1 00:00:08,230 --> 00:00:15,760 [Auto-generated transcript. Edits may have been applied for clarity.] So I'm very happy to and pleased to have Ruth Ann with us today for this third annual, uh, Digital Enlightenment Studies lecture. 2 00:00:15,840 --> 00:00:19,419 We're really taking a broad view of the enlightenment, uh, in that sense. 3 00:00:19,420 --> 00:00:20,559 But that's, uh, that's fine. 4 00:00:20,560 --> 00:00:29,550 Ruth is a professor of literary history and digital humanities at the Queen Mary, uh, and probably best known, at least in some circles, for her, uh, 5 00:00:29,620 --> 00:00:35,290 directorship of the living with machines project, which was the UK Ukri, uh, 6 00:00:35,290 --> 00:00:40,480 big deal for five years between the British Library and the Turing Institute, which did all sorts of wonderful things. 7 00:00:40,490 --> 00:00:46,720 I'm sure we'll hear a bit about it. Uh, she's also worked a lot on the Tudor papers with her partner, Sebastian. 8 00:00:46,990 --> 00:00:48,760 Wonderful book. That was out a few years ago. 9 00:00:49,090 --> 00:00:58,329 Uh, Tudor Networks of Power and, uh, and a wonderful little book, uh, that I use a lot in teaching, uh, called the Network Turn with, uh, 10 00:00:58,330 --> 00:01:02,709 Scott Weinger and Nicole Coleman and Sebastian, which I recommend highly, uh, 11 00:01:02,710 --> 00:01:08,260 and has done great work for, uh, proselytising network analysis for the humanities. 12 00:01:08,470 --> 00:01:12,280 And it's wonderful. So we're very happy to have Ruth here today. Uh, please join me in welcoming. 13 00:01:18,580 --> 00:01:20,860 And thank you so much to Glen for inviting me. 14 00:01:20,890 --> 00:01:28,270 Um, today I want to share some work that I'm kind of just completing, although I've been tinkering with that for about ten years, embarrassingly. 15 00:01:28,420 --> 00:01:35,740 Um, so it's a research problem that sits at the intersection of archival studies and early modern history and digital humanities. 16 00:01:36,010 --> 00:01:37,360 The problem of aliases. 17 00:01:37,840 --> 00:01:45,100 Specifically, I'll be thinking about how we might identify people hidden behind aliases in large scale correspondence collections, 18 00:01:45,460 --> 00:01:49,120 and what happens when we bring computational methods to bear on that challenge. 19 00:01:49,900 --> 00:01:52,750 So as you see, the material I work on is a 16th century. 20 00:01:52,750 --> 00:01:59,530 So not very enlightenment, but the methods I'm going to show you today kind of are applicable to correspondence collections of any period. 21 00:02:01,980 --> 00:02:06,270 So my argument is that aliases are a natural consequence of communication. 22 00:02:06,270 --> 00:02:13,260 So as new communication technologies emerge, so do techniques to evade surveillance. 23 00:02:13,770 --> 00:02:19,020 Today we have things like encrypted messaging apps, anonymous browsers and burner phones. 24 00:02:19,020 --> 00:02:24,870 But that dynamic is not a new one. So in the early modern period, the technologies were paper and ink, 25 00:02:25,470 --> 00:02:30,090 and there were a variety of technologies employed to evade scrutiny, including the alias. 26 00:02:30,990 --> 00:02:36,600 Now, while some browser technologies have been built for whistleblowers, journalists and just normal people who want to, 27 00:02:36,720 --> 00:02:42,360 um, avoid being, uh, surveilled, they're also, of course, used by criminals. 28 00:02:42,780 --> 00:02:49,439 And in much the same way the evidence left in archives shows us the aliases we use, but for a variety of reasons. 29 00:02:49,440 --> 00:02:53,579 But abuse by state agents and spies, they were used by conspirators, 30 00:02:53,580 --> 00:02:58,140 but also authors and theologians expressing opinions that might have been controversial. 31 00:03:00,810 --> 00:03:05,580 Now intersection of letters was actually a fairly widespread practice in the early modern period. 32 00:03:05,580 --> 00:03:11,400 So, um, in a study I did with Rachel Madeira and Sebastian on at MIT, my longtime collaborator and partner, 33 00:03:11,700 --> 00:03:21,480 um, we showed that in excess of 5% of papers in, in the Elizabethan set papers foreign were intercept. 34 00:03:21,510 --> 00:03:28,440 So we developed a new method to show this. Um, and that's a lot, uh, and because this was a very real threat for people sending letters, 35 00:03:28,740 --> 00:03:32,430 various technologies evolved to protect letters from prying eyes. 36 00:03:33,120 --> 00:03:38,310 The most famous histories of conspiracy in this period period are littered with those technologies and strategies. 37 00:03:38,610 --> 00:03:41,280 We might think of the Babington Plot that you probably have heard of, 38 00:03:41,280 --> 00:03:45,300 which was the the means by which Mary, Queen of Scots was ultimately brought down. 39 00:03:45,420 --> 00:03:52,560 This was um, employed encrypted letters. Um, it said, between aliases often hidden in beer barrels. 40 00:03:52,710 --> 00:03:58,260 And these were intercepted, copied, uh, by the double agent, uh, Gilbert Gifford, uh, 41 00:03:58,260 --> 00:04:02,820 who was also known by various names and deciphered by Walsingham as code breaker Thomas Phillips. 42 00:04:02,830 --> 00:04:04,260 We'll hear a bit more about later. 43 00:04:05,580 --> 00:04:10,740 Other strategies employed in this period were things like a letter lock, in, which you might think of as a tamper evident technology. 44 00:04:12,240 --> 00:04:16,680 But beyond these dramatic episodes, aliases were a widespread use. 45 00:04:17,430 --> 00:04:21,030 And this because of that, that we've come to this, um, this method. 46 00:04:23,040 --> 00:04:27,550 So if historians and users present three kinds of challenge. 47 00:04:27,600 --> 00:04:30,510 So this first, this aliases that actually haven't been solved. 48 00:04:31,020 --> 00:04:38,430 The second is aliases that have been identified in scholarship, but for which the information has not been folded back into finding Aids. 49 00:04:38,880 --> 00:04:41,970 Um, or it's not easily discoverable, uh, for other scholars. 50 00:04:42,180 --> 00:04:47,760 The third is where the identity of the aliases are disputed or theories are difficult to verify. 