:46:11.797 --> 00:46:12.718 They could help each other. 856 00:46:13.158 --> 00:46:17.061 So can you just develop a little bit about what can... 857 00:46:18.666 --> 00:46:23.790 Maybe from the theory of Herbert Simon or the development of artificial intelligence, 858 00:46:24.070 --> 00:46:27.172 how can it be inspired by human psychology? 859 00:46:27.793 --> 00:46:27.953 Yes. 860 00:46:28.654 --> 00:46:28.774 So 861 00:46:29.574 --> 00:46:35.859 I started out from the distinction that Jimmy Savage made between small worlds and large worlds. 862 00:46:37.019 --> 00:46:37.700 In a small world, 863 00:46:37.700 --> 00:46:38.561 you can optimize. 864 00:46:39.081 --> 00:46:40.762 Complex methods will work. 865 00:46:41.663 --> 00:46:46.046 And that's why so many decision theories study. 866 00:46:46.482 --> 00:46:47.783 Choices being in game balls, 867 00:46:47.863 --> 00:46:48.704 it's a small world. 868 00:46:49.464 --> 00:46:51.986 But whether that translates into the real world is not clear. 869 00:46:52.947 --> 00:46:53.668 In a large world, 870 00:46:53.708 --> 00:46:54.649 you can't optimize. 871 00:46:55.730 --> 00:46:58.292 And that's what's called the management rule cover-up. 872 00:47:00.053 --> 00:47:00.854 In this situation, 873 00:47:00.874 --> 00:47:01.834 you need heuristics. 874 00:47:02.435 --> 00:47:06.718 The human mind evolved not to deal with game balls, 875 00:47:07.239 --> 00:47:08.260 well-defined game balls, 876 00:47:08.300 --> 00:47:09.260 but with uncertainty. 877 00:47:10.762 --> 00:47:13.744 And that's why we have this adaptive tool ball. 878 00:47:14.364 --> 00:47:14.485 So, 879 00:47:15.225 --> 00:47:16.066 having said that... 880 00:47:17.502 --> 00:47:18.263 The question is, 881 00:47:18.943 --> 00:47:23.946 where can complex algorithms be likely successful and where not? 882 00:47:25.407 --> 00:47:29.949 And that's what I call the difference between a stable world and an instable world. 883 00:47:30.590 --> 00:47:35.692 And it corresponds to the small world of Savage and others and the large world. 884 00:47:37.033 --> 00:47:43.777 The big successes of AI are in well-defined worlds and stable world. 885 00:47:44.077 --> 00:47:45.898 A well-defined world is one like chess. 886 00:47:46.138 --> 00:47:50.781 and go and a stable world is where tomorrow is likely like yes. 887 00:47:53.303 --> 00:47:56.405 Instable worlds include predicting the flu, 888 00:47:57.946 --> 00:47:58.567 predicting 889 00:47:59.167 --> 00:48:02.910 COVID-19, predicting human behavior, 890 00:48:03.910 --> 00:48:05.572 predicting recidivism. 891 00:48:06.652 --> 00:48:07.713 In these areas, 892 00:48:09.114 --> 00:48:14.238 the success of complex algorithms is the key. 893 00:48:15.106 --> 00:48:15.967 is not there. 894 00:48:17.968 --> 00:48:20.611 It is in well-defined problems. 895 00:48:21.952 --> 00:48:25.494 Or like in large language models where there is a language there, 896 00:48:25.515 --> 00:48:28.177 there is a matrix between correlations, 897 00:48:28.217 --> 00:48:29.358 and that's fairly stable. 898 00:48:30.739 --> 00:48:32.220 But just to give an example, 899 00:48:33.781 --> 00:48:40.747 the research in Princeton universities asked the research community, 900 00:48:41.087 --> 00:48:42.688 the machine learning community, 901 00:48:43.329 --> 00:48:44.630 to predict the future 902 00:48:45.062 --> 00:48:46.884 of so-called fragile families. 903 00:48:48.887 --> 00:48:51.389 And they delivered, 904 00:48:52.331 --> 00:48:52.451 yeah, 905 00:48:52.711 --> 00:48:54.173 what can be called big data, 906 00:48:54.573 --> 00:48:57.016 millions of data points for these families. 907 00:48:57.016 --> 00:49:00.360 So fragile families are usually families with only one 908 00:49:01.188 --> 00:49:01.608 parent. 