WEBVTT

1
00:00:02.610 --> 00:00:10.410
Hannah Pyman: Thank you for joining us for our third 'Newcomers Presents' webinar. My name is Hannah and I'm the scholarly communications coordinator in the library.

2
00:00:10.980 --> 00:00:18.480
Hannah Pyman: My colleague Keziah Gibbs and myself will be managing the webinar today, so if you have any problems or questions, please just send us a chat message.

3
00:00:19.380 --> 00:00:28.350
Hannah Pyman: We encourage you to ask questions via the Q&A function here in Zoom when the panelists are presenting, and we'll be asking these questions to our speakers at the end of the webinar.

4
00:00:29.220 --> 00:00:39.690
Hannah Pyman: With us today. We have Pablo Cabrera Álvarez, Lukas Griessl,  Sabrina Rau, and Shahin Salarian, who all have very interesting and different topics to talk about.

5
00:00:41.580 --> 00:00:52.260
Hannah Pyman: Our first speaker today is Pablo Cabrera Álvarez, a Senior Research officer at the institute for economic and social research, who is about to complete a PhD in social sciences.

6
00:00:52.950 --> 00:01:03.180
Hannah Pyman: Pablo will talk about polls and US presidential elections. So I'm going to hand over to you now Pablo, if you want to start sharing your screen.

7
00:01:04.620 --> 00:01:05.190
Pablo Cabrera Álvarez: Thank you.

8
00:01:06.750 --> 00:01:08.400
Pablo Cabrera Álvarez: I'll just share the screen first.

9
00:01:13.980 --> 00:01:14.460
Pablo Cabrera Álvarez: Is that OK?

10
00:01:15.120 --> 00:01:15.360
Yeah.

11
00:01:17.310 --> 00:01:24.270
Pablo Cabrera Álvarez: Perfect. So hi, everyone. Thank you very much for inviting me to this seminar.

12
00:01:25.290 --> 00:01:41.370
Pablo Cabrera Álvarez: My name is Pablo Cabrera and I work at the Institute for Social and Economic Research. I normally do methological research and surveys, and longitudinal surveys, but today I'm going to talk about something more close to what I did for my PhD.

13
00:01:41.910 --> 00:01:59.400
Pablo Cabrera Álvarez: I'm going to talk about the recent presidential election in the US and how polls performed at the election. So yes, to start, like the pause before the election

14
00:02:00.330 --> 00:02:04.830
Pablo Cabrera Álvarez: During a scenario in which Joe Biden, the Democratic candidate was going to

15
00:02:05.580 --> 00:02:15.900
Pablo Cabrera Álvarez: have a comfortable margin, he was going to easily win the election, but finally we have to wait more than seven days for the final outcome.

16
00:02:16.380 --> 00:02:31.560
Pablo Cabrera Álvarez: And there is some voices saying that polls failed again. So today I want to, I want to see whether those polls failed and what reasons can be behind that failure.

17
00:02:32.460 --> 00:02:36.570
Pablo Cabrera Álvarez: First of all, I want to zoom our. I want to zoom out and

18
00:02:37.050 --> 00:02:50.790
Pablo Cabrera Álvarez: show you that polling failures are not new in the last century. There have been manual elections, many countries, many locations in which polls deviated from their final election result.

19
00:02:51.570 --> 00:03:01.440
Pablo Cabrera Álvarez: Just to focus on a couple of them. Recently in the UK, in 2015 polls predicted a tie between

20
00:03:02.250 --> 00:03:15.990
Pablo Cabrera Álvarez: the Labour candidate Ed Miliband and the Conservative David Cameron finally, on election day, David Cameron won by seven percentage points of the popular vote.

21
00:03:16.770 --> 00:03:40.560
Pablo Cabrera Álvarez: Similar happened in the US four years ago. Poll models all predictive victory for Hilary Clinton, the Democratic candidate. But at the end of the day it was stolen by Trump, who won the White House. So I want to get back to that moment to

22
00:03:42.180 --> 00:03:52.590
Pablo Cabrera Álvarez: see what went wrong. Our starting point for what happened in this election that election, what went wrong, was that the polls

23
00:03:53.070 --> 00:04:05.610
Pablo Cabrera Álvarez: incorrectly estimated the national vote, popular vote, and they moved within the margin, expected margin of error of the polls. However, in some

24
00:04:06.570 --> 00:04:14.100
Pablo Cabrera Álvarez: key states, battleground states, the polls deviated much more from the actual results. And what happened there was

25
00:04:14.850 --> 00:04:22.200
Pablo Cabrera Álvarez: some research after the election and they found that there was first of all a late swing,

26
00:04:22.830 --> 00:04:30.360
Pablo Cabrera Álvarez: so some Trump voters decided to vote in the very last minute. So some of the polls couldn't reflect that movement of the electorate.

27
00:04:31.140 --> 00:04:46.440
Pablo Cabrera Álvarez: Then there was also problems in the sample composition. So, for example, voters with college education or higher education were over represented in the polls and these voters

28
00:04:47.070 --> 00:04:51.570
Pablo Cabrera Álvarez: tend to vote for democratic candidates in a higher proportion.

29
00:04:52.110 --> 00:05:05.880
Pablo Cabrera Álvarez: And also there was the shy Trump theory, which says that some Trump voters are ashamed of voting, for admitting that they are going to vote for Trump, and they try to hide their voting intention.

30
00:05:06.450 --> 00:05:13.260
Pablo Cabrera Álvarez: So after this diagnostic what pollsters did was to

31
00:05:13.770 --> 00:05:33.750
Pablo Cabrera Álvarez: make some amendments and change their methods. So, for example, they ensured that the sample was representative in terms of education and also in terms of where the sample comes from. Is it from rural or urban areas? These two parts were important

32
00:05:34.890 --> 00:05:37.800
Pablo Cabrera Álvarez: to understand the electoral behaviour in the US.

33
00:05:38.220 --> 00:05:44.730
Pablo Cabrera Álvarez: Also, some of them updated the sampling methods. So, for example, they were generating random telephone numbers to call people.

34
00:05:44.970 --> 00:05:58.920
Pablo Cabrera Álvarez: And they moved to all the types of sampling. So, for example, using address based sampling, sending mail, recruiting people into a panel, doing surveys, and some of them even switched modes, or

35
00:06:00.990 --> 00:06:11.940
Pablo Cabrera Álvarez: used together different modes of administration. Some of them moved from landlines to cell phones.

36
00:06:13.170 --> 00:06:22.500
Pablo Cabrera Álvarez: Some of them combined different modes, for example, used SMS, internet, internet service, and also a telephone service altogether.

37
00:06:22.980 --> 00:06:36.780
Pablo Cabrera Álvarez: So what was the result? What happened actually in 2020? So what happened is that Donald Trump's voting share was underestimated in all key battleground states, as well as in the national average.

38
00:06:37.590 --> 00:06:48.240
Pablo Cabrera Álvarez: This is something very, very similar to what happened four years ago, four years ago in 2016.