51 00:04:49,970 --> 00:04:54,110 And so what I want to present today is, um, a computational heuristic, uh, 52 00:04:54,110 --> 00:04:57,830 designed to address these challenges, which I co-developed with, with Sebastian. 53 00:04:58,190 --> 00:05:01,190 Um, and the method doesn't solve aliases automatically. 54 00:05:01,190 --> 00:05:06,490 I'm not really in the method of kind of automatic methods. Instead, it generates candidate matches. 55 00:05:06,740 --> 00:05:12,290 Uh, that helps guide the story in allowing them to kind of test and explore and refine hypotheses. 56 00:05:14,480 --> 00:05:18,920 We developed this method on the data behind, uh, to the networks of power projects. 57 00:05:18,920 --> 00:05:22,999 So this book, um, applies quantitative network analysis, 58 00:05:23,000 --> 00:05:29,090 but also other computational methods to the study of, uh, the letters in the Tudor State Papers. 59 00:05:29,090 --> 00:05:33,020 Um, so these started out as the working papers of the secretaries of state. 60 00:05:33,440 --> 00:05:37,020 Um, they've been catalogued and they were digitised as state papers online. 61 00:05:37,080 --> 00:05:43,010 And because, uh, if we if we, um, subscribe to these kinds of resources, we can actually ask the underlying data. 62 00:05:43,010 --> 00:05:46,339 So we ask the underlying data, uh, the XML and the scans. 63 00:05:46,340 --> 00:05:56,180 And from that we extracted 182,747 letters, uh, which were from and to over 20,000 people spread across Europe and beyond. 64 00:05:56,690 --> 00:06:03,409 But that kind of quick overview of the data belies what was, uh, a long and arduous data cleaning process. 65 00:06:03,410 --> 00:06:09,559 Um, 18 months probably. Um, we first needed to clean people's names, needs to disambiguate them and duplicate them. 66 00:06:09,560 --> 00:06:13,760 We needed to do the same for place names. We also enriched the data. 67 00:06:13,760 --> 00:06:16,880 We sort of geocode the places. We added linked data to people. 68 00:06:17,750 --> 00:06:23,810 Um, so that the book we have an interactive visualisation now, aliases were part of that data cleaning process. 69 00:06:24,020 --> 00:06:28,850 And we might think of aliases as a kind of advanced deduplication problem. 70 00:06:29,210 --> 00:06:36,830 Uh, because people are known by multiple names designed to avoid being collapsed onto one identity. 71 00:06:37,250 --> 00:06:42,650 And we can see exactly the kind of problem that that poses for us, uh, with the visualisation in this next figure. 72 00:06:46,660 --> 00:06:52,870 Um, so this, um, is a network visualisation of the different aliases that have been attributed, 73 00:06:53,380 --> 00:06:57,670 uh, with different levels of, uh, controversy to a man called William Sterile. 74 00:06:57,670 --> 00:07:02,020 And I'm going to come back to the tricky, this tricky character, um, later in the talk. 75 00:07:02,350 --> 00:07:08,860 But what you need to know now is that all of the yellow nodes have been suggested as aliases of sterile, 76 00:07:09,220 --> 00:07:12,640 and if they were collapsed, the network would look like this. 77 00:07:14,140 --> 00:07:25,390 Uh, but they have not. Um, so, uh, this shows you exactly what aliases do they have the function of breaking connections and network, 78 00:07:25,390 --> 00:07:31,870 obscuring identities and creating dead ends not only for anyone surveilling the letter writers, 79 00:07:31,870 --> 00:07:36,550 but also for any historians later working with those archives. 80 00:07:36,820 --> 00:07:44,469 Um, and, hey, uh, the the sort of, um, leftmost, uh, two little they're called dyads where you've got an, 81 00:07:44,470 --> 00:07:48,910 um, an an entity with two, two nodes connected with one edge. 82 00:07:49,330 --> 00:07:57,430 They're the kind of perfect case. What's happening here is the alias is used so strictly, um, that they are directional as well. 83 00:07:57,430 --> 00:08:03,340 That person A, writing to person B is probably using a different alias for person than person B, writing to person I. 84 00:08:04,870 --> 00:08:13,300 If you use a system like this perfectly, it's very, very, very difficult to collapse it back into the kind of network I just showed you. 85 00:08:13,660 --> 00:08:16,540 But fortunately, people are lazy. People make mistakes. 86 00:08:16,900 --> 00:08:25,270 Um, and so once we can start piecing together little bits of evidence to start collapsing the network together, this is where our method comes in. 87 00:08:27,470 --> 00:08:33,470 So that network visualisation of sterile aliases and neighbours showed us the problem. 88 00:08:34,010 --> 00:08:38,810 But networks also offer a solution through the method of quantitative network analysis. 89 00:08:39,230 --> 00:08:47,690 And so the idea, the idea behind a heuristic is if x communicates x, y and x communicates with z, 90 00:08:48,110 --> 00:08:52,520 but y and z do not communicate, it is possible they are the same person. 91 00:08:53,000 --> 00:08:58,670 And this builds on a theory by um, grant of assets like his stand out publication. 92 00:08:58,670 --> 00:09:02,660 It's like massively cited uh, courts for the strength of weak ties. 93 00:09:02,780 --> 00:09:05,750 But in that publication, he has something called the forbidden triad. 94 00:09:06,320 --> 00:09:16,850 Um, and he says, um, that, uh, if X and Y correspond and, um, and sorry, that's a little bit if if X has strong social ties with Y and Z, 95 00:09:17,030 --> 00:09:21,290 there will also be some kind of social tie, either weak or strong, between y and Z. 96 00:09:21,620 --> 00:09:27,860 And he describes social networks where there was an absence of the y z type as the forbidden triad, 97 00:09:28,040 --> 00:09:32,390 precisely because it deviates from observed patterns in social network data. 98 00:09:32,750 --> 00:09:39,560 So we're kind of riffing on that theory. So for those of you who might be familiar with the state papers or other kinds of state collections, 99 00:09:39,950 --> 00:09:44,059 you might be thinking of reasons why there wouldn't be connections between Y 100 00:09:44,060 --> 00:09:48,260 and Z precisely because they started out as the Secretary of State's papers. 