1 00:00:00,790 --> 00:00:04,600 So welcome, everyone, to today's strategy lecture. 2 00:00:04,600 --> 00:00:12,040 In case you don't know, the Strachey she lectures are a distinguished lecture series named for Christopher Strachey, 3 00:00:12,040 --> 00:00:20,140 a pioneer of computer science, known in particular for his contributions to programming languages and the notational semantics. 4 00:00:20,140 --> 00:00:26,440 And the founder of The Discipline at Oxford, the first professor of computing science in Oxford Street. 5 00:00:26,440 --> 00:00:34,240 He lectures a generously supported by Oxford Asset Management. The computer science department is very grateful for their support. 6 00:00:34,240 --> 00:00:36,700 Under normal circumstances, pre pandemic. 7 00:00:36,700 --> 00:00:44,590 This enabled us to hold lectures in the splendid venue and with very nice refreshments enjoyed by all the participants. 8 00:00:44,590 --> 00:00:50,470 We don't have those that is today. But what we do have is is an excellent lecturer. 9 00:00:50,470 --> 00:00:56,860 So it's my great pleasure to introduce today's straight lecturer, Professor Sonia Smedes. 10 00:00:56,860 --> 00:01:06,640 Sonia is Professor of Logic and Epistemology at the Institute for Logic, Language and Computation LLC at the University of Amsterdam. 11 00:01:06,640 --> 00:01:16,480 And since 2016, she has been the director of L.L.C., which is one of the major world centres for the study of logic and computation. 12 00:01:16,480 --> 00:01:20,800 Amongst her many activities and recognitions, too many to mention them all. 13 00:01:20,800 --> 00:01:26,500 But let me just mention that she was awarded the book of On the Human Prise for her work in Quantum Logic. 14 00:01:26,500 --> 00:01:37,270 She has held a Rosalind Franklin fellowship and she was elected to membership of the Academia Europea, the European Academy in 2019. 15 00:01:37,270 --> 00:01:48,700 Sonia's work has spanned a range of topics, notably dynamic quantum logic, dynamic epistemic logic, formal epistemology and social epistemology. 16 00:01:48,700 --> 00:01:54,460 Much of her recent work focuses on the use of logic to characterise and study information flow, 17 00:01:54,460 --> 00:02:02,380 opinion formation, social influence, informational cascades, homophily and polarisation in social networks. 18 00:02:02,380 --> 00:02:07,100 And the title of her lecture today is, I think, very resonant for us all. 19 00:02:07,100 --> 00:02:10,720 The Quest for Truth in the Information Age. 20 00:02:10,720 --> 00:02:19,460 Just before I hand over to Sonia, let me mention that if you have a question, please use the Q&A facility to send your question. 21 00:02:19,460 --> 00:02:28,510 You can find the Q&A panel by clicking on the icon with a question mark in it on the right hand side, top of your screen. 22 00:02:28,510 --> 00:02:35,000 After the lecture, we will have a Q&A session where I will read questions to Sonia and she will respond. 23 00:02:35,000 --> 00:02:40,600 OK, without further ado, I have a great pleasure in handing over to Sonia for the lecture. 24 00:02:40,600 --> 00:02:53,640 Sonia, please. OK, well, thank you so much for this invitation and thank you some some for a very nice introduction. 25 00:02:53,640 --> 00:02:59,190 It's a really great honour to give the stretchy lecture and computing science today. 26 00:02:59,190 --> 00:03:09,300 So the title is The Quest for Truth in the Information Age. So let me start by reflecting on worldwide connectivity as we know it today. 27 00:03:09,300 --> 00:03:18,840 So there are many advantages in worldwide connectivity, from finding information, from finding high school friends, 28 00:03:18,840 --> 00:03:27,090 creating new friends, finding jobs, finding supports, etc. So plenty of advantages, but with disadvantages. 29 00:03:27,090 --> 00:03:30,690 Also on the down side, there are many disadvantages as well. 30 00:03:30,690 --> 00:03:44,100 So we see the spread of misinformation. We see online believe manipulation, online radicalisation, many cases of online bullying in schools, 31 00:03:44,100 --> 00:03:50,430 the formation of informational cascade's and echo chambers to which we come back later on. 32 00:03:50,430 --> 00:03:55,230 So that brings me to the question can truth survive the information age? 33 00:03:55,230 --> 00:04:02,610 Now, answering this question requires a study of the mechanisms for mass information 34 00:04:02,610 --> 00:04:09,630 aggregation and information exchange in relation to our ability to track the truth. 35 00:04:09,630 --> 00:04:16,110 So I will focus on the social and epistemic benefits and costs of such mechanisms, 36 00:04:16,110 --> 00:04:24,030 and I will reflect upon the successes and failures of individual versus collective rationality. 37 00:04:24,030 --> 00:04:31,890 So the approach that I adopt brings together many insights from a series of different fields. 38 00:04:31,890 --> 00:04:43,290 So one is I as a magician, I use logic to model the dynamics of information and we look specifically at knowledge and belief. 39 00:04:43,290 --> 00:04:49,560 We take input from belief, revision, theory and philosophy and another area from philosophy. 40 00:04:49,560 --> 00:04:54,540 We use formal epistemology and a particular social epistemology. 41 00:04:54,540 --> 00:04:57,390 Epistemology is a theory of knowledge. 42 00:04:57,390 --> 00:05:07,890 We use input also from judgement, aggregation and social choice theory, and our latest work uses more input from network theory. 43 00:05:07,890 --> 00:05:14,520 So let's focus on this aspect of tracking the truth and pose the first question. 44 00:05:14,520 --> 00:05:18,810 So how the group successes and failures to track the truth? 45 00:05:18,810 --> 00:05:23,730 How are they related to individual successes and failures to track the truth? 46 00:05:23,730 --> 00:05:31,110 Not answering this means that we have to look at what is known as wisdom versus madness of crowds, 47 00:05:31,110 --> 00:05:35,250 and maybe more specifically, what actually does a group know? 48 00:05:35,250 --> 00:05:41,880 So what is the knowledge that a group can possess? So I'll call this the group's potential knowledge. 49 00:05:41,880 --> 00:05:48,570 And what I mean by that is the knowledge that the members of a group may come to possess 50 00:05:48,570 --> 00:05:55,770 by combining their individual knowledge and using their joint epistemic capability. 51 00:05:55,770 --> 00:05:59,550 But in what sense can a group process beliefs or knowledge? 52 00:05:59,550 --> 00:06:03,600 Of course, you can infer this by looking at the group's behaviour. 53 00:06:03,600 --> 00:06:13,710 This is the joint actions that they can do, that the members of the group can perform after combining their individual knowledge. 54 00:06:13,710 --> 00:06:19,050 So that still brings the question, so how smart is a group? 55 00:06:19,050 --> 00:06:25,650 So does a group no more than each of the members that that is in that group. 56 00:06:25,650 --> 00:06:38,910 So on the one hand, we can say yes, yes, because the new information that was initially unknown to any of the agents may be combined, 57 00:06:38,910 --> 00:06:46,110 may be obtained by combining individual different pieces of private information possessed by the different agents. 58 00:06:46,110 --> 00:06:50,940 So in potentially we know more as a group than each of us individually. 59 00:06:50,940 --> 00:06:59,400 But the opposite may also happen when some, maybe all individual knowledge is made inaccessible to the group. 60 00:06:59,400 --> 00:07:04,470 And now you can think, well, maybe at first sight the opposite never happens. 61 00:07:04,470 --> 00:07:10,500 Maybe the group's potential knowledge includes everything known by any of the members. 62 00:07:10,500 --> 00:07:14,500 So that exactly is what is called wisdom of the crowds. 63 00:07:14,500 --> 00:07:20,130 And there's a lot of popularising literature on that also book with this title. 64 00:07:20,130 --> 00:07:24,870 So the wisdom of the crowds is this idea that the potential knowledge of the group is 65 00:07:24,870 --> 00:07:30,390 typically much higher than even the knowledge of the most expert member of the group. 66 00:07:30,390 --> 00:07:38,070 Now, if you looking into this a little bit more detail, you see that there are different examples of wisdom of the crowds. 67 00:07:38,070 --> 00:07:44,320 And I will make a distinction between two kinds of examples. And each of those is based on. 68 00:07:44,320 --> 00:07:50,740 Different forms of information aggregation, and they have different explanations. 69 00:07:50,740 --> 00:08:02,200 So if we focus on the first type, then you would say, OK, a different way of aggregating information would be by communication. 70 00:08:02,200 --> 00:08:08,410 So a group can actualise a piece of group knowledge by communication on the one hand. 71 00:08:08,410 --> 00:08:17,260 And on the other hand, it would use a different method for aggregation, which would be pulling the opinions together, some form of judgement, 72 00:08:17,260 --> 00:08:26,110 aggregation, and how good a group will be in realising its epistemic potential or bringing about the potential group. 73 00:08:26,110 --> 00:08:31,630 Knowledge will depend on each of these methods, but also on different factors. 