39
00:06:48.900 --> 00:07:07.500
Pablo Cabrera Álvarez: So our question here is whether the diagnostic that the pollsters did four years ago was not good enough to tackle the errors, or what they tried to do didn't fix, or we have moved to a different scenario in which we have new sources of bias.

40
00:07:09.000 --> 00:07:18.420
Pablo Cabrera Álvarez: And about this I want to spend the last couple of minutes talking about some hypotheses that have

41
00:07:19.530 --> 00:07:30.360
Pablo Cabrera Álvarez: raised in the last few days related to what happened in this election. The first one is completely new, of course is about the COVID-19 pandemic. So what happened with the pandemic?

42
00:07:30.660 --> 00:07:41.730
Pablo Cabrera Álvarez: It happened that some people given their local situation could stay on the path that they were going to vote, and finally didn't go to the poll, or that they were going to vote by mail, which is

43
00:07:43.530 --> 00:07:53.340
Pablo Cabrera Álvarez: one of the possibilities, or anticipated vote, and finally, they didn't. What's the point that these people worried about COVID-19 were mostly democrat voters.

44
00:07:54.720 --> 00:08:06.390
Pablo Cabrera Álvarez: Another theory that has been raised is the shy Trump voters. This was four years ago, there was little evidence to support this, and this time I think that is not the case.

45
00:08:07.080 --> 00:08:20.040
Pablo Cabrera Álvarez: There is no large majority of people that are ashamed of admitting that they are going to vote for Trump. So there another two hypotheses. The first one is the likelihood to vote.

46
00:08:21.540 --> 00:08:33.240
Pablo Cabrera Álvarez: So the likelihood to vote is very important in polling. Why, because we need to estimate the voting, sure, so we need to know who's going vote. If you support a community if you don't go to the poll

47
00:08:33.810 --> 00:08:44.310
Pablo Cabrera Álvarez: you are not going to count. So the point there is that some Trump voters could have said that maybe they were not going to go to the polls for sure, but finally

48
00:08:44.880 --> 00:09:01.560
Pablo Cabrera Álvarez: they went. While thee democrat voters, with the same likelihood to vote, didn't go. This difference could be one of the reasons of that deviation. And finally, there is a very interesting hypotheses about the lack of trust in institutions.

49
00:09:02.970 --> 00:09:10.770
Pablo Cabrera Álvarez: So Trump has been moving in this framework and establishment, right, so he has tried to say that

50
00:09:11.940 --> 00:09:22.860
Pablo Cabrera Álvarez: the Congress, the administration, all of them are pushing against him. So it has made on effect in the last four years, the trust of

51
00:09:23.610 --> 00:09:29.400
Pablo Cabrera Álvarez: Republican voters on institutions like the Congress or like the

52
00:09:30.270 --> 00:09:40.350
Pablo Cabrera Álvarez: public administration, has to, but not only these, this also includes the media, and the media an an imposter. So it could be that there is a core of Trump voters

53
00:09:40.710 --> 00:09:59.250
Pablo Cabrera Álvarez: refusing to take part in service, not because they want to hide that they are Trump voters, or they aren't interested, but because they are making some kind of activists refusing to take part in this kind of what they would call manipulated polls, right.

54
00:10:00.390 --> 00:10:18.750
Pablo Cabrera Álvarez: So I just want to finish saying that there is this debate whether we are asking too much to polls, right. So whether polls are a good method to predict elections and political action when we have our heads.

55
00:10:19.530 --> 00:10:25.710
Pablo Cabrera Álvarez: And I stop it here. Thank you very much, and over to you, Hannah. Thanks.

56
00:10:28.230 --> 00:10:31.350
Hannah Pyman: Thank you so much Pablo, that was great, really interesting.

57
00:10:31.860 --> 00:10:40.260
Hannah Pyman: And as I said at the start, there'll be time for questions at the end. So if anyone does have any questions for Pablo please do put them in the Q&A.

58
00:10:41.910 --> 00:10:57.330
Hannah Pyman: Okay, so our second speaker today is Lukas Griessl. Lukas is a PhD student in sociology, and his talk is entitled 'statistical controversies: sampling and quantification'. So whenever you're ready. Lukas.

59
00:11:00.660 --> 00:11:01.020
Lukas Griessl: Thank you Hannah.

60
00:11:02.100 --> 00:11:04.440
Lukas Griessl: I will share my screen as well.

61
00:11:08.670 --> 00:11:09.210
Hannah Pyman: Perfect, thank you.

62
00:11:09.720 --> 00:11:10.710
Lukas Griessl: Cool. Great.

63
00:11:11.610 --> 00:11:15.060
Lukas Griessl: So, yeah. Hi, everyone. My name is Lukas and

64
00:11:16.320 --> 00:11:24.660
Lukas Griessl: as Hannah said I'm a PhD student in sociology here at Essex, at the end of my first year. And today in the presentation, I would like to

65
00:11:25.590 --> 00:11:34.740
Lukas Griessl: talk about a part of my PhD research, which deals with controversies within the field of statistics, especially with controversies in

66
00:11:35.250 --> 00:11:51.120
Lukas Griessl: regarding sampling strategies and it's great that Pablo was talking before me, because I think my topic seamlessly follows up on what Pablo just talked about, even though I'm not particularly dealing with failed polls

67
00:11:52.740 --> 00:11:55.500
Lukas Griessl: I'm rather dealing with a 

68
00:11:57.180 --> 00:12:08.550
Lukas Griessl: controversy diverging polls. And what I do is get an external perspective on statistics [unknown] and pollsters, but

69
00:12:08.970 --> 00:12:22.590
Lukas Griessl: I am to get insights into certain controversies and issues. So what my project is a qualitative research that explores debates in a quantitative discipline, so to speak.

70
00:12:23.730 --> 00:12:30.780
Lukas Griessl: So as part of this project I explore different case studies dealing with unknown uncertain and contested numbers and thereby with

71
00:12:31.470 --> 00:12:40.290
Lukas Griessl: inner statistical controversies around these numbers. I will then look at different case studies, whereas in the first case study that I'm researching

72
00:12:41.040 --> 00:12:50.070
Lukas Griessl: I look at controversy mainly in survey statistics around the question as to whether we can draw inferences from non probability sampling.

73
00:12:50.760 --> 00:12:57.660
Lukas Griessl: To give us a brief explanation of this it's about the design versus a model based approach to statistical inference.

74
00:12:58.650 --> 00:13:15.180
Lukas Griessl: The design approaches around the classical way of statistical sampling, meaning that possible respondents are randomly selected giving each possible respondent the same or a known chance of being included in the sample. For non probability samples, on the other hand, we don't have this.

75
00:13:16.530 --> 00:13:17.580
Lukas Griessl: And the

76
00:13:18.930 --> 00:13:39.420
Lukas Griessl: random aspect has largely being disregarded. Instead several adjustments are being made on the raw data in order to make it more representative. The way of adjusting for example, in order to weight certain responses from underrepresented groups higher.