101 00:09:48,260 --> 00:09:57,740 Do you not think of Secretary of State as the X note? Um, that's true, but the state papers also folded in all sorts of other administrative packets. 102 00:09:58,070 --> 00:10:01,820 Um, uh, and it, uh, it did lots of interception. 103 00:10:01,820 --> 00:10:08,600 It sees lots of people's private ah, correspondence meaning that very frequently we find that triads are, in fact, closed. 104 00:10:08,750 --> 00:10:12,380 So open triads are a clue to us. Um, and that's why we begin. 105 00:10:12,620 --> 00:10:23,420 So this is how the method works. So the method, very simply put, systematically searches for open triads amongst the 20,000 plus notes. 106 00:10:24,140 --> 00:10:31,010 Now we developed two variants of this method because we have kind of different alias problems that we know about or might be looking for. 107 00:10:31,310 --> 00:10:38,330 So the first one begins the designated identity. That's a suspected alias, something like an initial or a or a symbol. 108 00:10:39,980 --> 00:10:46,580 Um. Uh, and this returns as a single list of people that fulfil the criteria, that is, 109 00:10:46,580 --> 00:10:50,570 pairs of correspondence that are unconnected but which share neighbours. 110 00:10:52,280 --> 00:10:59,179 We also developed the second variant, which is a kind of more exploratory research tool, which might begin with a given, 111 00:10:59,180 --> 00:11:04,940 uh, person in the network who's known to, uh, correspond with people who might have used aliases. 112 00:11:05,180 --> 00:11:11,780 Um, what we do is we, we scour their entire neighbourhood for people who fulfil that criteria, 113 00:11:11,780 --> 00:11:15,080 i.e. people who don't, who share neighbours better correspond with one another. 114 00:11:15,200 --> 00:11:19,430 And what that gives us is a list of pairs of nodes to explore. 115 00:11:20,630 --> 00:11:25,730 Uh, what we need to do is verify all of those, uh, candidates. 116 00:11:27,200 --> 00:11:32,330 And as this description shows, the method is, uh, necessarily a very simple heuristic. 117 00:11:32,340 --> 00:11:37,820 Um, uh, the suggested matches for each candidate can often be numerous, especially around secretaries of state. 118 00:11:38,150 --> 00:11:45,110 Uh, but when trying to make matches where evidence is uncertain and indeed designated to, uh, evade scrutiny, 119 00:11:45,410 --> 00:11:49,780 um, it is better to have lots of false negatives and to find a few positive matches amongst them. 120 00:11:49,790 --> 00:11:52,790 So that's why we're kind of designed as that kind of open net kind of approach. 121 00:11:53,690 --> 00:11:57,950 And as I've suggested, a crucial step in this methodology therefore is verification. 122 00:11:57,950 --> 00:12:03,770 So each candidate has to be checked by one by one, of course, and can compose as many puzzles as answers. 123 00:12:04,340 --> 00:12:13,910 In some cases the identity can be verified by handwriting, um, by endorsements, um where internal information acts giveaway. 124 00:12:14,120 --> 00:12:19,670 And also we might look to contextual information where we know that a given person is in a given place at a certain time, 125 00:12:19,820 --> 00:12:22,250 and makes it likely that they would have access to that knowledge. 126 00:12:23,810 --> 00:12:30,230 And as of suggested, this method works best on regions of the network that have small information hubs like diplomats, 127 00:12:30,440 --> 00:12:35,750 more miners, secretaries, um agents, unknown conspirators. 128 00:12:36,830 --> 00:12:43,820 Uh, as I'll discuss the next few minutes when dealing with um people, agents associated with a principal secretary. 129 00:12:43,940 --> 00:12:49,970 The method often presents us hundreds of candidates, which take longer to work through, uh, but still is useful. 130 00:12:49,980 --> 00:12:56,840 Um, but, um, in all cases, what this method does is it makes tractable something that is a very, 131 00:12:56,840 --> 00:13:01,190 very large and complex problem and actually makes it not suitable, uh, for users. 132 00:13:03,110 --> 00:13:05,120 So let's begin looking at the method in action. 133 00:13:05,730 --> 00:13:11,720 Uh, sprinkled with the state papers, uh, tantalising letters in which the senders or recipients are identified by initials, 134 00:13:11,930 --> 00:13:15,230 symbols, or vaguely described as an intelligence that will spy. 135 00:13:16,100 --> 00:13:23,540 Many of these figures, unsurprisingly, sit within the orbit of principal secretaries, channelling sensitive political and military information. 136 00:13:23,990 --> 00:13:29,570 Yes, precisely. The proximity to power that makes alias is used in these contexts so difficult to unravel. 137 00:13:31,250 --> 00:13:34,310 So the nature of the challenge becomes clear when we look at a few examples. 138 00:13:35,270 --> 00:13:42,590 Um, and it's quite good to show you where there has been a proposal in scholarship, because it allows us to kind of test a hypothesis. 139 00:13:43,070 --> 00:13:49,549 Um, and specifically one we're using here is one that hasn't been folded back into the, into, uh, the finding aid. 140 00:13:49,550 --> 00:13:52,940 So in the state papers, it looks like a separate identity. 141 00:13:53,330 --> 00:13:59,270 Um, so in the case of the method, returns a 704 candidate. 142 00:13:59,270 --> 00:14:06,290 So that's quite a lot. Um, but crucially, it does include Robert Barnard, um, who's already proposed in scholarship. 143 00:14:06,500 --> 00:14:12,380 Um, and that identification is strengthened, uh, by handwriting comparison with an earlier letter signed by Barnard. 144 00:14:12,740 --> 00:14:16,879 Um, by what we know of his career. He also operated under other aliases. 145 00:14:16,880 --> 00:14:24,080 So things like Robert Woodward, Robert Wood, and he was a double agent working for Walsingham and is known after the Babington Plot. 146 00:14:24,620 --> 00:14:30,500 Now it makes a useful case for testing, um, the method, um, it shows us, uh, 147 00:14:30,860 --> 00:14:34,580 that the method can return us the right answer, but amongst a lot of candidates. 148 00:14:34,790 --> 00:14:37,729 And we can narrow this by looking at kind of narrowing the temporal window, 149 00:14:37,730 --> 00:14:42,110 which often narrows things down to something like 20 or 40, uh, matches instead. 