909 00:49:02.469 --> 00:49:05.370 And this was real prediction, 910 00:49:05.450 --> 00:49:06.690 not out of sample, 911 00:49:07.191 --> 00:49:09.071 but real prediction of the future, 912 00:49:09.792 --> 00:49:10.092 where, 913 00:49:11.072 --> 00:49:11.733 for instance, 914 00:49:11.913 --> 00:49:22.797 the child's grade point average at age 15 was predicted or whether the mother still has a job or a house, 915 00:49:23.177 --> 00:49:23.657 a home, 916 00:49:24.798 --> 00:49:25.738 in a certain time. 917 00:49:27.379 --> 00:49:28.660 They had, 918 00:49:29.360 --> 00:49:29.660 I think, 919 00:49:29.944 --> 00:49:30.884 If I remember correctly, 920 00:49:30.904 --> 00:49:38.226 1,600 submissions and from with mostly highly complicated algorithms. 921 00:49:39.027 --> 00:49:48.129 And the result was that most of these machine learning algorithms could not beat a very simple algorithm. 922 00:49:48.509 --> 00:49:52.530 They just looked at three or four data points or maybe just two, 923 00:49:53.070 --> 00:49:56.791 such as how did the child perform six years ago. 924 00:50:00.585 --> 00:50:09.390 And these are examples where machine learning or complex algorithms in general don't do very well. 925 00:50:09.670 --> 00:50:10.690 And we need to see this. 926 00:50:11.090 --> 00:50:12.011 In our studies, 927 00:50:12.031 --> 00:50:12.491 for instance, 928 00:50:12.531 --> 00:50:14.752 closer to management, 929 00:50:15.333 --> 00:50:16.874 there is a common problem. 930 00:50:17.334 --> 00:50:19.355 You have a huge database. 931 00:50:19.435 --> 00:50:27.099 So a company has a huge database and wants to know which of the customers will likely purchase, 932 00:50:27.139 --> 00:50:28.520 make purchases again. 933 00:50:30.229 --> 00:50:32.650 And how do you predict that? 934 00:50:33.111 --> 00:50:34.792 This is all the high uncertainty. 935 00:50:34.792 --> 00:50:37.314 There are so many factors that predict this. 936 00:50:37.874 --> 00:50:39.455 And then there are two answers again. 937 00:50:40.176 --> 00:50:43.538 I use complex models or 938 00:50:44.258 --> 00:50:45.519 I'll use, 939 00:50:46.100 --> 00:50:46.900 I investigate. 940 00:50:46.900 --> 00:50:49.002 Now we're getting to psychological AI. 941 00:50:49.582 --> 00:50:50.763 That's the term I use. 942 00:50:51.704 --> 00:50:57.948 We don't start with logistic regressions or random forests. 943 00:50:58.708 --> 00:51:01.589 or other machine learning and try which one does better. 944 00:51:01.970 --> 00:51:05.291 But we looked how experienced managers do that. 945 00:51:06.932 --> 00:51:16.736 And the answer is most of the managers use a simple heuristic that's called the hiatus heuristic, 946 00:51:17.356 --> 00:51:21.578 which is if someone hasn't bought anything in the last nine months, 947 00:51:22.979 --> 00:51:23.179 out, 948 00:51:23.839 --> 00:51:24.460 otherwise in. 949 00:51:25.528 --> 00:51:26.949 That's again one reason, 950 00:51:27.029 --> 00:51:27.870 like Elon Musk, 951 00:51:28.351 --> 00:51:28.871 one reason. 952 00:51:29.952 --> 00:51:34.836 And one reason heuristics can be powerful under high uncertainty, 953 00:51:35.597 --> 00:51:38.560 not in a world where you know everything. 954 00:51:39.460 --> 00:51:39.580 So, 955 00:51:40.261 --> 00:51:52.492 and we have shown with 24 companies that this hiatus heuristic predicted future outcomes better than 956 00:51:53.112 --> 00:51:55.614 the typical machine learning algorithms. 957 00:51:55.755 --> 00:51:58.197 So that is random forests, 958 00:51:59.157 --> 00:52:02.360 regularized logistic regression in that case. 959 00:52:03.841 --> 00:52:07.424 And so the inside of that is the following. 960 00:52:08.605 --> 00:52:18.493 Complex algorithms can work very well and better than humans in certain areas, 961 00:52:18.753 --> 00:52:20.034 which are called stable worlds, 962 00:52:20.555 --> 00:52:21.876 but not necessarily. 963 00:52:22.312 --> 00:52:22.732 everywhere. 