74 00:08:31,630 --> 00:08:36,520 So the different factors that play a role are well, first of all, 75 00:08:36,520 --> 00:08:45,100 it's important what the topology of the communication network will be like, who will talk to whom, but also trust plays a role. 76 00:08:45,100 --> 00:08:48,730 So the mutual trust graph is important. 77 00:08:48,730 --> 00:08:56,840 So how trustworthy do you think that the information is that you get from another agent or a subgroup of agents? 78 00:08:56,840 --> 00:09:05,120 And self trust will play a role, so this includes the threshold of what is needed for an agent to change her beliefs. 79 00:09:05,120 --> 00:09:15,800 So from which point onwards, can you be influenced or are too influential to be subjected to information that comes from others? 80 00:09:15,800 --> 00:09:23,450 And the last point here on that slide is about the the agents epistemic interest plays a role. 81 00:09:23,450 --> 00:09:35,440 This means her own agenda, her own questions, what she is interested in and how she filters the information that she gets her learning goals. 82 00:09:35,440 --> 00:09:44,860 So if we look a little bit more closer, then you see that these two different ways of obtaining or pulling information together, 83 00:09:44,860 --> 00:09:52,930 one via communication and the other via some aggregation method, they lead to two different notions of group knowledge. 84 00:09:52,930 --> 00:10:00,970 So the first one, the first notion of group knowledge that we see would be the one that is based on communication. 85 00:10:00,970 --> 00:10:09,040 So so here you see that the agents will have a piece of information or a piece 86 00:10:09,040 --> 00:10:14,980 of evidence that typically is truthful and the agents share their knowledge. 87 00:10:14,980 --> 00:10:21,250 And examples of this type of group knowledge would be joint authorship on a paper. 88 00:10:21,250 --> 00:10:28,840 So agents that collaborate on the project, such as maybe a big science project like the Human Genome Project, 89 00:10:28,840 --> 00:10:38,650 maybe proof of a mathematical theorem for us last year, but also a deliberation that happens in the hiring committee. 90 00:10:38,650 --> 00:10:43,990 So these are all explanations were individual experts. 91 00:10:43,990 --> 00:10:48,320 They collaborate together, they share their knowledge and they communicate doing so. 92 00:10:48,320 --> 00:10:55,300 And that brings evolved form of group knowledge. And the form that it brings about is typically described as. 93 00:10:55,300 --> 00:11:02,260 So the explanation in epistemic logic and epistemology that is given for this is distributed knowledge. 94 00:11:02,260 --> 00:11:11,230 So distributed knowledge is a specific form that is based on information sharing via communication between agents were as an example. 95 00:11:11,230 --> 00:11:19,420 I know p you know, that implies. Q So we pull the information together and we use our reasoning powers. 96 00:11:19,420 --> 00:11:27,650 So we do multiple opponents in this case. This means we both know P and by modest performance, we both know Q. 97 00:11:27,650 --> 00:11:32,690 So that was the first type of knowledge. Now the second type is an entirely different type. 98 00:11:32,690 --> 00:11:38,480 So here, in contrast to the first type, communication is not allowed. 99 00:11:38,480 --> 00:11:45,800 So here we see agents that aggregate independent opinions so they aggregate their beliefs. 100 00:11:45,800 --> 00:11:52,430 And these beliefs are based on private observations of pieces of evidence that they have. 101 00:11:52,430 --> 00:12:01,280 But in this case, the evidence is soft. So it's possibly that the evidence is fallible and it is important here. 102 00:12:01,280 --> 00:12:10,610 This is a different type of group knowledge because it requires agents to pull together their information based on independent observations. 103 00:12:10,610 --> 00:12:18,020 So the evidence that they have, they make an independent observation of this and they base their opinions on that. 104 00:12:18,020 --> 00:12:25,880 So examples that you see in the literature are they include the verification of experimental results. 105 00:12:25,880 --> 00:12:33,590 And in the popular science literature, you will see examples that refer to there is an important example, 106 00:12:33,590 --> 00:12:42,680 the case described by Francis Galton in nineteen of six, he went to a livestock fair or that was an ox on display. 107 00:12:42,680 --> 00:12:50,330 And what Francis did was he collected all the so there was a contest where people could win a prise 108 00:12:50,330 --> 00:12:56,030 if they guessed the weight of the ox correctly and they each had to ride their guests on the ballot. 109 00:12:56,030 --> 00:13:04,460 So Francis collected all these ballots. And what he saw was that in the end, if you would take the mean of all the guesses, 110 00:13:04,460 --> 00:13:12,440 take the average, then that actually gets to the almost exact weight of the ox. 111 00:13:12,440 --> 00:13:22,280 So he saw that collecting that independent information from all the agents and then saying this gets you to almost a perfect answer. 112 00:13:22,280 --> 00:13:28,610 Other examples in the literature are, for instance, counting jellybeans in a jar. 113 00:13:28,610 --> 00:13:34,550 This is also an example that was done by Jack Trainer already in the 1920s. 114 00:13:34,550 --> 00:13:43,100 In his classroom, he took a jelly bean jar to his classroom, asked the students to to give an estimate of the number of beans in the jar. 115 00:13:43,100 --> 00:13:50,330 And he saw that if you take the mean of all the gases, you get almost the correct answer. 116 00:13:50,330 --> 00:13:56,960 So here indeed, this works very well. This method for aggregating the beliefs works very well. 117 00:13:56,960 --> 00:14:07,580 But it is based on this idea that no communication is allowed and the guesses of the individual participants are independent. 118 00:14:07,580 --> 00:14:13,280 Now, we can illustrate that with a more concrete example. 119 00:14:13,280 --> 00:14:20,300 So this will be what I call the first or an example. So let's take a sequence of individual agents. 120 00:14:20,300 --> 00:14:27,230 And I enumerate these agents a want to. And and the idea is that these agents entered into a room. 121 00:14:27,230 --> 00:14:31,730 So they stand in a line. So it's a sequence. They enter into that room. 122 00:14:31,730 --> 00:14:39,630 And it is common knowledge that in that room there is one of two possible urns they don't know and they can't see which one from the outside. 123 00:14:39,630 --> 00:14:53,750 So it is either urn w an urn w contains two white walls, one black wool or A or B and urn B contains two black males and one white ball. 124 00:14:53,750 --> 00:15:00,770 So that is common knowledge that one of these answers in the room and in person goes into the room there, 125 00:15:00,770 --> 00:15:06,890 a ball from the urn looks at it and makes a guess about which urn is placed in the room. 126 00:15:06,890 --> 00:15:11,680 So the guess is either W or B. And in this case, the guesses are secret. 127 00:15:11,680 --> 00:15:23,390 So no communication is allowed. So imagine the launch sequence of agents, each entering and having a guest that may be writing that guess on a ballot. 128 00:15:23,390 --> 00:15:28,610 Now, at the end, the poll is taken of all the people's guesses. 129 00:15:28,610 --> 00:15:35,300 And what we see is that the majority guests actually will then be what we call the potential group knowledge. 130 00:15:35,300 --> 00:15:42,440 Given the size of the group gets to infinity, the group gets virtual knowledge of the real estate, 131 00:15:42,440 --> 00:15:48,800 with a probability approaching to one of the size of the group is large enough then the truth. 132 00:15:48,800 --> 00:15:53,270 Tracking power of this group via this method is excellent. 133 00:15:53,270 --> 00:16:04,550 And the standard explanation for this can be given by some variation of council says here, which is essentially based on the law of large numbers. 134 00:16:04,550 --> 00:16:13,700 And what it says here is that when you perform many independent observations, the individual errors are the pieces of private evidence. 135 00:16:13,700 --> 00:16:20,550 Supporting the false hypothesis will be outnumbered by the truth supporting evidence. 136 00:16:20,550 --> 00:16:27,680 So have statement of the theorem a little bit more specifically correct state. 137 00:16:27,680 --> 00:16:35,660 So the term Sesto, the probability that the majority selects the correct alternative approaches to one when the size 138 00:16:35,660 --> 00:16:42,260 of the group and goes to infinity whenever the following three constraints of conditions are met, 139 00:16:42,260 --> 00:16:51,890 namely, one, that each individual makes an independent decision based only on private information, 140 00:16:51,890 --> 00:16:59,450 the likelihood of the private signal to be correct has to be bigger than than a half. 141 00:16:59,450 --> 00:17:09,670 And in this case, the choice has to be binary. So we have two mutually exclusive alternatives that people vote on or that they choose from. 142 00:17:09,670 --> 00:17:11,440 And that's an important term. 143 00:17:11,440 --> 00:17:22,180 It has also been used in a lot of different contexts, so the term of Canossa has been used to argue in favour of epistemic democracy. 