150 00:14:42,950 --> 00:14:50,690 But it does also show us the power of the alias, how you can hide in plain sight because of the crowd that around you. 151 00:14:54,020 --> 00:14:58,250 Now. What alias is used in correspondence to secretaries often generate daunting candidate lists. 152 00:14:58,520 --> 00:15:00,380 The method gives us much more uh. 153 00:15:00,710 --> 00:15:08,450 Uh tractable candidate lists when applied to cases connected to less prolific correspondence, but often still gives us implausible answers. 154 00:15:08,630 --> 00:15:11,030 And so there's a couple of examples we have here. 155 00:15:11,060 --> 00:15:18,020 We have the case of NB, um, which was either not been solved or not, is very difficult to discover in scholarship. 156 00:15:18,350 --> 00:15:21,470 Um, he wrote to Thomas Phillips in 1584. 157 00:15:21,800 --> 00:15:25,370 We get 18 candidates back, uh, including Thomas Bond's. 158 00:15:25,370 --> 00:15:30,620 Um, and we find an endorsement to B. And we know that Phillips often used that to signify bonds. 159 00:15:30,980 --> 00:15:36,379 Another example is HB, uh, which is slightly more tentative, but I think is still the correct response. 160 00:15:36,380 --> 00:15:41,150 He wrote to William Davidson, who was, um, the, uh, agent in the Netherlands. 161 00:15:41,180 --> 00:15:47,990 He was a diplomat in the Netherlands in this period. And among the candidates, I think it was exactly 100 candidates, actually. 162 00:15:48,230 --> 00:15:55,040 Uh, we find Nicholas Breunig, um, and this is very possible because he sent Davidson two letters in roughly the same period. 163 00:15:55,580 --> 00:16:00,500 Um, it's similarly written in French. It has a very comparable hand, though not identical. 164 00:16:00,920 --> 00:16:04,040 And the way that runic uh, styles is n is very like an H. 165 00:16:04,040 --> 00:16:08,599 So the C, it feels like a feasible, uh, match. Now, 166 00:16:08,600 --> 00:16:15,229 the payoff of resolving such aliases lies in what it allows us to see once identities 167 00:16:15,230 --> 00:16:19,880 are consolidated in a network topology previously fractured and obscured. 168 00:16:20,120 --> 00:16:27,740 Um, it becomes legible. Um, and for us to understand the full scope of a person's contacts, reach and access to knowledge. 169 00:16:28,400 --> 00:16:33,110 And sometimes this also, uh, exposes a different side of a figure that we thought we knew. 170 00:16:33,290 --> 00:16:36,620 Uh, one such case is the alias 88. 171 00:16:37,370 --> 00:16:41,420 Um, so here we get a really huge, uh, candidate list. 172 00:16:41,630 --> 00:16:45,830 But fortunately, I found what I thought was the solution within the within the first 100. 173 00:16:46,370 --> 00:16:54,980 Uh, and this is on the basis of handwriting and then finding a valediction that was identical to the one used by Jack LeClair. 174 00:16:55,370 --> 00:17:03,140 Um, and so, to our knowledge, this hasn't been identified before. And it gives us new information about Japan's, uh, covert diplomacy. 175 00:17:03,530 --> 00:17:09,650 Uh, what makes this case particularly interesting is that Japan is not one of Walsingham, the shadowy intelligences. 176 00:17:09,950 --> 00:17:13,760 He was. Um, in fact, uh, Secretary of state to Henry of Navarre. 177 00:17:14,270 --> 00:17:22,579 So earlier the same year, 1588, Japan had written openly to both Walsingham and Berkeley seeking support, uh, 178 00:17:22,580 --> 00:17:27,980 for Navarre, um, during the Eighth War of religion, which was fought over the right to the French throne. 179 00:17:29,630 --> 00:17:37,220 And the letter from eight. The HD eight S is markedly different from those earlier letters, um, from, uh, Japan. 180 00:17:37,730 --> 00:17:42,200 Um, we we might think of it as kind of, um, quite oblique. 181 00:17:42,590 --> 00:17:49,430 Um, it's so hard to understand the meaning as the the catalogue editor dramatically writes, if I understand this dark letter a right, 182 00:17:49,670 --> 00:17:53,749 it points out the measures for keeping the king and evar constant to the, uh, puts the party. 183 00:17:53,750 --> 00:17:58,040 So it's quite hard to see what's being said is kind of hidden. 184 00:17:58,520 --> 00:18:04,370 Um, and it's very clear that, um, uh, Japan is not acting in his capacity as the secretary of state anymore. 185 00:18:04,790 --> 00:18:12,049 Rather, he's actually coordinating with a foreign power to manage intense pressure on Navarre to convert to Catholicism in order to see, 186 00:18:12,050 --> 00:18:18,500 uh, secure the throne. Uh, and we know that he did not ultimately, um, convert to Catholicism in 1593. 187 00:18:18,830 --> 00:18:27,200 But this letter from the 88 alias suggests that there were already reasons to suspect that he would convert back in 1588, 188 00:18:27,560 --> 00:18:32,060 thereby frustrating plans for a European Protestant coalition. 189 00:18:32,960 --> 00:18:36,650 And so we can see exactly why he might choose an alias. He wants to protect himself. 190 00:18:37,160 --> 00:18:42,139 His actions may have been perceived as disloyal to Navarre. He also wants to protect the Protestant alliance, 191 00:18:42,140 --> 00:18:45,379 as information about Navarro's feared conversion could have risked their 192 00:18:45,380 --> 00:18:49,550 relationship with their existing allies or have been leveraged by their enemies. 193 00:18:52,400 --> 00:18:55,610 But of course, aliases were not only used by agents of the state. 194 00:18:57,940 --> 00:19:00,349 But also isn't. Bursaries. Elizabeth. 195 00:19:00,350 --> 00:19:07,960 The first faced repeated plots from Catholics against her position on the throne, both at home and abroad during her reign. 196 00:19:08,740 --> 00:19:16,990 And while these conspiratorial networks have been extensively studied, many aliases to which they've operated remain unresolved or disputed. 197 00:19:17,290 --> 00:19:20,860 And the method provides a kind of systematic way of intervening in these debates. 198 00:19:23,020 --> 00:19:26,979 So I don't want to spend the whole talk kind of showing you piecemeal, like we've sold this alias. 