77
00:13:40.650 --> 00:13:54.480
Lukas Griessl: So, what we can say is that while statistical theory provides justification for confidence and probability samples as a function of the survey design, inferences based on nonprobability samples are dependent on models for validity.

78
00:13:55.290 --> 00:14:14.820
Lukas Griessl: But there's on the other hand no certainty as to whether the model represents what it ought to represent. And this is where the controversy lies in. So I frame this controversy in my PhD around debate that took place in Germany, that is still taking place in Germany.

79
00:14:16.500 --> 00:14:27.210
Lukas Griessl: It's an actual debate between two or more than two of the rather traditional and established polling snstitutes in Germany and the new

80
00:14:28.050 --> 00:14:41.220
Lukas Griessl: startup that does this is like non probability online polls. And the debate started off in 2018 after a picture between two football players from the German national team,

81
00:14:41.700 --> 00:14:54.480
Lukas Griessl: which we can see here, Ilkay Gundogan and Mesut Ozil with the Turkish President Erdogan. And this picture that took place in London caused quite a huge public debate in Germany,

82
00:14:55.410 --> 00:15:17.730
Lukas Griessl: as people, and also certain media and newspapers, kind of question the loyality of these football players towards the German national team. It's a very big and interesting debate in its own right, but I just use it as a starting point to show how this controversy in Germany unfolded.

83
00:15:19.080 --> 00:15:22.230
Lukas Griessl: And within this controversy there were several

84
00:15:23.550 --> 00:15:24.870
Lukas Griessl: opinion poll

85
00:15:26.610 --> 00:15:27.990
Lukas Griessl: surveys being made.

86
00:15:29.340 --> 00:15:41.040
Lukas Griessl: And in order to kind of better understand the attitudes of the chairman population towards these football players, and that was, I mean that was very, very relevant because it was shortly before the 2018

87
00:15:42.210 --> 00:16:05.220
Lukas Griessl: World Cup. And the first one conducted by Forsa, which is a traditional polling institute that does random sampling, they do kind of this classical telephone interviews, and they found that 25% of the German population said that

88
00:16:08.310 --> 00:16:13.140
Lukas Griessl: these two football players should not be nominated for the upcoming World Cup. Whereas another

89
00:16:14.220 --> 00:16:30.300
Lukas Griessl: poll conducted by Civey, which is a, they do internet polls and non probability polls in which 80% said, to the question, should they after this picture continue to play with the

90
00:16:31.650 --> 00:16:47.520
Lukas Griessl: German national team said no. So we have two completely different numbers. We have a difference of more than 55%. And now the question is, who got it wrong, or did both get it wrong? And can we know this? 

91
00:16:48.960 --> 00:17:06.240
Lukas Griessl: And this marked the beginning of a very heated and charged debate between these two polling institutes in Germany and also found its way to the chairman press council. But yeah, I won't go too much into detail, I'd rather talk about what I'm interested in.

92
00:17:07.590 --> 00:17:09.210
Lukas Griessl: So in order to

93
00:17:10.710 --> 00:17:20.610
Lukas Griessl: understand this debate about inferences from non probability sampling, what I'm interested in is a question as to how statistical inferences are being justified,

94
00:17:21.030 --> 00:17:29.190
Lukas Griessl: and which epistemic cultures, which are culture of how knowlege is being made, are developed and maintained in the field of statistics.

95
00:17:30.030 --> 00:17:36.030
Lukas Griessl: And in order to do so, I draw on different areas such as the social sciences of quantification.

96
00:17:36.750 --> 00:17:46.860
Lukas Griessl: Which has one of its basic convictions that numbers are based on conventions, but nevertheless represent a reality. So this presents us with a kind of a paradox between

97
00:17:47.250 --> 00:18:05.130
Lukas Griessl: what we can call a constructivist and the realistic view on numbers. Which I do not want to solve, but which I want to shed more light on through interviews that I conduct with statistians, pollsters, and other stakeholders. So how I do this, I do interviews with

98
00:18:06.330 --> 00:18:09.870
Lukas Griessl: statisticians, mainly academic statisticians,

99
00:18:10.950 --> 00:18:22.440
Lukas Griessl: online, because that's the only way to do this right now, and I'm talking to them about questions such as what does statistics mean to you? What is statistics?

100
00:18:22.770 --> 00:18:28.170
Lukas Griessl: How do they deal with contested numbers? How do they justify their position within this controversy?

101
00:18:29.070 --> 00:18:42.840
Lukas Griessl: I also asked them more technical questions about which I'm mostly interested in the way how they present them to me. Do they present them to me in the way 'this is how we do it', or do they present to me in the way, 'there's different ways to do it'?

102
00:18:43.650 --> 00:18:58.260
Lukas Griessl: The results that I get are very interesting, and I would just like to point out some kind of response that I get to give you a feeling of what these people are telling me. So one chairman professor,

103
00:18:59.340 --> 00:19:05.310
Lukas Griessl: I just anonymized him for the moment, because it's a translation I didn't want to attribute it for now.

104
00:19:05.880 --> 00:19:09.810
Lukas Griessl: So this person told me "the fascinating thing is that they then find lunatics to believe that.

105
00:19:10.080 --> 00:19:18.240
Lukas Griessl: It's just like homeopathy, you have somebody who makes the belief somethin, and of course there are people who believe it, it's cheap, it's fast, it's sexy, but the problem is they don't explain it.

106
00:19:18.750 --> 00:19:28.950
Lukas Griessl: There's no reproductive proof. There's no predictive power. And the interesting thing is simply that there are people who believe them. It's really believing. There's no mathematics behind it, nor is there any empirically convincing evidence that it works.

107
00:19:30.240 --> 00:19:41.400
Lukas Griessl: And to maintain this despite the absence of evidence, you've been looking at homeopaths for, I don't know 120 years, and why should it go any faster in statistics? Machine learning is exactly the same at the moment.

108
00:19:41.880 --> 00:19:46.890
Lukas Griessl: People throw the data in the box, there is a predictive model and then they wonder why it doesn't work for the next set of data.

109
00:19:47.490 --> 00:19:54.600
Lukas Griessl: Machine learning people are people who have yet to discover inferential statistics. And this is exactly what you see with those who believe in the

110
00:19:55.170 --> 00:20:06.510
Lukas Griessl: numbers produced by Civey. If you want a number, I will tell you a number. If you shake me long enough. I'll tell you 42, which is surprisingly wide applicable. I found this quote particularly interesting

111
00:20:06.990 --> 00:20:18.060
Lukas Griessl: especially when we look at which kind of associations, this person draws on, saying that people believing certain numbers are lunatics, associating this approach with homeopathy.

112
00:20:18.780 --> 00:20:30.420
Lukas Griessl: Saying that there's an absence of evidence, and the last sentence was particularly funny making this reference to the Hitchhiker's Guide to the Galaxy.