144 00:17:22,180 --> 00:17:31,690 That is the idea that in normal circumstances, the democratic forms of belief aggregation are indeed conducive to the truth. 145 00:17:31,690 --> 00:17:40,000 So I refer you here to the work of Christian listing Goulden for this and what they did in their work. 146 00:17:40,000 --> 00:17:49,240 I've quoted here, I've put the reference for the paper is that they actually extended the term in their argument to go beyond the binary choice. 147 00:17:49,240 --> 00:17:53,540 That was the last condition in the in the statement of the term. 148 00:17:53,540 --> 00:18:03,500 And so they looked at voting scenarios that involved many correct options and they still saw that that mattered. 149 00:18:03,500 --> 00:18:07,130 So this extended version of the concept here is very good. 150 00:18:07,130 --> 00:18:16,810 And as a as a belief aggregation method that is conducive to the truth. 151 00:18:16,810 --> 00:18:22,300 Now, what if the jury tarim, if we look at it? 152 00:18:22,300 --> 00:18:30,730 Of course, it assumes that the agent's opinions are based only on independent observations, like when the people entered that room in the urn example, 153 00:18:30,730 --> 00:18:40,510 they look at the bull and they have to make their guess independently and they have their own private independent evidence for doing so. 154 00:18:40,510 --> 00:18:44,620 Now, what happens if we allow some communication of observations? 155 00:18:44,620 --> 00:18:48,100 Right in the example to now, communication was not allowed. 156 00:18:48,100 --> 00:18:59,420 If we do that, if we allow communication, then the opinions will no longer be independent, which means that the Condorcet jury theorem will not apply. 157 00:18:59,420 --> 00:19:03,740 And intuitively, we might think that after communication, 158 00:19:03,740 --> 00:19:11,920 the agents entering the room in their own example later on now they would have maybe more information than just their private observations. 159 00:19:11,920 --> 00:19:17,270 So you would intuitively expect that the group's knowledge would increase. 160 00:19:17,270 --> 00:19:24,530 However, as we'll soon see when I extend the example, this is actually not what happens. 161 00:19:24,530 --> 00:19:27,620 So that brings me to the fragility of group knowledge. 162 00:19:27,620 --> 00:19:37,670 So we've talked about two forms of wisdom of the crowds and two specific forms that we're linked to, two types of group knowledge. 163 00:19:37,670 --> 00:19:45,260 So both types are actually very fragile because any change in the conditions that are necessary 164 00:19:45,260 --> 00:19:53,000 to bring about or actualise that form of group knowledge can lead to under optimal results. 165 00:19:53,000 --> 00:19:59,870 So let's look at the first type of group knowledge, which is the one that is based on communication. 166 00:19:59,870 --> 00:20:09,410 Now, in this case, if the information exchange is limited by the network or by the degree of trust or by the agents agendas, 167 00:20:09,410 --> 00:20:17,270 then the group will only actualise something much less than full distributed knowledge. 168 00:20:17,270 --> 00:20:26,720 So this type of group knowledge can fail. And the reason for the failure is that we would say that it this some kind of selective hearing. 169 00:20:26,720 --> 00:20:37,670 So in this case, the agents listen only to certain sources, mainly only their friends. 170 00:20:37,670 --> 00:20:45,020 They listen only if it fits the information, fits their own interests or fits their own agenda. 171 00:20:45,020 --> 00:20:49,070 And maybe they only accept new information if it is consistent with their own 172 00:20:49,070 --> 00:20:56,570 information that they already had from before or their pre-existing beliefs. 173 00:20:56,570 --> 00:21:00,980 So in my work with my co-authors, 174 00:21:00,980 --> 00:21:08,720 we looked at the specific example which we called the curse of the committee because of many deliberating committees. 175 00:21:08,720 --> 00:21:14,690 After long sessions of deliberations, information sharing, a lot of argumentation, 176 00:21:14,690 --> 00:21:19,640 it often happens that the only thing at the end that the committee agrees on 177 00:21:19,640 --> 00:21:24,920 is actually what they already agreed on before the start of the deliberation. 178 00:21:24,920 --> 00:21:32,510 Of course, in these cases, the potential knowledge of the group seems to boil down to the groups prior common knowledge. 179 00:21:32,510 --> 00:21:36,290 So the information that was common knowledge at the start of the deliberation, 180 00:21:36,290 --> 00:21:41,360 but not at the end, and that is typically much poorer than distributed knowledge. 181 00:21:41,360 --> 00:21:49,250 Distributed knowledge is the ideal of what they could reach. So in a sense, this is the worst possible outcome. 182 00:21:49,250 --> 00:21:59,750 And now one reason why this happens is the selective hearing principal were an agent's actual or potential 183 00:21:59,750 --> 00:22:07,580 knowledge is limited by her own epistemic interests and her own questions and the things that she's interested in. 184 00:22:07,580 --> 00:22:15,620 So the liberating agents here are process only the information that is relevant to their own questions or issues. 185 00:22:15,620 --> 00:22:19,880 So in joint work with example Baltar and Rachel body, 186 00:22:19,880 --> 00:22:26,840 we explain this example of the curse of the committee by looking at the at each of the agents epistemic 187 00:22:26,840 --> 00:22:33,710 interests that they have to actually give a full formalisation of this which are not doing this talk, 188 00:22:33,710 --> 00:22:44,300 but I will give a reference to it. So we use a version of interrogative epistemic logic that is an epistemic logic with an interrogator component, 189 00:22:44,300 --> 00:22:49,100 which means we are also modelling the questions on the issues that the agents have, 190 00:22:49,100 --> 00:22:55,070 and we put a specific constraint on how we represent the knowledge of the agents, 191 00:22:55,070 --> 00:23:00,830 including the fact that one can only know things that are actually relevant to the issues 192 00:23:00,830 --> 00:23:07,340 that other agent has so relevant to the questions or the agenda of that specific agent. 193 00:23:07,340 --> 00:23:12,640 And the same constraint also applies on the dynamics that we that we model. 194 00:23:12,640 --> 00:23:24,320 So so the dynamics means that here we look at public announcements and that has to comply with the selective hearing principle. 195 00:23:24,320 --> 00:23:32,450 So an agent only learns from an announcement if it's relevant and only learns from an 196 00:23:32,450 --> 00:23:38,580 announcement what the relevant consequences are relevant to the agent's own agenda. 197 00:23:38,580 --> 00:23:45,370 When you do that, then it's easy to see that this indeed can lead to radical disagreement. 198 00:23:45,370 --> 00:23:56,140 Now in the paper rules to formalise what the actual group knowledge is in such an agenda driven context, 199 00:23:56,140 --> 00:24:05,530 so what we call the the groups, the group Jews epistemic potential or group potential group knowledge, 200 00:24:05,530 --> 00:24:13,930 we the this case superscripts that will amount to whatever will become common knowledge after each 201 00:24:13,930 --> 00:24:22,630 agent in the group updates her knowledge with the group's information that answers her questions. 202 00:24:22,630 --> 00:24:32,920 So this potential will typically lie in between initial common knowledge and what is known in the literature as distributed knowledge. 203 00:24:32,920 --> 00:24:38,470 So what we saw in our paper is that distributed knowledge and common knowledge, 204 00:24:38,470 --> 00:24:44,110 which are the main concepts from epistemic logic and the standard literature, 205 00:24:44,110 --> 00:24:49,150 they can actually be seen as too extreme versions of this potential group knowledge, 206 00:24:49,150 --> 00:24:55,210 giving us the upper and the lower limits of the group's epistemic potential. 207 00:24:55,210 --> 00:25:04,150 So if every agent in the group is interested in all the information processed by all everybody else, so when the agendas are very cohesive, 208 00:25:04,150 --> 00:25:12,520 then the potential norm, which is maximal and then the potential group knowledge will be equally distributed knowledge. 209 00:25:12,520 --> 00:25:21,760 But on the other side, if the agents have very dissimilar agendas, namely the questions they are interested in are orthogonal to each other. 210 00:25:21,760 --> 00:25:28,360 So no agent has opened questions that can actually be answered by the group than a potential group. 211 00:25:28,360 --> 00:25:29,770 Knowledge is very minimal. 212 00:25:29,770 --> 00:25:38,490 And in that case, as in the curse of the committee example, they will fall back on what was already common knowledge of START. 