199 00:19:26,980 --> 00:19:33,760 I've sold this alias. Um, but there's a couple of examples here that, um, show what's useful about the method in particular. 200 00:19:34,060 --> 00:19:46,480 Um, it's worth showing again, the method where we find, um, for example, symbols and uh, people just given, uh, initials. 201 00:19:46,720 --> 00:19:49,250 Um, we find among those, uh, likely matches. 202 00:19:49,250 --> 00:19:57,300 So we've got an example of F here is probably, uh, Thomas Fitzherbert, a slightly more tentatively, we think na uh, is should be map to human. 203 00:19:57,580 --> 00:20:04,420 And this is consequential because they were both important leaders within the Catholic exile community in this period. 204 00:20:04,420 --> 00:20:08,410 Um, uh, Fitzherbert was, of course, active in the cause, Mary, Queen of Scots. 205 00:20:08,620 --> 00:20:10,209 And we're going to come back to Hugo in later. 206 00:20:10,210 --> 00:20:20,380 But he was a very kind of important information hub in the, um, funnelling of information between Spain and England, um, in, in the 1590s especially. 207 00:20:21,950 --> 00:20:25,430 And by resolving these aliases, we expand the information we have about these men. 208 00:20:25,580 --> 00:20:30,290 And once again, we're verifying it through handwriting letter form signatures, letter contents. 209 00:20:30,290 --> 00:20:34,280 You know. And in one of them, we've got information about Gilbert Gifford's arrest in Paris. 210 00:20:34,580 --> 00:20:40,160 And we know from contextual information that Fitzherbert was in Paris at that time and would have access to that knowledge. 211 00:20:42,570 --> 00:20:47,490 But I also want to bring up an example of case where we could not find a match because it's an instructive case. 212 00:20:47,790 --> 00:20:56,700 Um, so we have a case of SB, um, and he, um, what would this was in a single letter to address, uh, Charles Paget. 213 00:20:57,270 --> 00:21:03,780 Um, it came with 22 candidates, but it was not possible to decisively match any of these. 214 00:21:03,900 --> 00:21:09,030 The letter is a copy, so we can't rely on handwriting, um, as all the contextual clues are lacking. 215 00:21:10,110 --> 00:21:15,600 But the candidate list itself is 22. Names were revealing because they showed us two important things. 216 00:21:15,610 --> 00:21:21,629 Firstly, it shows us that the f v letter was likely to have been part of an intercepted packet, uh, 217 00:21:21,630 --> 00:21:29,820 because it occurs, uh, two of the other candidate matches W, B and more, um, were sent within the three day window. 218 00:21:30,030 --> 00:21:35,189 So it was likely that part of an intercept packet, um, and also two further candidate matches show more. 219 00:21:35,190 --> 00:21:41,610 And Martin have already by identified in scholarship but not folded back into state papers online as aliases of Thomas Barnes. 220 00:21:41,910 --> 00:21:45,239 So what it's showing us as even without a firm identification, 221 00:21:45,240 --> 00:21:51,540 we're seeing something about the community dynamics around Charles Paget, um, the recipient of this letter. 222 00:21:52,380 --> 00:22:00,510 But in so doing, it also shows us why the second variant of our method may be more useful for thinking about conspirator networks. 223 00:22:00,900 --> 00:22:06,900 Um, so just to remind you of what's different about the second variant, it costs a wider net. 224 00:22:07,230 --> 00:22:12,630 So if you begin with a single person, you can cast the net in their entire community around them. 225 00:22:13,170 --> 00:22:17,910 Um, anybody who they correspond with who may have been an alias. 226 00:22:18,120 --> 00:22:25,860 And that's really good because it doesn't require us to know in advance which names are aliases, because some aliases are not symbols and initials. 227 00:22:26,040 --> 00:22:29,189 Um, as in the case of Sherman and Martin. 228 00:22:29,190 --> 00:22:32,730 Uh, Barness pseudonyms don't announce themselves as aliases. 229 00:22:33,890 --> 00:22:37,760 Um, and it helps us to uncover things like letter bundling and things like that. 230 00:22:38,990 --> 00:22:48,649 Um, which brings us to the case of William Sarah, which we started with that figure, that, um, network diagram, steroids case, uh, 231 00:22:48,650 --> 00:22:56,920 illustrates how profoundly alias attribution can reshape historical interpretation depending on which identities are mapped onto him. 232 00:22:56,930 --> 00:23:01,459 Sterile appears either as a minor and perhaps not very efficient conspirator. 233 00:23:01,460 --> 00:23:07,880 Um, uh, intelligence, uh, or perhaps someone actually working actively with Catholic conspirators. 234 00:23:08,180 --> 00:23:11,510 Um, following, uh, Patrick Martin and John Fiennes's suggestions. 235 00:23:11,780 --> 00:23:18,980 Uh, he was a man managing multiple personas. Um, and on both it provides a systematic way of testing these competing interpretations. 236 00:23:19,130 --> 00:23:27,440 So was he in a ineffectual, uh, in intelligence or was he actually conspiring with the Catholics he was reporting against? 237 00:23:28,010 --> 00:23:32,149 So this started out as part of the Earl of Essex, uh, shadow Foreign Office. 238 00:23:32,150 --> 00:23:37,550 So he, um, in the 1590s, he wanted to set himself up as a successor to Walsingham. 239 00:23:38,450 --> 00:23:44,030 Um, and it was designed, um, as an independent intelligence operation, uh, with Francis Bacon, 240 00:23:44,300 --> 00:23:50,720 um, designed to interpret, uh, shape and redirect information circling amongst Catholic exiles. 241 00:23:51,110 --> 00:24:00,620 And central to this endeavour were Thomas Phillips and William Sterile, whose coordinated use of aliases created such a massive puzzle for us, 242 00:24:00,740 --> 00:24:07,730 uh, working that now the strategy developed was straightforward in conception, uh, but sophisticated in execution. 243 00:24:07,880 --> 00:24:12,380 Sterile would pose as an agent sympathetic to Catholic exiles in Flanders, 244 00:24:12,650 --> 00:24:18,560 feeding them false intelligence reports, um, while reporting back emerging plots against Elizabeth. 