964 00:52:24.054 --> 00:52:37.705 And we can inform AI by studying how experts make decisions and then putting that heuristic as an algorithm. 965 00:52:37.785 --> 00:52:40.467 And the example I just gave is a very simple algorithm. 966 00:52:40.467 --> 00:52:41.468 You just have one. 967 00:52:42.028 --> 00:52:44.651 You still have to estimate the time, 968 00:52:44.731 --> 00:52:46.692 like is it nine months or six months, 969 00:52:46.712 --> 00:52:47.733 depending on the problem. 970 00:52:49.234 --> 00:52:50.876 But it gives you a different thinking. 971 00:52:51.376 --> 00:52:51.796 For instance, 972 00:52:51.796 --> 00:52:55.039 we also have shown that with the same heuristic, 973 00:52:55.119 --> 00:52:58.001 so take just the most recent, 974 00:52:59.383 --> 00:53:00.463 we could predict 975 00:53:01.624 --> 00:53:03.426 Google, so the spread of the flu, 976 00:53:04.126 --> 00:53:05.688 better than Google Flu Trends. 977 00:53:06.688 --> 00:53:07.049 Remember, 978 00:53:07.129 --> 00:53:09.511 Google Flu Trends reigned for eight years. 979 00:53:10.431 --> 00:53:15.736 I was closed down in 2015 and had the idea that with big data, 980 00:53:17.186 --> 00:53:21.367 You can predict the flu-related doctor visits, 981 00:53:21.767 --> 00:53:23.388 which is a useful problem. 982 00:53:23.428 --> 00:53:26.049 So it's what's called now-casting. 983 00:53:26.749 --> 00:53:27.449 What's now? 984 00:53:27.549 --> 00:53:29.489 What's now happening that we know that? 985 00:53:30.710 --> 00:53:35.591 And we have seen that a simple heuristic, 986 00:53:35.691 --> 00:53:36.051 that is, 987 00:53:36.711 --> 00:53:41.013 that we know that humans use in situations of high uncertainty, 988 00:53:41.433 --> 00:53:45.954 namely just take the most recent data point seriously. 989 00:53:46.322 --> 00:53:47.243 and ignore the rest. 990 00:53:47.263 --> 00:53:59.292 That could predict the spread of the flu better and every year better than Google flu trends and for all the updates of Google flu trends. 991 00:54:00.072 --> 00:54:03.995 The Google engineers updated it when something unexpected happened, 992 00:54:04.515 --> 00:54:05.116 the swine flu, 993 00:54:06.257 --> 00:54:07.137 and it couldn't. 994 00:54:08.218 --> 00:54:09.279 With big data, 995 00:54:09.579 --> 00:54:11.200 you are like a big tanker. 996 00:54:11.641 --> 00:54:13.122 You can't steer around. 997 00:54:13.898 --> 00:54:16.800 But the reason is that he can adapt to anything that's new. 998 00:54:18.641 --> 00:54:18.761 So, 999 00:54:19.601 --> 00:54:22.823 psychological AI is the idea to, 1000 00:54:24.984 --> 00:54:27.325 if the problem is about uncertainty, 1001 00:54:28.446 --> 00:54:31.828 psychological AI doesn't help you for playing chess or golf. 1002 00:54:32.828 --> 00:54:39.772 That was Robert Simon's one error that he thought that it would help in these areas. 1003 00:54:40.513 --> 00:54:42.534 But it helps you under uncertainty. 1004 00:54:43.262 --> 00:54:48.504 So find out how experienced people solve problems and its heuristics and answer, 1005 00:54:48.704 --> 00:54:51.746 because it has to be sufficiently simple to be robust. 1006 00:54:52.626 --> 00:54:59.229 And then model this as an algorithm and compare your complex algorithms with the simple ones. 1007 00:54:59.389 --> 00:55:02.730 And often you will find you actually do better simple. 1008 00:55:03.370 --> 00:55:06.172 And that has political consequences because, 1009 00:55:07.252 --> 00:55:07.872 as you may know, 1010 00:55:08.353 --> 00:55:08.473 the 1011 00:55:08.813 --> 00:55:11.574 European Union has a Digital Data Act. 1012 00:55:12.166 --> 00:55:14.168 And one of the issues is transparency, 1013 00:55:14.568 --> 00:55:16.909 meaning understandability for the public, 1014 00:55:17.490 --> 00:55:18.511 for credit scoring. 