213 00:25:38,490 --> 00:25:49,140 So the lesson that we learn from this is that a group formed of agents with very dissimilar agendas will have a very poor epistemic power, 214 00:25:49,140 --> 00:26:00,060 they say, actually with what they already knew. And when you have a committee or a group with a very cohesive agenda, with dissimilar expertise, 215 00:26:00,060 --> 00:26:06,600 still need specific experts coming together, they will have a very strong epistemic potential. 216 00:26:06,600 --> 00:26:08,790 They can actually learn from each other. 217 00:26:08,790 --> 00:26:18,450 And here I gave also the reference to the paper in which we explore these results, where we make this formally precise. 218 00:26:18,450 --> 00:26:19,880 So the curse of the comment, 219 00:26:19,880 --> 00:26:29,550 the example that we investigate that can actually be related to what is known as the common knowledge of fact and social psychology. 220 00:26:29,550 --> 00:26:35,790 So that common knowledge of fact refers to the case where groups tend to focus more 221 00:26:35,790 --> 00:26:41,490 on information that is already common rather than on sharing their private evidence. 222 00:26:41,490 --> 00:26:47,400 Now, the specific relation to this common knowledge of facts should still be explored because 223 00:26:47,400 --> 00:26:53,430 it's my understanding that they actually talk about another notion of knowledge, 224 00:26:53,430 --> 00:26:59,070 which is not really common, but more something like mutual knowledge of what everybody knows. 225 00:26:59,070 --> 00:27:07,440 But still, it's interesting that this effect has been has been studied and points out that information that the committee had already before the 226 00:27:07,440 --> 00:27:17,850 deliberations starts is something that the group tends to focus more on that rather than on sharing their own private information. 227 00:27:17,850 --> 00:27:24,770 So if I go back now to our information society, then what we see happening is that. 228 00:27:24,770 --> 00:27:36,350 People choose with whom to associate because information technology also enhances visibility for more choice based on our preferences or agendas, 229 00:27:36,350 --> 00:27:41,060 and having more choice is a good thing. At least that's what we think. 230 00:27:41,060 --> 00:27:49,310 But these examples also show that information exchange can be filtered through these choices. 231 00:27:49,310 --> 00:27:58,310 So more connectivity without more epistemic sophistication can actually accelerate effects of polarisation. 232 00:27:58,310 --> 00:28:09,470 The formation of Cascades echo chambers. So these examples bring me to the to the fragility of group knowledge again. 233 00:28:09,470 --> 00:28:17,690 But now let's look at the second type. So we've we have already seen the fragility of group knowledge that is based on communication and 234 00:28:17,690 --> 00:28:23,990 what can go wrong when people filter incoming information through their agendas and interests. 235 00:28:23,990 --> 00:28:28,610 But let's look at the second type of group knowledge, which is not based on communication, 236 00:28:28,610 --> 00:28:39,590 but which is based on some judgement aggregation method of voting out where people have to bring together independent opinions. 237 00:28:39,590 --> 00:28:47,930 So this type is also very fragile and also prone to failure because in this case, any breach of the agent's independence. 238 00:28:47,930 --> 00:28:52,220 So any communication can lead the group astray. 239 00:28:52,220 --> 00:29:03,980 And typical examples here are indeed the formation of an informational cascade, which you can think of as some form of rational epistemic bandwagon's. 240 00:29:03,980 --> 00:29:15,110 Forms of groupthink and notion of pluralistic ignorance is the scenario where we point to a lack of common knowledge leading to group. 241 00:29:15,110 --> 00:29:26,680 The group suppresses generally shared results. And another example would be radical disagreement, which will lead to group polarisation. 242 00:29:26,680 --> 00:29:36,850 So let's zoom in on the first example here on the formation of an informational cascade and on its social and epistemic features. 243 00:29:36,850 --> 00:29:40,150 So so imagine a sequence of people. 244 00:29:40,150 --> 00:29:54,130 So the line of of agents as an example and the informational cascade happens when the people in that line base their 245 00:29:54,130 --> 00:30:02,770 decisions by observing the decisions of the predecessor and they follow their the decisions of their predecessors, 246 00:30:02,770 --> 00:30:09,490 ignoring their own private evidence that they have. So these situations happen. 247 00:30:09,490 --> 00:30:16,360 And it is not necessarily the case that this is just mindless imitation that we're imitating our predecessors. 248 00:30:16,360 --> 00:30:20,650 Now, these people are not brainwashed. They're not part of mass hysteria. 249 00:30:20,650 --> 00:30:29,240 So actually, this happens in cases where people are assumed to be rational and they have a rational choice. 250 00:30:29,240 --> 00:30:33,810 So let's go back to our own example, which is very fit to study this, 251 00:30:33,810 --> 00:30:42,500 so take the same or an example that I have had before where the standard explanation was given by the commerce security here, 252 00:30:42,500 --> 00:30:47,150 except now I will allow a limited form of communication. 253 00:30:47,150 --> 00:30:55,830 So when each agent makes her guess for W.O. or be the guest now is immediately publicly announced. 254 00:30:55,830 --> 00:31:00,050 So they go into the room, the drawable they look at it and they make a guess. 255 00:31:00,050 --> 00:31:05,900 But instead of writing it on a ballot, they go out and they make a public announcement. 256 00:31:05,900 --> 00:31:13,730 And what we see is that an informational cascade arises with a probability of at least one of the nine. 257 00:31:13,730 --> 00:31:18,250 If the first two votes that are drawn are of the minority type. 258 00:31:18,250 --> 00:31:27,080 So suppose that the person in the room is or w and the first two agents who enter the room, they take a ball and it is a black ball. 259 00:31:27,080 --> 00:31:35,260 Then the first two agents rationally will guess B. And they make their guesses publicly. 260 00:31:35,260 --> 00:31:39,100 There is a public announcement of their guesses to everybody else, everybody else can hear it, 261 00:31:39,100 --> 00:31:50,710 then all the agents will rationally follow and they will all select the wrong option B. So the wrong option will then be unanimously selected. 262 00:31:50,710 --> 00:31:58,700 So now we can ask the question, is this rational? And according to an analysis, actually, the answer is yes, 263 00:31:58,700 --> 00:32:06,020 because given the available information that the agents have been agents that follow probability 264 00:32:06,020 --> 00:32:12,200 theory and based based update's rule when they are interested in individual truth tracking, 265 00:32:12,200 --> 00:32:22,160 they will behave exactly in this way. So this means that individual Bayesian rationality can lead the whole group to the wrong answer. 266 00:32:22,160 --> 00:32:36,980 So to group irrationality, some people throw doubt on the Bayesian proof because it doesn't make explicit the higher order reasoning of the agents. 267 00:32:36,980 --> 00:32:43,520 So you can do the whole analysis of what happens in the unexampled right down a right of the proof. 268 00:32:43,520 --> 00:32:48,530 And what you will see is that it has credence is there for a simple belief, 269 00:32:48,530 --> 00:32:53,240 but it doesn't take into account the high order reasoning in higher order beliefs of the agents. 270 00:32:53,240 --> 00:33:01,490 So the question we ask ourselves is that if agents who who have a theory of the 271 00:33:01,490 --> 00:33:06,740 mind and who can reflect on the overall protocol that they can reflect on, 272 00:33:06,740 --> 00:33:16,010 on the whole scenario that is happening and they can realise that they are part of a cascade and they reason about other agents minds. 273 00:33:16,010 --> 00:33:20,180 Can they avoid the cascade so they can piece everything together? 274 00:33:20,180 --> 00:33:24,080 They have a full theory of the mind. Can they avoid the cascade? 275 00:33:24,080 --> 00:33:31,610 And that can happen for a number of cases. The answer may be yes, but it is not for an example. 276 00:33:31,610 --> 00:33:35,510 So whatever reasoning powers that we give our agents, 277 00:33:35,510 --> 00:33:43,850 it's not going to help them to make a rational decision to break this cascade of what we did in our paper. 278 00:33:43,850 --> 00:33:50,750 So I've given the reference to the paper here. Below the slides is that we we have a proof for that. 279 00:33:50,750 --> 00:33:56,060 So we use dynamic epistemic logic to prove three, prove the argument. 