113
00:20:31.740 --> 00:20:50.280
Lukas Griessl: "I tell you, 42", like just, I can just, just throw a number out and this is the kind of research that these people are doing. And what's interesting is that I get similar responses from people from the other camp within this controversy saying that

114
00:20:51.570 --> 00:21:02.730
Lukas Griessl: the probability camp has very big issues and has been completely ignoring them. And yeah I think, yeah, that's, I've come to the end.

115
00:21:04.140 --> 00:21:15.270
Lukas Griessl: Yeah, this is, I think, should give you a little insight into what I'm interested in, and what I'm doing. And yeah, I say thanks and I'm

116
00:21:17.550 --> 00:21:19.830
Lukas Griessl: And I'm looking forward to the discussion. Thank you.

117
00:21:21.780 --> 00:21:26.190
Hannah Pyman: Thanks Lukas! Another really great presentation and did lead on well Pablo's.

118
00:21:27.870 --> 00:21:44.340
Hannah Pyman: Okay. And so our third speaker today is Sabrina Rau, a senior research officer in the human rights, big data and technology project based at the law school. And Sabrina is going to talk about the responsibility of tech companies during COVID-19.

119
00:21:46.920 --> 00:21:49.500
Hannah Pyman: Perfect, thanks Sabrina, we can see your slides now.

120
00:21:53.700 --> 00:21:54.720
Hannah Pyman: Oh, you're muted.

121
00:22:02.490 --> 00:22:03.600
Sabrina Rau: There we go. Can you hear me now?

122
00:22:03.990 --> 00:22:05.670
Hannah Pyman: Yes, we can hear you. Great.

123
00:22:06.270 --> 00:22:07.410
Sabrina Rau: Thanks for having me.

124
00:22:08.520 --> 00:22:19.530
Sabrina Rau: Yeah, as you said, I'm a Senior Research officer in the human rights, big data, and technolocy project and I look at the opportunities and challenges that big data and data analytics and AI

125
00:22:20.010 --> 00:22:28.710
Sabrina Rau: have on human rights, both positive and negative. This paper that I'm presenting today was actually part of a series of papers that was published in July,

126
00:22:29.850 --> 00:22:43.200
Sabrina Rau: called the Essex dialogues on COVID-19 in which I presented on this topic. And the question mainly arose because of the role that tech companies are playing in our lives right now, even enabling the seminar right now through zoom.

127
00:22:43.620 --> 00:22:54.150
Sabrina Rau: Do tech companies bear special responsibilities if they are key enablers of certain human rights that would otherwise not possible in this time, such as the right to education right to work?

128
00:22:55.170 --> 00:22:56.640
Sabrina Rau: So, for

129
00:23:00.750 --> 00:23:07.290
Sabrina Rau: this presentation, I'm going to go through what tech companies have been doing in this time and what has changed.

130
00:23:08.010 --> 00:23:14.880
Sabrina Rau: Some examination of the special obligations that tech companies have, particularly under the UN guiding principles on business and human rights.

131
00:23:15.330 --> 00:23:29.430
Sabrina Rau: And three types of human rights impact assessments that tech companies should be conducting. There is a whole element of this that is focused on states which is going to be looked at in the next part of my research, but this presentation will be solely on the responsibility of companies.

132
00:23:30.990 --> 00:23:39.240
Sabrina Rau: So in terms of looking at changing consumer behavior, we've all noticed that things have changed since COVID has started, since we've been a lockdown,

133
00:23:40.410 --> 00:23:52.080
Sabrina Rau: as social isolation started. We've not been able to engage face to face, and everything has been happening online, on a personal level, but also in

134
00:23:52.710 --> 00:24:10.950
Sabrina Rau: other ways, such as the government employing contact tracing apps, sanitization things, making government services online, rather than face to face. These two diagrams show right at the beginning of COVID just from February, March, this huge spike in

135
00:24:12.240 --> 00:24:23.880
Sabrina Rau: uses of these online communication platforms. As well as on the right side, you see group call time in Italy, they were published by Facebook which increased dramatically.

136
00:24:24.600 --> 00:24:38.850
Sabrina Rau: Talking about statistics, it wasn't 42%, but web browsing generally increased over 70% and social media engagement over 61% over normal usage rates in the month of March.

137
00:24:40.890 --> 00:24:48.660
Sabrina Rau: So, these, these are dramatic changes that have happened. And actually, the zoom CEO Eric Yuan in a public ledger noted that

138
00:24:48.840 --> 00:25:01.830
Sabrina Rau: Zoom services were actually built for large enterprises, they were not built for individuals to socialize, for education, for kids. It was built for large enterprises and the purpose for which it was built,

139
00:25:02.190 --> 00:25:11.490
Sabrina Rau: and the types of risks that might be set up as a result, can change dramatically depending on users. And while they're used to getting spikes, such as Facebook

140
00:25:12.210 --> 00:25:27.330
Sabrina Rau: has published very openly saying that, for example, during new years they experience high peaks in usage, they're prepared for it while in COVID-19 they were unprepared for the types of usage that their services would be used for.

141
00:25:29.070 --> 00:25:37.890
Sabrina Rau: So, in one way, the technologies have been providing huge opportunities in the way that we're still able to continue our personal life, and also the way that

142
00:25:38.130 --> 00:25:50.490
Sabrina Rau: governments and businesses have been able to work in this time by providing communication work, education, cultural life that would otherwise not be possible. And additionally, for example, a lot of social media

143
00:25:51.960 --> 00:26:01.170
Sabrina Rau: platforms actually engaged with the WHO in raising awareness of health and safety guidelines and promoting government

144
00:26:01.740 --> 00:26:15.450
Sabrina Rau: content or other content that might be misinformation, disinformation. And additionally, you know, helping states in the Republic functions such as courts Dr surgery and schools, enabling these online platforms. Oh, and

145
00:26:16.920 --> 00:26:17.580
Sabrina Rau: Sorry.

146
00:26:18.600 --> 00:26:27.630
Sabrina Rau: Engaging drones for sanitation and contact tracing apps. But while all of these positive sides, there's also a lot of tremendous harm that can be done.

147
00:26:27.930 --> 00:26:36.030
Sabrina Rau: And that many times is associated with privacy and freedom of expression, but can range to a whole range of human rights, which can be seen on the right here.

148
00:26:38.280 --> 00:26:51.270
Sabrina Rau: The harms that can be, that are associated, are twofold. One are harms associated with use and deployment of these tech product and services and second is harms resulting from the lack of access to the services.

149
00:26:51.750 --> 00:26:58.770
Sabrina Rau: The second is quite significant as the digital divide is very important. It's not everyone that has access to technology or internet,

150
00:26:59.010 --> 00:27:08.100
Sabrina Rau: which might mean that a lot of the services that are available that are enabling certain rights are not accessible to everyone. And that was something that the state is actually responsible for.

151
00:27:08.670 --> 00:27:18.930
Sabrina Rau: And this is, this is something that I won't touch more upon, it's more about the businesses and how they do things what they do or what they don't do and how that might cause harm.