245 00:24:19,310 --> 00:24:27,709 So this plan built on, uh, sterilise existing contacts in the Netherlands, particularly amongst figures already suspected, um, of, uh, 246 00:24:27,710 --> 00:24:35,450 involvement in intelligence gathering and conspiracy, including Stanley Holt Sherwood, Edwin Fitzherbert, um, a figure known as Captain Jakes. 247 00:24:38,050 --> 00:24:44,890 So to appreciate the stealth of this operation and the difficulties it poses for modern reconstruction is necessary not to map, 248 00:24:45,300 --> 00:24:52,090 uh, not just a single correspondence network, but multiple overlapping networks associated with the aliases used by sterile, uh, 249 00:24:52,090 --> 00:24:57,490 which, uh, all, uh, listed here, um, with various levels of scholarly backing. 250 00:24:57,640 --> 00:25:03,070 Our method, of course, does not work on this series of, um, three dyads I'm talking about here. 251 00:25:03,250 --> 00:25:06,700 But once we start collapsing the network, this is when the method comes into its own. 252 00:25:07,900 --> 00:25:13,150 So fortunately, the archive itself preserves traces of how their system operated. 253 00:25:13,360 --> 00:25:17,800 And therefore, some of the aliases have already been noticed by the editors of the calendars state papers. 254 00:25:18,850 --> 00:25:25,270 So, for instance, names have been scrubbed and replaced. So this allows us to map the sterile name to, um, Henry. 255 00:25:25,270 --> 00:25:31,179 Some, uh, we have a few endorsements are letters uh, left by Philips, for example, 256 00:25:31,180 --> 00:25:35,930 was a letter sent by uh Franklin, which is endorsed by Philips as to some. 257 00:25:37,470 --> 00:25:46,080 And we also have polygraph evidence which suggests that Harry, which could be confirmed through the ham box in comparison with the Sandman identity. 258 00:25:46,830 --> 00:25:51,569 Um, so this is this is an alias has long been associated with him, 259 00:25:51,570 --> 00:25:57,030 and we can kind of confidently say that these are all the same men operating under the same, uh, under different names. 260 00:25:58,410 --> 00:26:02,250 And what becomes clear by looking at these is that the aces were not randomly divided. 261 00:26:02,370 --> 00:26:09,700 They were relational. For example, uh, the alias Terrell was, uh, the name used for Phillips's internal records. 262 00:26:09,720 --> 00:26:15,720 That's where he's at. That's why he's endorsing, uh, some some men, um, uh, uh, sterile. 263 00:26:16,050 --> 00:26:22,950 Um, his the name that he was given in the XL community, but, uh, comparison was Robinson. 264 00:26:23,370 --> 00:26:27,990 Either that's how he was known or that's how he instructed that community to call him. 265 00:26:28,650 --> 00:26:33,959 Um, and let's us from the Robinson alias. What, uh, was said to not Phillips, but to Morris. 266 00:26:33,960 --> 00:26:37,740 So Phillips there was operating under the Morris, uh, name. 267 00:26:37,950 --> 00:26:44,310 So, in other words, identity was carefully segmented by audience and sometimes even by direction of communication. 268 00:26:46,510 --> 00:26:49,690 Now more complex is the identity of Peter Hollins. 269 00:26:49,930 --> 00:26:53,830 So while traditionally this has been attributed solely to Philips, 270 00:26:54,100 --> 00:27:02,530 Martin and Fitness have argued that Helen's actually functioned as a shared or composite identity used collaboratively both by sterile and Philips. 271 00:27:03,100 --> 00:27:09,340 Um, now there's good evidence that several of Philips did, in fact, uh, operate shared aliases. 272 00:27:09,730 --> 00:27:17,799 So the which of name, for example, which was used by um sterile for outgoing letters in 1602 only was also used as what 273 00:27:17,800 --> 00:27:21,760 we might think of as a kind of inbox or like something like a post office box. 274 00:27:22,030 --> 00:27:27,340 Um, years earlier, the Helen's alias appears to have operated in a similar way, 275 00:27:27,610 --> 00:27:33,200 being used primarily for incoming correspondence in the period 1597 to 1602. 276 00:27:33,220 --> 00:27:40,030 It's never used as an outgoing, um, alias, but unlike which, um Helen's had this kind of fully realised persona. 277 00:27:40,030 --> 00:27:47,290 He was a merchant, a Koopman from London, uh, and different contacts wrote to it, but the two different hands likely managed it. 278 00:27:47,740 --> 00:27:52,030 Although to the outside world it seemed like one single coherent persona. 279 00:27:54,410 --> 00:28:02,000 So from these securely identified aliases, um, we've been able to map Dan Robinson. 280 00:28:02,000 --> 00:28:05,600 Sam which and Franklin and Helen's were single node there. 281 00:28:05,600 --> 00:28:12,170 So we get this consolidated network at the top, and we see already how sterile is a small information hub. 282 00:28:12,770 --> 00:28:17,180 We've also collapsed loops to the aliases of Maurice, William and Talon's. 283 00:28:17,570 --> 00:28:20,270 But we still have these challenges, uh, down the bottom. 284 00:28:21,380 --> 00:28:32,360 Um, uh, also based on the, um, testimony of, uh, Felipe's, um, we have also collapsed on a couple of aliases on Hugh Owen. 285 00:28:32,510 --> 00:28:39,200 Um, we know that he was Pettitt, uh, that he also used the Thomas Nichols and, uh, Nichols aliases. 286 00:28:39,380 --> 00:28:44,690 So this is a this is a partially collapsed network based on material evidence. 287 00:28:45,200 --> 00:28:51,679 However, it's important to note that these verified identities, for the most part, only cover incoming correspondence. 288 00:28:51,680 --> 00:28:58,700 Two Sarah Phillips and Phillips testimony given in 1603, in response to accusations of conspiracy. 289 00:28:58,700 --> 00:29:04,100 But I would note that I wouldn't believe that the correspondence travelling in the opposite direction to come from a friend of sterols. 290 00:29:04,340 --> 00:29:06,260 But under what name are they written? 291 00:29:06,800 --> 00:29:17,330 No letters from Helens are coming, and this is why we come to the more contested, um, uh, the more contested, um, examples. 292 00:29:18,530 --> 00:29:24,680 Martin and Fitness have argued that further aliases such as Fenner Cordell and Vincent Orwell and even Anthony Rivers, 293 00:29:24,980 --> 00:29:26,390 uh, should be attributed to sterile. 