280 00:33:56,060 --> 00:34:06,470 But now in an epistemic logic setting where you can make the reasoning powers of the agents explicit so the agents have unlimited reflective power, 281 00:34:06,470 --> 00:34:13,880 these are super duper agents. And I mean the tools that we have for this, they work very well because indeed, 282 00:34:13,880 --> 00:34:19,580 epistemic logic incorporates all the levels of knowledge and belief that agents have. 283 00:34:19,580 --> 00:34:27,290 And dynamic epistemic logic allows us to add the knowledge and belief that agents have about what is going on, 284 00:34:27,290 --> 00:34:36,560 about the protocol, about the informational events. And the conclusion that we have is exactly the same conclusion as you would get in the case. 285 00:34:36,560 --> 00:34:41,660 But now that the full logical omniscience, higher order reasoning, 286 00:34:41,660 --> 00:34:52,400 reflecting that you're part of an informational cascade is not enough for you to rationally make a different decision and get out of this cascade. 287 00:34:52,400 --> 00:35:01,460 If you look at the informational cascade, note that the the strength of evidence that agents have doesn't get higher, 288 00:35:01,460 --> 00:35:11,090 doesn't rise as the cascade progresses, because the third guess in this example itself is actually uninformative. 289 00:35:11,090 --> 00:35:15,710 Only the first two guesses, if you think about example, are informative. 290 00:35:15,710 --> 00:35:23,900 So the information of the fourth person or the person on the line is actually exactly the same. 291 00:35:23,900 --> 00:35:33,470 Of course, a reality we see that people are reinforced when they see millions of other agents and from them all making the same decision, 292 00:35:33,470 --> 00:35:42,560 even though maybe the only informative part are the decisions made by the by the first two agents and not everybody else around it. 293 00:35:42,560 --> 00:35:48,950 But it might increase. In reality, we see that people tend to have a stronger level of belief, 294 00:35:48,950 --> 00:35:54,380 even though that is done in that case for the unexampled, that would not be justified. 295 00:35:54,380 --> 00:36:02,150 What we did in our work is, well, it's a different paper here on the dynamic epistemic logics of diffusion and prediction and social 296 00:36:02,150 --> 00:36:11,120 networks is that we generalise the case of Cascade's in in this linear networks to arbitrary networks. 297 00:36:11,120 --> 00:36:16,370 So arbitrary, I mean, any type of topology, not necessarily linear line. 298 00:36:16,370 --> 00:36:25,610 So we looked at how fashion and behaviour propagates in a network of agents with its theory of the mind. 299 00:36:25,610 --> 00:36:33,110 And in particular here, these agents can anticipate or predict the behaviour of this of their friends. 300 00:36:33,110 --> 00:36:35,300 So they use their higher order, 301 00:36:35,300 --> 00:36:44,300 reasoning power to get reason about what the other agents are going to do and they can predict the behaviour of the others. 302 00:36:44,300 --> 00:36:51,080 So what we saw the conclusion of that paper is that knowledge does make a difference. 303 00:36:51,080 --> 00:36:53,830 So if you look at the speed of how fast? 304 00:36:53,830 --> 00:37:03,640 It emerges then the epistemic ingredient is very important because knowledge can speed up the spread of this cascade formation, 305 00:37:03,640 --> 00:37:10,560 so the spread of behaviour while on the other hand, uncertainty can slow it down. 306 00:37:10,560 --> 00:37:17,430 Even more, if it is common knowledge that everybody knows everybody's preferences or behaviour, 307 00:37:17,430 --> 00:37:21,900 and of course, also they need to have some knowledge about the network. 308 00:37:21,900 --> 00:37:27,460 So the friendship relation. So who's connected to who then, those gates? 309 00:37:27,460 --> 00:37:34,890 Could the principle be reached immediately? So that will immediately speed up the formation of the of the whole cascade. 310 00:37:34,890 --> 00:37:42,810 So, again, the analysis that we did for that paper is fully based on dynamic epistemic logic where you can make 311 00:37:42,810 --> 00:37:51,750 these aspects also the prediction of behaviour of others fully explicit in the logical language. 312 00:37:51,750 --> 00:37:58,620 Now, these cascades happened in reality, and there are many examples of newspapers are full of them. 313 00:37:58,620 --> 00:38:03,510 But let me give you one example that comes from the Netherlands. 314 00:38:03,510 --> 00:38:09,570 So this is an example that hit the newspapers and it was called Project X. 315 00:38:09,570 --> 00:38:21,420 So it brings me to 2012. So we we are in the city of Hama and Hanna is close to Thorning in the north northeast of the Netherlands. 316 00:38:21,420 --> 00:38:31,920 So in 2012, Miftah invited all her Facebook friends for her 16th birthday party and she made the 317 00:38:31,920 --> 00:38:38,640 message public on Facebook so that friends could bring in principle other friends. 318 00:38:38,640 --> 00:38:45,570 And what we saw is that one friend invited five hundred other people, so five hundred other friends. 319 00:38:45,570 --> 00:38:49,770 And then a day later, sixteen thousand people signed up for the party. 320 00:38:49,770 --> 00:38:57,030 So by now, Metho and her parents, they were getting afraid of this and said, like, OK, this is getting out of hand. 321 00:38:57,030 --> 00:39:02,070 This was not the intention. We are going to cancel the party. The party was cancelled. 322 00:39:02,070 --> 00:39:06,870 But what happened on Facebook is that other organisers took over. 323 00:39:06,870 --> 00:39:12,210 They called it Project X and they they kept sending out the announcement for the party. 324 00:39:12,210 --> 00:39:15,630 So very soon. Thirty thousand people signed up. 325 00:39:15,630 --> 00:39:21,690 And exactly on her birthday, four hundred thousand messages were posted on Twitter. 326 00:39:21,690 --> 00:39:26,460 And in the evening for the birthday party, still thousands came to harden. 327 00:39:26,460 --> 00:39:33,640 And this and the very bad side ended with riots with the police metho and her friends and her family. 328 00:39:33,640 --> 00:39:44,360 They had to leave the town. They couldn't. The whole street, I think, was evacuated, or at least her family had to evacuate. 329 00:39:44,360 --> 00:39:54,200 Examples of this type and saw, especially via Facebook and Twitter, we can see these online cascade's emerging. 330 00:39:54,200 --> 00:39:58,430 If you look back at the classic example, specifically in the iPhone example, 331 00:39:58,430 --> 00:40:08,090 then we saw that the cascade happened because agents were rewarded for their individual through tracking capabilities. 332 00:40:08,090 --> 00:40:13,880 So they were I mean, we didn't make the reward or the payoff explicit in their own example. 333 00:40:13,880 --> 00:40:25,190 But the way to set up is they actually only looked at their individual evidence and they tried to get each of them individually to to make it correct. 334 00:40:25,190 --> 00:40:30,170 Guess for the earth to be W or B now, you could change that. 335 00:40:30,170 --> 00:40:36,620 You could make the playoffs explicit and change these payoffs, rewarding the agents, 336 00:40:36,620 --> 00:40:43,700 not when they got to the individual rights of their individual guests would be correct, 337 00:40:43,700 --> 00:40:51,300 but maybe rewarding the agents only if and only if the whole majority, the whole group tracked the truth. 338 00:40:51,300 --> 00:41:00,170 And in that case, the cascade should not form because rational players will then disregard the information that 339 00:41:00,170 --> 00:41:05,930 they received themselves and they'll take the information that they got themselves into account. 340 00:41:05,930 --> 00:41:15,560 But disregard the guesses, the information that we received from the others, simply guessing the urn that matches to the colour that they saw. 341 00:41:15,560 --> 00:41:20,350 And they should take full advantage of the recessed herem, which was perfect, 342 00:41:20,350 --> 00:41:27,170 the tracking the truth when they disregard the communication of the others. 343 00:41:27,170 --> 00:41:35,590 Now, that means that if we look closer at the role of communication in this, so. 344 00:41:35,590 --> 00:41:42,490 In the cascading earn example, agents were allowed to communicate and then the Cascades could form, 345 00:41:42,490 --> 00:41:48,460 but it happened because there they were allowed to communicate their opinions, 346 00:41:48,460 --> 00:41:57,900 but not their justification, not their private evidence or their private reasons for making a certain guess. 347 00:41:57,900 --> 00:42:03,510 So communicate communities who do communicate all their private evidence, 348 00:42:03,510 --> 00:42:09,120 you could think while this solves maybe the problem in the informational and the unexampled, 349 00:42:09,120 --> 00:42:18,570 where we see that informational cascade happening because of people in that case would announce their private evidence again, should not happen. 350 00:42:18,570 --> 00:42:20,790 But that is not always the case. 351 00:42:20,790 --> 00:42:30,510 So communities who do communicate the private evidence aren't necessarily always guaranteed to track the truth and to do better. 352 00:42:30,510 --> 00:42:37,140 So there are many other cascades that can happen where to enrich too much communication 353 00:42:37,140 --> 00:42:45,030 actually is counterproductive and where too much communication leads us to a problem. 354 00:42:45,030 --> 00:42:51,960 So this is in formal epistemology and in social epistemology. 