152
00:27:21.930 --> 00:27:31.650
Sabrina Rau: So when we examine the potential obligations of companies, we look at the UN guiding principles on business and human rights.

153
00:27:32.070 --> 00:27:41.430
Sabrina Rau: The UN working group on business and human rights, has explained that some businesses have a special role in the situation because of the nature of their product or services.

154
00:27:41.820 --> 00:27:52.830
Sabrina Rau: Referring specifically to providing life saving products, but under what category do you put products and services which act as the only means of enabling the enjoyment and fulfillment of human rights?

155
00:27:54.870 --> 00:28:02.490
Sabrina Rau: So under Article 14 you have the responsibility to respect human rights, 

156
00:28:03.030 --> 00:28:09.750
Sabrina Rau: which applies to all businesses, regardless of the sector, size, operational context, ownership, and structure.

157
00:28:10.200 --> 00:28:21.120
Sabrina Rau: However, the scale and complexity of the means in which the enterprise meet that responsibility may veary important to these factors and with the severity of the enterprise's adverse human rights impacts.

158
00:28:22.260 --> 00:28:32.940
Sabrina Rau: Now, that is a really important aspect of it because the UNGPs that respond to this respect applies to all businesses. That means human rights due diligence obligations

159
00:28:33.300 --> 00:28:36.240
Sabrina Rau: and the requirement to the human rights impact assessments.

160
00:28:37.110 --> 00:28:48.000
Sabrina Rau: However, whether you have 10 customers or 10 million customers, you still have responsibility to respect human rights.

161
00:28:48.420 --> 00:28:56.130
Sabrina Rau: However, the way, how you practice due diligence, how you conduct human rights impacts and the extent to which you

162
00:28:56.850 --> 00:29:06.780
Sabrina Rau: have participation and consultation as part of your process and how often you renew it very much depends on sector size, operation, and context ownership and struture.

163
00:29:07.590 --> 00:29:14.220
Sabrina Rau: While you might say that the context ownership and structure of these large tech companies hasn't changed,

164
00:29:14.850 --> 00:29:34.950
Sabrina Rau: the size and operational context very much has in terms of the number of users, changing the types of users, changing locations in which the services are all used are all changing, which means that, you know, in the first instance, it's important to assess where the impacts are.

165
00:29:37.290 --> 00:29:51.510
Sabrina Rau: In relation to this due diligence process, just to explain shortly, is the process whereby a company assessed as its human rights impact, acts upon the findings, transparently reports on those impacts,

166
00:29:52.380 --> 00:29:59.820
Sabrina Rau: and provides effective remedies to effective rights holders. So in this way, the impact assessment is the very first part of it.

167
00:30:00.840 --> 00:30:10.440
Sabrina Rau: And due to in this pandemic, we have three types, to make a bit more practical, there's three types of human rights impact assessment that companies should be assessing right now.

168
00:30:11.040 --> 00:30:19.020
Sabrina Rau: The first one is a baseline assessment, baseline human rights impact assessment, for new products, such as contact tracing apps. So before the deployment,

169
00:30:19.260 --> 00:30:25.260
Sabrina Rau: they should look at what the potential impacts could be, and to mitigate those impacts before they're roled out.

170
00:30:25.770 --> 00:30:36.330
Sabrina Rau: Such as contact tracing apps. Secondly, renewed human rights impact assessments for existing products, such as Zoom services. Does having a different audience

171
00:30:36.900 --> 00:30:47.460
Sabrina Rau: and being accessible in different locations, change the type of risk users are exposed to? Or are the risks to children who might have to use this platform be different than their typical enterprises, for example?

172
00:30:47.850 --> 00:30:53.940
Sabrina Rau: These are new types of considerations that need to be taken into account. And additionally humna rights impact

173
00:30:54.360 --> 00:31:00.720
Sabrina Rau: assessments for new and existing based business relationships because companies are not only legally responsible

174
00:31:01.110 --> 00:31:18.600
Sabrina Rau: for harm which they cause or contribute themselves, but also harm which they are linked to which means that if, within their supply chain or partnerships, a partner along their supply chain, there are harms to human rights a company can still be responsible for that harm.

175
00:31:20.550 --> 00:31:35.730
Sabrina Rau: And all of these things are basically the first. So if we would see more human rights impact assessment from the very beginning of COVID through the use of these to launch the increased use of these technologies, you might be reducing the types of harms that you see today.

176
00:31:36.900 --> 00:31:37.410
Sabrina Rau: Erm...

177
00:31:38.670 --> 00:31:39.150
Sabrina Rau: Sorry.

178
00:31:40.920 --> 00:31:46.680
Sabrina Rau: So lastly, in order to really harness the opportunities that technology presents

179
00:31:48.180 --> 00:31:58.980
Sabrina Rau: especially in this time and beyond, we need to build in adequate safeguards. And at the very beginning of that should be human rights impact assessments and embedded human rights based approaches.

180
00:32:00.630 --> 00:32:00.810
Thanks.

181
00:32:04.350 --> 00:32:05.190
Hannah Pyman: Thanks Sabrina.

182
00:32:06.450 --> 00:32:08.340
Hannah Pyman: That was really interesting. Thank you.

183
00:32:10.440 --> 00:32:11.040
Hannah Pyman: Okay.

184
00:32:13.320 --> 00:32:20.940
Hannah Pyman: Remember to put your questions in the Q&A. Our next speaker is Shahin Salarian, who's our final speaker of the day.

185
00:32:21.600 --> 00:32:35.400
Hannah Pyman: Shahin is a PhD candidate in the Computer Science and Electrical Engineering Department. Shain's presentation today focuses on full duplex communication systems. So I'm going to hand over to you now.

186
00:32:37.200 --> 00:32:38.100
Shahin Salarian: Thank you, Hannah.

187
00:32:40.050 --> 00:32:40.860
Shahin Salarian: Hello everyone, and good afternoon

188
00:32:43.980 --> 00:32:44.580
Shahin Salarian: Just going to share my screen.

189
00:33:01.140 --> 00:33:01.440
Shahin Salarian: Okay.

190
00:33:03.120 --> 00:33:22.470
Shahin Salarian: My name is Shahin. I'm a PhD researcher at RF and Microwave research lab in the computer science and electronic engineering, at the University of Essex. My PhD is about full-duplex communication system under the supervision of Professor Dariush Mirshekar-Syahkal.

191
00:33:25.830 --> 00:33:34.170
Shahin Salarian: What I'm presenting here is firstly I will go through some of the new applications and trends in communication systems,

192
00:33:34.620 --> 00:33:52.740
Shahin Salarian: and the need for full duplex communication systems. The advantages of this full duplex communication system, and the main challenge for implementing this. This is self interference and the combination of the signal the receive signal in the receiver.

193
00:33:55.140 --> 00:34:01.140
Shahin Salarian: And also different software's and method for mitigating this self interference signal.