1015 00:55:19.151 --> 00:55:21.793 You want to know why you have a bad score. 1016 00:55:22.834 --> 00:55:24.375 You might want to improve it, 1017 00:55:24.855 --> 00:55:25.636 but it's black box. 1018 00:55:26.737 --> 00:55:30.920 And I am working with the largest credit scores in Germany. 1019 00:55:31.880 --> 00:55:36.924 And we found that they are using highly complex algorithms. 1020 00:55:38.445 --> 00:55:39.686 And which is. 1021 00:55:41.271 --> 00:55:44.193 You can't even say that if you have more than two credit cards, 1022 00:55:44.734 --> 00:55:45.434 that hurts you. 1023 00:55:46.015 --> 00:55:47.556 It's only true in maybe 80, 1024 00:55:47.636 --> 00:55:48.817 90% of the cases, 1025 00:55:48.857 --> 00:55:50.759 because everyone is linked with everyone. 1026 00:55:51.500 --> 00:55:54.302 So people cannot know what to do. 1027 00:55:55.463 --> 00:56:00.367 And we found that if you have an algorithm, 1028 00:56:00.527 --> 00:56:08.174 if you delete all these interactions and reduce it to half a dozen or a few more variables, 1029 00:56:09.182 --> 00:56:10.242 You do as well and better. 1030 00:56:10.743 --> 00:56:11.623 And it's transparent. 1031 00:56:12.043 --> 00:56:13.044 People can understand. 1032 00:56:13.244 --> 00:56:14.444 They can change their behavior. 1033 00:56:15.325 --> 00:56:20.207 And you don't have to nudge them into something that they don't understand. 1034 00:56:20.927 --> 00:56:21.047 So 1035 00:56:21.767 --> 00:56:33.492 I think that simplicity has also societal value and allows people to understand much better when they're being scolded. 1036 00:56:33.972 --> 00:56:35.853 And that was a very insightful conclusion. 1037 00:56:36.033 --> 00:56:37.554 Thank you very much for... 1038 00:56:38.174 --> 00:56:39.074 your participation, 1039 00:56:39.154 --> 00:56:40.215 Professor Gigerenzer. 1040 00:56:41.275 --> 00:56:44.316 Is there maybe anything you'd like to leave our listeners with, 1041 00:56:44.536 --> 00:56:47.216 like perhaps where they can find more about your work? 1042 00:56:48.297 --> 00:56:48.697 Well, 1043 00:56:48.697 --> 00:56:48.877 I mean, 1044 00:56:49.017 --> 00:56:53.998 you can start with the book that is about smart management, 1045 00:56:56.079 --> 00:56:59.220 and or you might read the book, 1046 00:56:59.260 --> 00:57:00.060 as you mentioned, 1047 00:57:00.060 --> 00:57:01.560 The Intelligence of Intuition, 1048 00:57:02.861 --> 00:57:05.722 or you can read some of my more scientific books. 1049 00:57:07.918 --> 00:57:10.060 that rationality for mortals, 1050 00:57:13.702 --> 00:57:14.383 but read. 1051 00:57:15.163 --> 00:57:15.663 And also, 1052 00:57:16.484 --> 00:57:25.590 I think the basic insights are that take uncertainty seriously and notice that we still, 1053 00:57:26.671 --> 00:57:29.933 the mainstream of behavioral economics doesn't do that. 1054 00:57:31.034 --> 00:57:31.434 Otherwise, 1055 00:57:31.434 --> 00:57:33.135 we wouldn't declare everything an error. 1056 00:57:34.356 --> 00:57:35.257 That is simple. 1057 00:57:38.142 --> 00:57:38.703 And second, 1058 00:57:39.023 --> 00:57:39.804 align to that, 1059 00:57:40.144 --> 00:57:42.405 take heuristics seriously. 1060 00:57:43.566 --> 00:57:47.089 They are not second best in a VUCA world. 1061 00:57:47.829 --> 00:57:49.771 They are the only thing we can do. 1062 00:57:50.751 --> 00:57:52.252 And the question is a different one. 1063 00:57:53.033 --> 00:57:54.474 Which heuristic makes sense? 1064 00:57:54.934 --> 00:58:00.898 Which is smart and which is not smart in this application? 1065 00:58:01.799 --> 00:58:02.379 And finally, 1066 00:58:02.820 --> 00:58:05.902 have the courage to make decisions. 1067 00:58:06.322 --> 00:58:09.653 and to stand up for them and take responsibility. 1068 00:58:10.756 --> 00:58:11.318 Be Good, 1069 00:58:11.599 --> 00:58:13.826 a podcast by the BVA Nudge Unit.