355 00:42:51,960 --> 00:42:59,520 This is what is called now the Solman effect. So it refers to Kevin Zillman, who works at the University of Pittsburgh, 356 00:42:59,520 --> 00:43:08,340 and he studied the effect of connectivity on the ability of the scientific community to track the truth, 357 00:43:08,340 --> 00:43:13,050 because we as scientists, we would think that more common connectivity in science is better. 358 00:43:13,050 --> 00:43:17,580 Right. So lots of conferences, lots of communication. That should be good. 359 00:43:17,580 --> 00:43:25,740 Now, what Kevin found and that shocked the community is that while the answer to this is actually no, 360 00:43:25,740 --> 00:43:33,180 there are contexts in which a community of scientists as a whole will be more reliable 361 00:43:33,180 --> 00:43:39,750 when the members are actually less aware of the experimental results of their colleagues. 362 00:43:39,750 --> 00:43:51,990 And the reason for that is that too early elimination of correct alternative theories or options or experiments that can actually block the 363 00:43:51,990 --> 00:44:00,540 group from tracking the truth so they influence each other and they might just decide not to explore these alternative options anymore, 364 00:44:00,540 --> 00:44:05,440 which might be the key for the whole community to get it right. 365 00:44:05,440 --> 00:44:10,480 So let me give a quote of Solomon from his 2007 paper. 366 00:44:10,480 --> 00:44:15,560 So he says that in circumstances where speed is very important, 367 00:44:15,560 --> 00:44:25,300 but where we think that our initial estimates are like likely very close to the truth, then connected groups of scientists will be more reliable. 368 00:44:25,300 --> 00:44:33,670 So if we already happened to be close to the truth, that will do well. On the other hand, where we want accuracy above all else, 369 00:44:33,670 --> 00:44:46,930 then we should prefer communities made up of more isolated individuals so that the number of conferences that we see exploding in our communities, 370 00:44:46,930 --> 00:44:54,420 maybe not necessarily always good for the whole community to get to the truth. 371 00:44:54,420 --> 00:44:59,670 An example of a cascade in which more communication is harmful, 372 00:44:59,670 --> 00:45:06,120 all the while there are many examples, but it would lead, for instance, to believe polarisation. 373 00:45:06,120 --> 00:45:18,600 So believe polarisation is an informational cascade that is then produced by the widespread mutual communication of a preliminary test results. 374 00:45:18,600 --> 00:45:27,870 So not the final results. And that leads to two early elimination of alternative options or theories that are out there. 375 00:45:27,870 --> 00:45:38,040 And together it leads to a strengthening of a biased majority opinion, and that biased majority opinion will not track the truth. 376 00:45:38,040 --> 00:45:49,410 So in these scenarios, more communication means that it is much more likely that polarisation can actually occur. 377 00:45:49,410 --> 00:45:53,430 So then the question is, how can we actually avoid this problem? 378 00:45:53,430 --> 00:46:04,230 So how can we get out of that? Well, Spirit Sparks Networks would be would be an answer one would need to preserve the opinions 379 00:46:04,230 --> 00:46:12,570 independence for long enough so that enough alternative theories and approaches are checked out. 380 00:46:12,570 --> 00:46:16,200 As a result, the probability of a cascade will then be lower. 381 00:46:16,200 --> 00:46:24,310 So on these past networks, they are more likely to be through tracking in the long run. 382 00:46:24,310 --> 00:46:35,380 Now, of course, there is a Trade-Off between efficiency and speed and on the one hand and the reliability on the other hand, 383 00:46:35,380 --> 00:46:42,830 by reliability I mean here the avoiding the avoidance of a cascading, merging or polarisation happening. 384 00:46:42,830 --> 00:46:49,040 So because these past networks there may be very good for science, 385 00:46:49,040 --> 00:46:55,310 but there are also very slow at converging to any common opinion whether that opinion is true or not. 386 00:46:55,310 --> 00:47:00,800 So a very connected network converges very fast. 387 00:47:00,800 --> 00:47:09,350 And so when it hits the truth, it does so in a very efficient way and much more efficient than a sports network. 388 00:47:09,350 --> 00:47:14,480 But we are, of course, not guaranteed that it will hit actually truth if it is very connected. 389 00:47:14,480 --> 00:47:19,400 That's exactly what Kevin shows in his work. 390 00:47:19,400 --> 00:47:28,670 So. So indeed, we have this trade off between speed and reliability and what the community would meet and also what the community of scientists would 391 00:47:28,670 --> 00:47:40,850 meet would be a communication network that lies somewhere in the middle ground between extreme sparsity and extreme connectivity. 392 00:47:40,850 --> 00:47:49,400 Now, of course, there are agencies out there that know these theories and they can use it against the scientists. 393 00:47:49,400 --> 00:47:58,040 So you can if you know that alternative theories are necessary to progress science and that 394 00:47:58,040 --> 00:48:05,180 more alternative theories need to be investigated before actually full opinions are formed, 395 00:48:05,180 --> 00:48:14,780 then that can be used to create doubt and to undermine scientific truths, maybe when they have already been established. 396 00:48:14,780 --> 00:48:20,360 So when a community is too slow in tracking and maybe even just to convey the truth, 397 00:48:20,360 --> 00:48:25,290 even when they know it already, it will leave room for unwanted effects. 398 00:48:25,290 --> 00:48:32,870 So so I recently read an article in The New York Times and the headline there of that 399 00:48:32,870 --> 00:48:39,440 article was Influencers say that they were urged to criticise the fire's SURFAXIN. 400 00:48:39,440 --> 00:48:49,910 So disinformation efforts to reduce public confidence and covid-19 vaccinates try to role social media commentators in France and Germany. 401 00:48:49,910 --> 00:48:59,930 So you see that this idea that diversity of attempts and theories is good for the progress of science. 402 00:48:59,930 --> 00:49:08,480 Less activity is good. We don't need to converge to the truth too fast can also be used against us. 403 00:49:08,480 --> 00:49:18,170 So, for instance, when people build an on their mind, a truth that has been already established and tried to communicate that very fast. 404 00:49:18,170 --> 00:49:24,980 So if you take stock now of the different epistemic and social dilemmas that I've touched upon, 405 00:49:24,980 --> 00:49:33,380 then a main factor that you'll see is that indeed rational individuals can form 406 00:49:33,380 --> 00:49:40,790 irrational groups and rational groups might well be composed of irrational individuals. 407 00:49:40,790 --> 00:49:48,930 This, of course, is nothing new in formal epistemology. This is known as the independence thesis. 408 00:49:48,930 --> 00:49:59,750 And this refers to a paper that was written and published in Philosophy of Science in 2011 by Michael Wilson Commensal and they could Dang's. 409 00:49:59,750 --> 00:50:10,290 So they also on their paper, give lots of examples that come from philosophy, science game theory, social choice theory. 410 00:50:10,290 --> 00:50:14,730 Now, the question is, what kind of what can logic do? 411 00:50:14,730 --> 00:50:21,390 So can logic help us understand these social epistemic dilemmas? 412 00:50:21,390 --> 00:50:31,210 And I would argue that, well, yes, logic can provide us with good models to reason and predict about this epistemic distortion. 413 00:50:31,210 --> 00:50:39,720 So for for a diagnosis that we want to to put on the problem, then, yes, logic is a very good tool. 414 00:50:39,720 --> 00:50:44,670 And examples are given by the examples that I gave in the presentation here. 415 00:50:44,670 --> 00:50:52,230 But so we use dynamic epistemic logic to investigate the formation of these informational cascades. 416 00:50:52,230 --> 00:51:05,790 We also study a and social properties in interaction when we analyse the the behaviour of both online behaviour of influencers. 417 00:51:05,790 --> 00:51:13,590 What triggers the formation of a polarised network of these filtering mechanisms that you might see in an echo chambers, 418 00:51:13,590 --> 00:51:24,810 the effects of homophily, etc.? So there is, of course, network theory out there and a lot of simulations and models are provided by network theory. 419 00:51:24,810 --> 00:51:30,360 But most of the models on the market don't take the epistemic ingredients into account. 420 00:51:30,360 --> 00:51:37,830 And I think for studying groups of rational agents, the study criteria are crucial. 421 00:51:37,830 --> 00:51:47,370 So that leaves another question. So can computer science help us verify and debunk our social software, diagnose the weaknesses, 422 00:51:47,370 --> 00:51:53,830 provide tools that could be used to prevent the worst forms of groupthink? 423 00:51:53,830 --> 00:51:58,860 And maybe that's a question for computer scientists and the audience. 424 00:51:58,860 --> 00:52:05,400 So that brings me to my last slide. So here's my conclusion. 