194
00:34:02.400 --> 00:34:04.680
Shahin Salarian: Among them, I will go to the

195
00:34:05.940 --> 00:34:17.940
Shahin Salarian: passive software, active, and digital and analog, and I will introduce some of our challenges in our research. And the final part is

196
00:34:18.630 --> 00:34:37.770
Shahin Salarian: the recent publication that they have a huge conference. This was aboutdual-beam orthogonal circular polarized antenna. And then I also consider diverse range of audience for this presentation. So I will start from some basics and then go to some more technical products.

197
00:34:39.630 --> 00:34:48.600
Shahin Salarian: As mentioned in the previous presentation, the data rates for communication systems are booming somehow, such as a smart phone,

198
00:34:49.200 --> 00:35:05.280
Shahin Salarian: the cities, smart homes, IoT devices and different applications such as rooms, etc. And also the new generation of mobile communication: 5G, 6G. All of them meet low latency and high data rates.

199
00:35:06.000 --> 00:35:17.940
Shahin Salarian: So we have the explosive increase in demand for the data traffic and considering the limited frequency spectrum that we have, we need some sort of

200
00:35:18.630 --> 00:35:36.750
Shahin Salarian: optimal efficient use of this spectrum. So the previous or traditional communication system that they use are half-duplex. Mainly time domain or frequency domain, which means that they they're divided in time or in frequency.

201
00:35:37.980 --> 00:35:53.160
Shahin Salarian: So full duplex is a kind of promising technology to achieve higher data rates, and it's normally in literature with different norms, such as the STAR which is the standard for simultaneous transmit and receive,

202
00:35:54.390 --> 00:35:58.440
Shahin Salarian: or in-band full duplex.

203
00:35:59.460 --> 00:36:06.690
Shahin Salarian: And it has different advantages. As mentioned, one of them is doubling the data, linking capacity, reducing the feedback,

204
00:36:08.100 --> 00:36:21.060
Shahin Salarian: and system Capacity Enhancement. As mentioned, the main challenge is self interference, which is the power leaks from their own transceiver to the own receiver.

205
00:36:21.510 --> 00:36:32.040
Shahin Salarian: So there should be some ways to reduce this effect. For some long time it was believed that this full-duplex communication system is

206
00:36:32.640 --> 00:36:44.520
Shahin Salarian: impossible to implement, but with some and progress and novelty it's possible somehow just to reduce this interference and implement a complex full-duplex system.

207
00:36:47.220 --> 00:36:55.890
Shahin Salarian: As we can see in this slide, there are different northern communication system, each of them consists of a transceiver and receiver.

208
00:36:56.700 --> 00:37:09.900
Shahin Salarian: And we have the same for all of the nodes and some system behind this. So from these transceiver and receiver, we should know that they have a kind of antenna which are radiating in some direction.

209
00:37:11.130 --> 00:37:29.820
Shahin Salarian: And depending on the technical specification, they might have a stronger beam in some direction and we can meet in some direction. So you can see three main things here. One of them is the desired signal from the other node transceiver which is shown in green arrow.

210
00:37:30.930 --> 00:37:40.230
Shahin Salarian: And it's somehow desired. And the other one is this red one, which is the leakage of the power from the own transceiver to the receiver which is not

211
00:37:41.100 --> 00:37:52.860
Shahin Salarian: wanted. It's somehow self interference that should be mitigated. And the other one is the scattering from the scatterer, such as buildings,  trees or everything which is showing here.

212
00:37:55.470 --> 00:38:06.780
Shahin Salarian: And depending on the signal that we are targeting and the stage that we are suppressing the signal, there are different classifications and categories for the methods.

213
00:38:07.740 --> 00:38:17.220
Shahin Salarian: For example, there is one category which is called passive for this passive techniques where it's mainly in the RF and the antenna part, the electromagnetic part.

214
00:38:17.700 --> 00:38:35.370
Shahin Salarian: And it's implemented in different ways, such as two antennas, or using some other additional components such as a circulator or duplexer. Or it can be also implemented in one antenna which is somehow integrated in its dual port.

215
00:38:39.510 --> 00:38:54.870
Shahin Salarian: The other technique is active cancellation, which is a subtraction of the transmitter signal from the receive signal. So because it can reduce the effects of the weakened direct leakage for the transceiver from the receiver.

216
00:38:56.520 --> 00:39:05.550
Shahin Salarian: Our main focus has been on the antenna design, the passive form so far, the isolation and construction that we are using such as SIW

217
00:39:10.380 --> 00:39:29.760
Shahin Salarian: and the surface waves mitigation. Now, I also want to focus on dual port feed, polarization diversity, on different layers, different surfaces, and all of these methods somehow add some sort of complexity, but there should be a kind of trade off between these.

218
00:39:36.090 --> 00:39:41.640
Shahin Salarian: And this is the result from our latest publication in EUCAP conference, which shows

219
00:39:43.170 --> 00:39:45.390
Shahin Salarian: dual-beam orthogonal circular polarized antenna.

220
00:39:47.100 --> 00:39:59.940
Shahin Salarian: As you can see in the structure, we use a monopole antenna, with some SIW structure, and it's on substrate with epsilon 2.2.

221
00:40:02.250 --> 00:40:12.270
Shahin Salarian: And you can see it has two different polarizations, and both of them are central and the axial ratio is zero three which is required for circulartity. That was all, thank you.

222
00:40:21.990 --> 00:40:22.590
Hannah Pyman: Thank you Shahin, that was great.

223
00:40:26.880 --> 00:40:38.340
Hannah Pyman: Okay, so I think safe to say we have definitely had a broad range of topics today, and covered a lot of really interesting content. So thank you to all four of you, that was great. And I'm going to hand over to Kez now,

224
00:40:38.430 --> 00:40:40.170
and see if we have any questions.

225
00:40:41.190 --> 00:40:54.090
Keziah Gibbs: Yeah, I've had a few questions come in. So I'll start with one I've got one for Pablo. So do you think that polls are getting moree difficult to run accurately because of the increased information available to voters these days?

226
00:40:57.870 --> 00:40:58.740
Okay, this

227
00:41:00.000 --> 00:41:00.690
is a very good question.

228
00:41:04.380 --> 00:41:19.530
Pablo Cabrera Álvarez: If the fact that voters have more information right now makes them more volatile, so then it's more likely that they change from one party to another, and this change happens really quick, so polls

229
00:41:19.890 --> 00:41:30.300
Pablo Cabrera Álvarez: cannot capture that moment, in that case, that could be an option. The point is, if we see our poll as a picture,

230
00:41:31.800 --> 00:41:42.210
Pablo Cabrera Álvarez: it should reflect the full complexity in a given moment. So if I give a moment, people are receiving a lot of information that they have

231
00:41:43.410 --> 00:41:52.110
Pablo Cabrera Álvarez: a clear voting intention. I think that should not be a problem. I think that the fact that there is much more information circulating

232
00:41:52.440 --> 00:42:15.570
Pablo Cabrera Álvarez: has to do with public opinion dynamics. So the point is important, of course, we have the public opinion dynamics, and we also have this the method, right. So I don't see that the methods should not take the picture of that happening. But of course, of the system that public opinion

233
00:42:16.710 --> 00:42:20.910
Pablo Cabrera Álvarez: trends are getting more complex, at some point, it could affect. So yes, and no!