294 00:29:26,750 --> 00:29:35,690 And so this matters because, um, rivers actually appears in historical record as someone who is regarded as very crucial to Catholic networks. 295 00:29:35,690 --> 00:29:39,050 And specifically, he's got his own ODB entry. 296 00:29:39,260 --> 00:29:42,680 Sterile does not have an academic entry, so this is consequential. 297 00:29:42,950 --> 00:29:45,589 So if this script, this would radically recast sterile, 298 00:29:45,590 --> 00:29:51,020 not as a marginal or ineffective intelligence, but as a figure deeply embedded in Catholic strategy. 299 00:29:51,020 --> 00:29:54,200 And so this is Martin and Venice's kind of argued big argument. 300 00:29:55,190 --> 00:30:00,080 His empowerment failures may have been calculated to, in fact, serve the exile community. 301 00:30:00,470 --> 00:30:06,799 Now, our purposes engaging with this claim is twofold. First, to test whether the method supports these identifications, a second, 302 00:30:06,800 --> 00:30:11,000 to see whether collapsing these aliases onto a single node enables further discoveries. 303 00:30:13,730 --> 00:30:17,720 Now we can justify a second round of consolidation using two pieces of evidence. 304 00:30:18,020 --> 00:30:22,440 Firstly, our method predicts that Coaldale could in fact be thinner. 305 00:30:22,670 --> 00:30:29,780 And you can see that if you imagine sort of diagonally folding this figure onto the yellow, um, you can see how that that would work. 306 00:30:29,870 --> 00:30:36,469 They could be the same person because they have no contacts. Secondly, the handwriting, I think it's absolutely. 307 00:30:36,470 --> 00:30:40,700 Uh, so these are from the, um, state papers on lies, the horrible quality. 308 00:30:40,700 --> 00:30:47,270 But I think it's very clear that this letter by Fenner is in the same hand as this letter by Coaldale and this letter by reverse. 309 00:30:47,840 --> 00:30:55,100 And I think looking at all of the letters by these guys on the hand and the layout, the format and the way in which they encode the date, 310 00:30:55,850 --> 00:31:00,920 we can definitely say that Fenner, Cordell and Rivers are the same person, 311 00:31:01,850 --> 00:31:07,430 but all of that is that cluster of aliases the same as the sterile cluster of aliases? 312 00:31:07,460 --> 00:31:10,580 If we can match one, then we can prove that they were all the same person. 313 00:31:12,730 --> 00:31:15,460 So the decisive evidence. Oh, I should have said, um. 314 00:31:15,700 --> 00:31:21,310 The really important thing is, is that the handwriting of these letters looks nothing like the sterile aliases. 315 00:31:21,820 --> 00:31:27,310 But that isn't damning evidence in itself, because we can see that sterile wrote in several hands. 316 00:31:27,670 --> 00:31:30,729 We don't see him writing in this hand, but we see him writing in several hands. 317 00:31:30,730 --> 00:31:39,640 He was able to disguise his handwriting. The evidence that cinches it, though, is a little detail, and it might seem like a little detail. 318 00:31:39,910 --> 00:31:47,290 Um, it's about gloves. Um, so in a letter sent to Sterile and Phillips under the Harlan's alias. 319 00:31:47,740 --> 00:31:56,410 Um. Uh. Oh, and sorry, in a letter sent to Cell and Phillips under their Harlan's alias, Owen, 320 00:31:56,410 --> 00:32:01,959 under his Petter alias, requests that some sample gloves be copied and made up into two. 321 00:32:01,960 --> 00:32:06,490 Doesn't pair. So I sent my host in Antwerp. And to whom? I give you charge to deliver my letters to the post. 322 00:32:06,850 --> 00:32:15,790 Two gloves of some resort to be sent unto your unto you, praying you to cause to be made of each fashion a dozen pairs, and that of the like letter. 323 00:32:17,710 --> 00:32:26,750 So later that month we have what looks like a reply, a letter sent under a different alias tenor to, uh, Piccolo and Eddino. 324 00:32:27,380 --> 00:32:31,990 Um, and we think it's a reply because he says, I have spoken your gloves, but they are not yet ready. 325 00:32:32,500 --> 00:32:40,090 Um, that would mean that Fenner potentially is Harlan's and petitioner, or at Lino is Pettit. 326 00:32:40,230 --> 00:32:44,500 Uh, we think the latter. Later that month, again, call Dale another alias. 327 00:32:44,500 --> 00:32:51,130 Again writes to Geibel saying, I wrote to you and Edlin for Brussels like forbid, and I shall expect it. 328 00:32:51,280 --> 00:32:55,780 And by the next opportunity I hope to send Edlin two dozen pair of gloves according to his pattern. 329 00:32:56,680 --> 00:32:57,940 So it's a little detail. 330 00:32:58,090 --> 00:33:09,520 Um, but I think these are the links in the chain that show that Collins is Fenner is cordial, which means that Anthony Rivers is also William Sterile. 331 00:33:10,360 --> 00:33:12,330 Um, as an aside, I don't think these are real gloves. 332 00:33:12,340 --> 00:33:20,350 I think that the gloves are a coded reference, um, probably to, um, manuscript texts for scribal publication. 333 00:33:20,740 --> 00:33:24,790 Um, that seems the most likely temptation and leaving them on, but I think means leaving them unbound. 334 00:33:25,210 --> 00:33:31,360 Um, so once we collapse the sterile identities and rerun the match, that something important happens. 335 00:33:31,600 --> 00:33:36,610 New candidate matches emerge. Um, and now these can be solved iteratively. 336 00:33:36,790 --> 00:33:41,290 And that's what began as a set of isolated puzzles become a dynamic process of reconstruction. 337 00:33:42,010 --> 00:33:44,820 Um, and I don't again, I don't want to go through things in details, 338 00:33:44,830 --> 00:33:49,680 but I think from this work we're able to suggest several new clusters and several new matches. 339 00:33:49,690 --> 00:33:55,460 So, for example, uh, for Hugh Owen, we've already collapsed in with Pettitte and Nichols. 340 00:33:55,750 --> 00:34:02,620 Uh, but we then can suggest additional matches, um, as you know, in the canon, which are almost certain and already suggested in scholarship. 341 00:34:02,620 --> 00:34:09,610 But also we want to suggest a couple more matches that aren't suggested yet, which might be John Leclerc and, uh, Carhart. 