425 00:52:05,400 --> 00:52:10,200 So let's look back at the question of the lecture. So will the facts. 426 00:52:10,200 --> 00:52:17,940 And then, I mean, the actual facts, things that are true in the world eventually emerge from the noisy network of irrelevant data, 427 00:52:17,940 --> 00:52:25,510 self confirming rumours, idiosyncratic likes, or, to put it more bluntly, will truth survive the information age? 428 00:52:25,510 --> 00:52:35,250 Now, I am an optimist. So according to me, knowing the mechanisms for mass information aggregation, for information exchange, 429 00:52:35,250 --> 00:52:43,710 we have to learn and know the epistemic benefits and costs and all that is just a start, but it is a very good start. 430 00:52:43,710 --> 00:52:50,040 It will hopefully allow us to perform some of the most urgent interventions very soon. 431 00:52:50,040 --> 00:52:53,920 I know that some interventions already are being taken. 432 00:52:53,920 --> 00:53:02,820 So there is also when you look at how the platform, the main social platforms are being organised, there is input from the legal side, 433 00:53:02,820 --> 00:53:11,910 their actions on decisions made by politicians, the platforms themselves, also how fact checking algorithms, tool to fight fake news, etc. 434 00:53:11,910 --> 00:53:22,830 So that is very good. But I also think that this epistemic factors need to be studied more and they play a crucial role in the formation of these 435 00:53:22,830 --> 00:53:32,540 online cascade's and all these negative side effects and distortions that we saw for the different types of wisdom of the crowds. 436 00:53:32,540 --> 00:53:36,300 So with this, I think this concludes my lecture. 437 00:53:36,300 --> 00:53:44,540 So thanks for listening and for attending, and I don't give the floor back to some. 438 00:53:44,540 --> 00:53:54,620 Thank you very much. But that was a wonderful lecture. And you showed us how logic and fútbol methods can reach deep into a lot of the prominent 439 00:53:54,620 --> 00:54:02,630 features of our information age in a way that I think many of us will be really illuminating. 440 00:54:02,630 --> 00:54:10,400 So, OK, so we we couldn't go to Q&A. 441 00:54:10,400 --> 00:54:23,180 We if you do have questions for Sonia, please send the questions on the on the Q&A panel, which you can find, as I said, with the icon. 442 00:54:23,180 --> 00:54:33,020 But the question mark. So let me just start with a couple that have been sent, one coming, I think, from a philosophical side, 443 00:54:33,020 --> 00:54:39,500 simply asking, since you speak of truth, what is the notion of truth that you're working with? 444 00:54:39,500 --> 00:54:47,180 Is it as simple as a simple bubble theory account, enough to fully encapsulate truth? 445 00:54:47,180 --> 00:54:51,890 So that's the question. Oh, yes. The notion of truth. 446 00:54:51,890 --> 00:54:57,110 I mean, it's the one that we use and logics of correspondence to the facts, I would say. 447 00:54:57,110 --> 00:55:02,240 So it's a very simple one. It's not so that many different notions of truth out there. 448 00:55:02,240 --> 00:55:07,410 So, for instance, coherence, theories of truth, where information needs to be coherent or things like that. 449 00:55:07,410 --> 00:55:12,650 So that's not the one I'm referring to, but just something that corresponds to basic facts and reality. 450 00:55:12,650 --> 00:55:23,090 So something that can always be checked. So, yes, of course, also on the logic side there, 451 00:55:23,090 --> 00:55:29,810 there are logics to deal with different theories of truth and to formalise that explicitly, but that we don't do so. 452 00:55:29,810 --> 00:55:39,060 It's always mixture. An external factor to the models that we have is somewhere a little point that says this is the true option. 453 00:55:39,060 --> 00:55:46,400 Yes, this is the model who has to build in that option and he needs to have that information available. 454 00:55:46,400 --> 00:55:52,070 So we assume that that is possible to get that information available, even if the agents don't know it. 455 00:55:52,070 --> 00:55:56,890 So agencies don't necessarily have to know that something is true. 456 00:55:56,890 --> 00:56:02,390 Knowledge is truthful, but not necessarily all the agents in the model will have access to that point, 457 00:56:02,390 --> 00:56:07,870 that is the real world, but the modellers should have that have access to that point. 458 00:56:07,870 --> 00:56:19,610 No. OK, thanks. So another question, what would your what would your suggestions be for us to come? 459 00:56:19,610 --> 00:56:25,780 I mean, this is the sort of the big practical question that, you know, they so the question is, 460 00:56:25,780 --> 00:56:38,140 what would your suggestions be for us to combat fake news in today's super connected society with all the powerful social media and, 461 00:56:38,140 --> 00:56:44,320 for example, Facebook, whose whole kind of mission is to make people more connected? 462 00:56:44,320 --> 00:56:48,320 So that's what they're all about. Oh, yeah. 463 00:56:48,320 --> 00:56:51,620 So there there are many things that you can do to combat fake news. 464 00:56:51,620 --> 00:57:02,590 So I think an important one is to educate people, self critical thinking, self reflection, sending our students to school, 465 00:57:02,590 --> 00:57:10,720 giving them logic classes from the beginning, that that is crucial for critical thinking because and I do think that will help. 466 00:57:10,720 --> 00:57:11,800 So that is one thing. 467 00:57:11,800 --> 00:57:22,330 So educate people so that they can they can think first when they see something, without retweeting it, without sending it further, 468 00:57:22,330 --> 00:57:30,370 if it is not coming from a reliable source or not from something that they're absolutely certain about. 469 00:57:30,370 --> 00:57:33,490 So that would be one thing then. 470 00:57:33,490 --> 00:57:44,110 Currently, I think some universities designed this fact checking games that people can play so people can then see how well they how good they are. 471 00:57:44,110 --> 00:57:50,600 In fact, checking some of the fake news themselves and can they discern the fake from the real news articles. 472 00:57:50,600 --> 00:57:56,710 So these games are excellent because means that people will then think about these things and they can they can learn. 473 00:57:56,710 --> 00:58:02,650 And are we also see that the younger generation actually is already much better at doing 474 00:58:02,650 --> 00:58:08,290 so than the ones that are growing up with all the social media and are fully connected. 475 00:58:08,290 --> 00:58:11,320 They will not necessarily believe every message that they see. 476 00:58:11,320 --> 00:58:23,110 So they already have built in better mechanisms of finding or ways of of eliminating the falsehoods from the truth. 477 00:58:23,110 --> 00:58:28,400 Now, of course, the platforms themselves can build fact checking algorithms, which they do. 478 00:58:28,400 --> 00:58:32,800 So there is a lot of work on the computer science side and that should continue. 479 00:58:32,800 --> 00:58:37,840 I think that is highly urgent that we do that, that should improve. 480 00:58:37,840 --> 00:58:44,230 So I don't have details because I didn't look at the algorithms and how they are designed and how they work. 481 00:58:44,230 --> 00:58:54,490 But that is, of course, very good. And then regulation is the other thing that the politicians need to regulate the information that is being spread, 482 00:58:54,490 --> 00:59:06,170 because the newspapers that are sending out information and the news channels, they have very, very clear and strict code of conduct. 483 00:59:06,170 --> 00:59:11,120 They they know what they are allowed to do. And they are not just broadcasting anything. 484 00:59:11,120 --> 00:59:22,390 So we know when a news agency is a reliable news agent, but on these new information platforms and social media, we don't have these these mechanisms. 485 00:59:22,390 --> 00:59:27,850 So they are totally not yet regulated up to the level that it should be in order to 486 00:59:27,850 --> 00:59:35,660 prevent people from being influenced by fake news and taking it as a truthful fact. 487 00:59:35,660 --> 00:59:39,280 So, yeah, so there are many things that we should do. 488 00:59:39,280 --> 00:59:43,250 And it starts with education. I think these games that are on the market. 489 00:59:43,250 --> 00:59:51,310 So there's a lot that people themselves can do and making them aware of this is is also very good. 490 00:59:51,310 --> 00:59:58,930 OK, great, thanks. We have a there was a there was another question by Mike and Mike, Mike Wooldridge, which is similar, I think, 491 00:59:58,930 --> 01:00:11,290 to what you've just been answering, sir, about this sort of work on automated fact checking and fake news, fake news detection and so on. 492 01:00:11,290 --> 01:00:17,280 I think he's mentioning in particular using deep learning to detect fake news. 493 01:00:17,280 --> 01:00:25,060 So he's asking if you think logic can play a practical role in this as well? 494 01:00:25,060 --> 01:00:27,790 Well, yes, I think that I mean, 495 01:00:27,790 --> 01:00:39,490 logic is the science for truth plays an essential role where I would say that the machine learning algorithms they don't necessarily have. 496 01:00:39,490 --> 01:00:45,070 So they might be able to deal with a large set of data. 497 01:00:45,070 --> 01:00:54,850 But the way that they deal with that data is not necessarily that they need to have an exact item that is labelled as the true one, right. 