234
00:42:23.700 --> 00:42:41.610
Keziah Gibbs: Yes. Okay. Thank you. Yeah, that's interesting. I'll just scroll through. I've got another question here, so for Lukas. Do you think that the media should be more transparent with the samplinf behind the statistics, or do you think it's too complicated for people to understand?

235
00:42:44.070 --> 00:42:52.620
Lukas Griessl: That's a very good point because especially in this controversy in Germany I was referring to, this question became, the,

236
00:42:53.340 --> 00:43:07.740
Lukas Griessl: like this, the case was in front at the chairman press Council, where some of these polling institutes were filing a claim against Civey, not against Civey, against

237
00:43:08.820 --> 00:43:20.160
Lukas Griessl: a German newspaper saying that they were using the survey without questioning the validity of the survey.

238
00:43:20.730 --> 00:43:38.430
Lukas Griessl: And the chairman press Council, in the end, decided in favor of this online newspaper, saying that it's not part of journalistic due diligence to delve into the technical and statistical dimension of the surveys.

239
00:43:39.570 --> 00:43:52.890
Lukas Griessl: So the this point has become a matter of public debate in Germany. And to me, I think, yes, they should be more transparent about this and I don't really think that

240
00:43:53.580 --> 00:44:04.620
Lukas Griessl: it's too complex. I think you can, I think you can make the like, you can make the uncertainty and certain surveys, more transparent, you can

241
00:44:06.210 --> 00:44:17.520
Lukas Griessl: make, you can maybe point towards the fact that there is a certain sample size, that there are certain methods

242
00:44:18.240 --> 00:44:34.320
Lukas Griessl: used and so I think this does not need to be like a very big information, or next to the presented survey. But yes, I think that journalists should and could easily be more transparent about this.

243
00:44:35.940 --> 00:44:36.960
Keziah Gibbs: Okay, thank you.

244
00:44:38.190 --> 00:44:42.930
Keziah Gibbs: And there's another question here. So this one's for Sabrina.

245
00:44:43.560 --> 00:45:00.210
Keziah Gibbs: If the tech companies continue to be used for teaching and medical services, for example, do you think that governments have a human rights responsibility to ensure everyone can access some form of technology to allow us to have these services?

246
00:45:03.210 --> 00:45:04.440
Sabrina Rau: Very good question.

247
00:45:05.520 --> 00:45:06.930
Sabrina Rau: Yes, absolutely.

248
00:45:09.990 --> 00:45:18.660
Sabrina Rau: Governments had a responsibility for this even before COVID and will continue to do so after it's very important to understand the world stage as well.

249
00:45:19.080 --> 00:45:29.820
Sabrina Rau: It's important to understand that under international law states are primary duty bearers, which means that they're responsible for protecting human rights, which means that if somebody doesn't have

250
00:45:30.360 --> 00:45:42.720
Sabrina Rau: accessible healthcare or, you know, education, then the State is responsible for making sure that there's education. The company isn't however responsible for

251
00:45:43.080 --> 00:45:56.280
Sabrina Rau: making sure that every single person has access to education. They're only responsible for ensuring that the service and the products that they provide don't cause harm under the do no harm principle, basically.

252
00:45:57.450 --> 00:46:11.190
Sabrina Rau: So I think we have to think about the types of safeguards that we can build in so we can really, you know, take the opportunities that technology is allowing us but ensuring that the states

253
00:46:11.730 --> 00:46:16.740
Sabrina Rau: protect human rights and protects from third party harm, for which businesses are part of

254
00:46:17.280 --> 00:46:32.970
Sabrina Rau: and to really mitigate those harms and also remedy those harms. That's often forgotten rights. So there's some really complex issues on how that works, and particularly in relation to public private partnerships, it's often the services that even governments

255
00:46:34.050 --> 00:46:40.530
Sabrina Rau: are was supplying, in terms of the digital government services are usually procured

256
00:46:40.890 --> 00:46:45.180
Sabrina Rau: by third parties, by companies externally, and they're creatued by companies.

257
00:46:45.390 --> 00:46:59.970
Sabrina Rau: So to what extent the government is responsible for the potential harms that occur as well must be considered to ensure that people who do experience harms as result of using the technology, and may not have access to those technologies, have effective rights.

258
00:47:02.010 --> 00:47:18.630
Keziah Gibbs: Great! Yeah. Very interesting. Thank you. And there's just another question here. So for Pablo. So if polls fail so much, then what is your opinion on the value of them. And do you think that polls themselves actually influence voters?

259
00:47:22.770 --> 00:47:25.080
Keziah Gibbs: You're muted. Sorry. Okay.

260
00:47:25.140 --> 00:47:27.960
Pablo Cabrera Álvarez: Very good question. Thank you. 

261
00:47:29.400 --> 00:47:35.640
Pablo Cabrera Álvarez: The first one I mean polls are useful to take a picture and to estimate

262
00:47:36.180 --> 00:47:47.220
Pablo Cabrera Álvarez: vote issues in a context, but we know, we know that there is a margin of error and this error is not just the fact that we are something from our population.

263
00:47:47.580 --> 00:48:01.080
Pablo Cabrera Álvarez: All the mathematics that LuKas has been has been explaining to us is much more. So it's what time do we do this poll, where you put the voting, is the essential question within the question.

264
00:48:02.040 --> 00:48:17.820
Pablo Cabrera Álvarez: All the context around and in it. So we know that a voting estimate has some error and that error is larger, actually, than the sampling error. So in a very narrow race, this is going to be an issue to estimate the right

265
00:48:19.020 --> 00:48:31.740
Pablo Cabrera Álvarez: to estimate or to predict the winner. So in that case, probably we are asking polls to do more that what they are really am prepared to do right.

266
00:48:32.100 --> 00:48:44.160
Pablo Cabrera Álvarez: So this is the first question. The second is about polling affecting. This is one of the big topics in public opinion, right. So people trying to use polls, so they can affect voters' behaviors.

267
00:48:45.150 --> 00:48:54.840
Pablo Cabrera Álvarez: The reality, I mean, first of all, they're methodological design to prove that hypothesis is very, very complex or even impossible.

268
00:48:56.970 --> 00:49:04.020
Pablo Cabrera Álvarez: Some people, of course, have been trying to show if that has an effect, and most of the analysis,

269
00:49:05.460 --> 00:49:20.610
Pablo Cabrera Álvarez: has got to the point that they  say that this is not likely to happen, right. So it needs to be people that are really into politics that poll information can change their behavior.

270
00:49:21.300 --> 00:49:40.500
Pablo Cabrera Álvarez: It happens, yes. It can switch an election? Probably not. Probably not. But, I mean, it's one of the hot topics of people say, trying to estimate what was the effect of this. Yeah, thanks.