342 00:34:10,090 --> 00:34:17,800 We have a alternate proposal for Carhart as well, which might be, uh, Derek, which has already been proposed as an alias of Richard Sherwood. 343 00:34:18,730 --> 00:34:22,900 We also suggest a cluster, um, of aliases. 344 00:34:23,110 --> 00:34:28,090 Uh, Garnett self a guy balls are likely Richard for Stegen. 345 00:34:28,480 --> 00:34:33,040 Um, it doesn't suggest Richard was a match, but he was actually really, really good at keeping his aliases separate. 346 00:34:33,370 --> 00:34:38,229 Um, but Gance, uh, South and Diebold, uh, definitely the same person. 347 00:34:38,230 --> 00:34:49,600 The handwriting is the same. Um, it shows, um, the intermediary role that, uh, the Stegen had in funnelling information to Brussels and to Rome. 348 00:34:50,770 --> 00:34:53,499 We have a third, uh, a further, uh, bundle. 349 00:34:53,500 --> 00:35:03,700 We think that correlates Piccolo and Giuseppe to singer, uh, William Baldwin and Marco, two singer and Akiko, uh, Robert Persons. 350 00:35:03,700 --> 00:35:07,660 And they were two very important leaders in the Catholic exile community. 351 00:35:07,840 --> 00:35:15,100 And finally, we have this figure of Ghazali, who we think is separate from all of his people, but has an intermediary role like the Stegen. 352 00:35:15,100 --> 00:35:22,030 And it could be either Derek Sherwood or Francis Harvey, who was a merchant in the in the Netherlands. 353 00:35:22,450 --> 00:35:27,170 Um. So we've made lots of proposals today. 354 00:35:27,530 --> 00:35:34,460 And taken together, I hope these select examples show how the method is useful for solving aliases and mapping possible theories. 355 00:35:34,670 --> 00:35:38,360 But it's not just solving about solving a set of archival puzzles. 356 00:35:38,510 --> 00:35:43,790 The method also helps to make visible the operational logic of secrecy itself. 357 00:35:44,540 --> 00:35:53,839 Seen from this broader perspective, the proliferation of aliases appears less as a series of ad hoc disguises than as a series of strategies, 358 00:35:53,840 --> 00:36:00,230 and these strategies and trends come into focus gradually through the case studies, but could be summarised as follow. 359 00:36:00,920 --> 00:36:07,010 The agents and conspirators commonly maintain multiple identities that these were tailored to different relationships. 360 00:36:07,640 --> 00:36:13,040 The incoming and outgoing correspondence often was often separated across different pseudonyms, 361 00:36:13,040 --> 00:36:22,100 to make it really difficult to reconstruct handwriting well, sometimes the disguised aliases might be shared or collaboratively maintained. 362 00:36:22,400 --> 00:36:30,559 Um, secure management of intelligence sometimes required individuals to hide additional identities when the collaborators and some aliases function. 363 00:36:30,560 --> 00:36:37,220 Something like a post office box that's attracted to single aliases might in fact be redistributed through intermediaries like the Stegen, 364 00:36:37,970 --> 00:36:41,840 and this infrastructure often invites the bundling of letters into packets. 365 00:36:43,750 --> 00:36:47,340 Uh, and these kind of strategies, you know, strategies like, um, 366 00:36:47,350 --> 00:36:52,210 this if you post office box meant that maybe people who were more peripheral in the network, 367 00:36:52,450 --> 00:36:58,440 maybe not at its centre and might not be entirely trusted, couldn't reconstruct the entire path that communication was taking. 368 00:36:58,450 --> 00:37:04,600 So it was a useful strategy. But more broadly, by showing how our method unfolds in practice. 369 00:37:05,020 --> 00:37:10,630 I have also shown the value of computational heuristics, uh, for this kind of historical research. 370 00:37:10,960 --> 00:37:15,230 Computational methods should never be substitutes, of course, for archival expertise. 371 00:37:15,310 --> 00:37:22,000 And here we show how archival skills and knowledge in the history of the period are absolutely crucial for verifying candidates. 372 00:37:22,570 --> 00:37:27,729 But the number of newly solved aliases and new hypotheses would not have been possible without this method. 373 00:37:27,730 --> 00:37:31,990 Because of the way it renders the search for candidates more tractable. 374 00:37:32,530 --> 00:37:35,979 Its value is not just the way that it may lead us to new matches, 375 00:37:35,980 --> 00:37:42,670 but also the way it allows us to explore and test hypotheses and, um, you know, counterfactual accounts of history. 376 00:37:43,630 --> 00:37:47,110 Um, I'd like to close by pointing out that this method is agnostic. 377 00:37:47,110 --> 00:37:54,849 It wasn't developed, uh, it was developed for the state papers, but would work on any kind of correspondence data set that is not an ego network. 378 00:37:54,850 --> 00:38:01,690 So not there isn't just one person. And that context has to have those kind of, uh, alter connections between them. 379 00:38:02,620 --> 00:38:05,110 Uh, but it's also not just limited to solving aliases. 380 00:38:05,320 --> 00:38:09,520 As I said at the beginning, aliases are what we might think of as an advanced deduplication issue. 381 00:38:09,700 --> 00:38:17,890 So you could use this as a method for data cleaning and to finding finding those, um, those, uh, duplicates entities in your data sets. 382 00:38:18,070 --> 00:38:24,219 But it's also, um, I've got Howard Hudson here. Say it's also been used in the networking archives projects to look at other 383 00:38:24,220 --> 00:38:27,670 community based reasons that people might not be corresponding with one another, 384 00:38:27,700 --> 00:38:31,000 maybe because they were actually avoiding corresponding with one another. 385 00:38:31,600 --> 00:38:38,890 Um, so my main takeaway is that I hope that other people will find this, uh, heuristic useful and pick it up, uh, for their own applications. 386 00:38:39,160 --> 00:38:42,790 So finally, I'd just like to thank again my collaborator, uh, Sebastian, uh, 387 00:38:42,790 --> 00:38:47,290 to our funders, this came out the two Networks of Power project funded by HRC. 388 00:38:47,290 --> 00:38:52,620 And to thank you all for listening. And I look forward to questions. Thank you.