498 01:00:54,850 --> 01:00:58,990 So logic inherently is about valid inference. 499 01:00:58,990 --> 01:01:04,300 The notion of validity and truth is built into the whole practise of how logic works. 500 01:01:04,300 --> 01:01:10,480 That is not necessarily the case in the machine learning algorithms, of course, for fact checking. 501 01:01:10,480 --> 01:01:20,260 They will have something that allows them to to find the source of the information that goes to to the the one that is labelled correct. 502 01:01:20,260 --> 01:01:23,920 But it will then be automatically from the outside. 503 01:01:23,920 --> 01:01:33,880 Still needs to be labelled as this is the correct one. I think that logic can play a role and it should play a role because for but 504 01:01:33,880 --> 01:01:38,050 that is more an answer for not only for fact checking in the fake news case, 505 01:01:38,050 --> 01:01:49,210 but for all of A.I. where you do want to make you want to give an explanation of why a certain decision is made. 506 01:01:49,210 --> 01:01:56,680 Right. And so the machine learning algorithms by themselves are not necessarily very good at exactly that point. 507 01:01:56,680 --> 01:02:00,790 So explaining how you get from input to output. 508 01:02:00,790 --> 01:02:05,800 So that whole track of the inferences is something where logic is very good. 509 01:02:05,800 --> 01:02:17,440 And so combining these two methods would be ideal, where you have one tool that can handle large datasets much better than what the logic side can do. 510 01:02:17,440 --> 01:02:22,070 But for the explanation part, according to me, logic is very good. 511 01:02:22,070 --> 01:02:27,350 So so ideally, you would build a new system that combines the two. 512 01:02:27,350 --> 01:02:32,650 I know that people are working on this, but yeah, that's actually a big challenge. 513 01:02:32,650 --> 01:02:37,370 So combining the old and new AI together. Yes. 514 01:02:37,370 --> 01:02:44,630 OK, thanks very much then. Well, actually, let me ask another question. 515 01:02:44,630 --> 01:02:51,950 This is, do you think there could be an application for research and market simulation? 516 01:02:51,950 --> 01:03:06,220 Um, I think so. I mean, so the one of the things that we are doing now is, for instance, using model checking to verify. 517 01:03:06,220 --> 01:03:13,180 So you use logic to specify certain properties of behaviour of the network, of the agents, 518 01:03:13,180 --> 01:03:18,670 of the beliefs that they have now, then you can apply that to a whole bunch of situations. 519 01:03:18,670 --> 01:03:23,490 And then the question is, OK, this gives us insight in the behaviour. 520 01:03:23,490 --> 01:03:35,990 And so you could use it, for instance, when when agents anticipate that you are going to buy the same type of shoes that as your friends, et cetera. 521 01:03:35,990 --> 01:03:44,230 So so if we can anticipate that a cascade is going to happen, then we can predict what the next fashion is going to be. 522 01:03:44,230 --> 01:03:46,240 Of course, it is very difficult to do that. 523 01:03:46,240 --> 01:03:53,740 Many factors play a role and we, the logicians typically abstract away and reason at a very high level of abstraction. 524 01:03:53,740 --> 01:04:03,310 But could the principle be used for that? Now, if you actually want an actual application, then I think so. 525 01:04:03,310 --> 01:04:12,250 Going back to the to the A.I. story, what you could do there is that in cases where you don't have a lot of data available and 526 01:04:12,250 --> 01:04:19,030 you want to reason very exactly about the behaviour and the specifications of the agents, 527 01:04:19,030 --> 01:04:26,480 you could do run a model tracking algorithm for that behaviour to check if it is true in a network. 528 01:04:26,480 --> 01:04:32,050 Right. So it will be much more accurate than some of the other algorithms that are out there. 529 01:04:32,050 --> 01:04:38,860 So then you would use some of the tools from computer science with the logical specification of these properties, 530 01:04:38,860 --> 01:04:45,100 and you would do the role model checking on these networks that can lead directly to some application. 531 01:04:45,100 --> 01:04:56,230 And you could. Yeah, it's not necessarily for predicting what the market's going to be or what will happen to cryptocurrency or with the next version. 532 01:04:56,230 --> 01:05:06,800 But so this is more on spotting the influencers or the bots in the network, for instance. 533 01:05:06,800 --> 01:05:16,460 Right. Yes, thanks. If I could add a question of I mean, I find this discussion fascinating on one of the other information, 534 01:05:16,460 --> 01:05:26,030 Kaskade example that you mentioned, but the models with communication and the way that this kind of defeated event sort of dynamic, 535 01:05:26,030 --> 01:05:32,960 dynamic, epistemic well, analysis with where you could model higher order beliefs in some sense, 536 01:05:32,960 --> 01:05:42,440 I'm following a comprehensive analysis that you still wouldn't kind of refute this or detect in some sense this information. 537 01:05:42,440 --> 01:05:49,710 Cascais, on the other hand, you know, we standing outside can see that there's something wrong here. 538 01:05:49,710 --> 01:05:50,420 So, I mean, 539 01:05:50,420 --> 01:06:00,800 does this suggest a further kind of level of analysis we could put back into a logic that would do a better job of detecting information flow? 540 01:06:00,800 --> 01:06:06,110 Cascades, Cascade's? Yeah, indeed. 541 01:06:06,110 --> 01:06:12,440 Of course. So if we would model the modellers that are modelling the phenomena so that some kind of. 542 01:06:12,440 --> 01:06:22,310 Yes, some kind of reflection, you need an extra reflection, which is not just what the agents that are making the decisions, 543 01:06:22,310 --> 01:06:30,650 but that is just the the ones from the outside who are looking at this and strategically may be reasoning about it. 544 01:06:30,650 --> 01:06:36,770 And we see this happening. Right. So agencies are looking at the formation of these cascades, 545 01:06:36,770 --> 01:06:44,660 because if I can predict the market and maybe just start market cascade so that everybody will buy the same type of shoes tomorrow, 546 01:06:44,660 --> 01:06:47,300 namely the grant that I might be producing, 547 01:06:47,300 --> 01:06:55,550 then I can make sure that I can manipulate the first agents in the group and hope that the North will follow. 548 01:06:55,550 --> 01:07:05,870 And I can set the trends right. So so these mechanisms mean that you don't understand how this works and there are models for it. 549 01:07:05,870 --> 01:07:13,010 But these are at a higher level because at the level of the agents themselves, they're higher order reasoning. 550 01:07:13,010 --> 01:07:16,130 Power is not going to break the code, but the ones from the outside, 551 01:07:16,130 --> 01:07:23,630 they can say they can see exactly where to intervene, where they could manipulate the formation of the cascade. 552 01:07:23,630 --> 01:07:31,370 They know exactly which information to drop out, at what point in order for the cascade to derail or to stop. 553 01:07:31,370 --> 01:07:36,080 So, yeah, so that we didn't do that in our logical setting. That would be very interesting. 554 01:07:36,080 --> 01:07:47,370 Yes. Shows how subtle the and rich the sort of modelling capacity that one is led to and analysing these situations. 555 01:07:47,370 --> 01:07:51,260 I think also having to make a comment, taking out what's already been said. 556 01:07:51,260 --> 01:07:58,850 We now see there's a whole genre of books that because there's been a lot of statistics around in the public domain with the pandemic's, 557 01:07:58,850 --> 01:08:03,960 we see a whole chapter of books covid trying to interpret this for the public at large. 558 01:08:03,960 --> 01:08:09,460 It occurs to me that maybe something else also brings in this epistemological 559 01:08:09,460 --> 01:08:14,330 ideological aspect would be a very valuable resource for the for the general public. 560 01:08:14,330 --> 01:08:24,890 So maybe in some of the spare time for all the other things that you're doing, you should think of doing that. 561 01:08:24,890 --> 01:08:34,360 OK, so do you have any any other questions for Sonia? 562 01:08:34,360 --> 01:08:40,330 OK, if not that I don't see any right now, then yes. 563 01:08:40,330 --> 01:08:51,820 Well, thank you very much again for a fascinating lecture. And it's really been shown that logic is going into these new and actually very central 564 01:08:51,820 --> 01:08:58,270 and important areas and in current life and doing so with a lot of incisive ideas. 565 01:08:58,270 --> 01:09:02,410 So thank you very much, Tanya. Thank you for the invitation. 566 01:09:02,410 --> 01:09:08,670 Was almost a pleasure to give this lecture. OK, so that will conclude there. 567 01:09:08,670 --> 01:09:11,401 OK, thanks, everyone.