271
00:49:41.520 --> 00:49:50.730
Keziah Gibbs: Great, thank you. And so that's all the questions that we've had so far. So thank you all for answering those. And so I'll hand back to Hannah.

272
00:49:51.540 --> 00:50:03.420
Hannah Pyman: Yeah, thank you. Thanks, Kez. Thanks for the questions, that was really interesting. And unless anyone else has any questions, unless any of the presenters have questions for each other?

273
00:50:04.770 --> 00:50:06.360
Lukas Griessl: Can I ask a question to Pablo.

274
00:50:10.320 --> 00:50:30.600
Lukas Griessl: So I am wondering whether we can say that in the 2016 election in the US, if the quality of the question in a survey was different as back in 2016 I would say it was a kind of a low cost decision whereas now it's a high cost decision because

275
00:50:31.890 --> 00:50:35.490
Lukas Griessl: when we look at the whole political turmoil, one can say that

276
00:50:36.630 --> 00:50:44.730
Lukas Griessl: it has become clear that the election will have a high impact on the lives of certain people, we are looking at COVID, at Black Lives Matter, so the

277
00:50:45.300 --> 00:51:03.870
Lukas Griessl: Election Results became a matter of life and death for certain people. So my question is, if the response of a survey back into, like the response people gave to the to the survey in 2016, had less truth and less reality to a response that was given in 2020?

278
00:51:07.890 --> 00:51:08.250
Lukas Griessl: Well,

279
00:51:08.610 --> 00:51:12.690
Pablo Cabrera Álvarez: That's hard to say in terms of evidence.

280
00:51:14.160 --> 00:51:30.120
Pablo Cabrera Álvarez: I think what you're you're trying to say, right, is that there are topics in this election are much more salient and then a lot of things have happened in this last four years in the US polarization has increased like crazy. We have a pandemic.

281
00:51:30.630 --> 00:51:42.780
Pablo Cabrera Álvarez: We have many issues that make voters be more aware of what they are doing in the election. And this was not the case in 2016. Is that the case Lukas, would you try to,

282
00:51:44.250 --> 00:51:47.520
Pablo Cabrera Álvarez: what you try to say? Okay.

283
00:51:49.050 --> 00:52:03.150
Pablo Cabrera Álvarez: Yes, I mean the context is very different. It doesn't seem so but the context, the electoral context is very different from what we have four years ago and what we had this time. I don't know if we can

284
00:52:03.900 --> 00:52:12.870
Pablo Cabrera Álvarez: go that far to say that what the voters say to the polls this time was more true because they were more engaged with the election somehow,

285
00:52:14.970 --> 00:52:29.580
Pablo Cabrera Álvarez: but certainly the salience of those topics affected how they responded to the polls. Indeed, we know that, for example, if respondents were asked about COVID issues that

286
00:52:30.090 --> 00:52:41.040
Pablo Cabrera Álvarez: tended to activate democrat voters. So the result of the poll would be slightly different to the question where the COVID question was

287
00:52:41.850 --> 00:52:51.810
Pablo Cabrera Álvarez: removed from the questioner. So I think if the topic on the content of the election has an effect on what has happened with the polls, I think, is

288
00:52:52.230 --> 00:52:59.040
Pablo Cabrera Álvarez: true, but we don't know in exactly what direction it has affected.

289
00:52:59.700 --> 00:53:14.820
Pablo Cabrera Álvarez: Because for example this election, much more salient. That's true. I mean, very hot topics, but at the same time, for example, their response rate to the polls are still very low. So I didn't make that much difference. All right.

290
00:53:17.160 --> 00:53:17.610
Lukas Griessl: Thank you.

291
00:53:21.420 --> 00:53:26.220
Hannah Pyman: Thanks. Okay. So unless there are any other questions at all?

292
00:53:29.220 --> 00:53:39.870
Pablo Cabrera Álvarez: Well, I do have a question. I mean, this is like cross questioning, because I think that Lukas and I shared quite a lot of the topic. Right, so

293
00:53:40.920 --> 00:53:53.670
Pablo Cabrera Álvarez: looking at the example, right, the example that you post like the trigger of the research, these questions about those, like Ozil, if he should join the German national team or not.

294
00:53:54.090 --> 00:54:09.090
Pablo Cabrera Álvarez: These two questions I want to ask you whether the debate focused on the sample composition. So, Civey modeling non-probability sample versus probabaility sample, or someone realized that maybe all the issues

295
00:54:09.780 --> 00:54:14.940
Pablo Cabrera Álvarez: should be affecting that difference of, for example, the wording of the question there

296
00:54:16.140 --> 00:54:25.290
Pablo Cabrera Álvarez: is, I don't know if the first question is, 'let's talk about the picture' or the position of the question within the question?

297
00:54:27.570 --> 00:54:35.580
Lukas Griessl: So both of these questions were formulated slightly differently, even though both surveys were published pretty much at the same time.

298
00:54:36.180 --> 00:54:59.340
Lukas Griessl: And I would say that there was a, on the one side rather a public debate around this, and then there was the inner statistical debate that had not so much to do with this picture, per se, but the divergent results of these two surveys

299
00:55:00.390 --> 00:55:00.900
Lukas Griessl: kind of

300
00:55:02.580 --> 00:55:04.500
Lukas Griessl: started off this

301
00:55:07.650 --> 00:55:15.090
Lukas Griessl: debate that was already there. 

302
00:55:16.440 --> 00:55:29.190
Lukas Griessl: So, this picture triggered a public debate around these football players. And this became a really huge huge thing actually.

303
00:55:29.700 --> 00:55:48.150
Lukas Griessl: And the kind of, the rather statistic or the rather technical debate, it was kind of triggered by this, but it was not so much, I would say, about the actual survey that was being conducted, it became a rather a broader and bigger discussion about something in general.

304
00:55:52.620 --> 00:55:54.510
Lukas Griessl: I think this is what what you meant.

305
00:55:56.220 --> 00:55:56.610
Yes.

306
00:55:58.770 --> 00:56:00.240
Hannah Pyman: Great. Thank you, guys.

307
00:56:01.680 --> 00:56:13.770
Hannah Pyman: Okay, I think we're just about out of time. And if anyone does think of any other questions, then you can always email me using newcomers@essex.ac.uk and I will pass on the questions to our presenters.

308
00:56:14.700 --> 00:56:23.040
Hannah Pyman: But otherwise, just to say thank you so much to all of our presenters today and it's safe to say we have had another very interesting newcomers presents session.

309
00:56:23.700 --> 00:56:34.740
Hannah Pyman: And don't forget that we have three more great topics that are going to be spoken about on Wednesday at 11am so if you haven't already, please do book your place at that session as well.

310
00:56:35.640 --> 00:56:46.590
Hannah Pyman: Thanks everyone for coming along and contributing some really great questions. That was a really good discussion there at the end. And so, thank you everyone for getting involved. I'm going to stop the recording now.