Abstract

A plenary panel, moderated by Sarah Weicksel of the American Historical Association (AHA), on which institutional changes, from graduate training to funding, would let computational history become a durable part of the discipline rather than a permanent margin. Matthew Connelly points to the post-2008 collapse of the job market and the risk to junior scholars who spend years cleaning data, Jo Guldi describes the lost promise of digital history because funding and hiring never followed, and Christopher Phillips says the long timescale of historical work and its focus on specific evidence slow change. The panel then argues for collectively chosen ambitious goals, including Guldi's proposed global index of climate harms, and Connelly compares the roughly 48 billion dollar annual budget of the National Institutes of Health (NIH) with the roughly 400 million dollar budget of the National Archives in making the case for a National Institute for History. Loren Moulds presents Discord poll results from attendees, and the audience asks how to present AI-assisted methods on the job market.

Chapters

Transcript

AI-generated transcript. It was produced from the recording by speech recognition and cleaned up with AI. It has not been edited by the speaker and may contain errors. The captions on YouTube are AI-generated in the same way. Each timestamp links to that moment in the video.

[0:15] Good morning, everybody. Morning, everybody. Welcome to day two of the AI and History Conference. I'm Louis Hyman. I'm really glad to see you all here again. And this morning, I'm very excited to have a panel with the illustrious people before you. Sarah Weicksel is the executive director of the AHA, and she's going to guide us through a conversation on thinking about what this means for the discipline of history with this panel before us. So thank you so much. Gee, I thought that we were going to get a really cool name right there. we'll think on a team name. Good morning, everyone. It's such a pleasure to be here again. And since I have the mic, I want to take an opportunity to just thank Louis for the opportunity to partner on this with him and for all of the amazing work he and his team here have done.

[1:17] I also want to say, and I also want to say thanks to Laura Ansley, who's here, as well as Hope Shannon, who's back at the AHA townhouse in D .C. for coordinating things on our end. It's been an amazing day yesterday, and I'm really looking forward to today. So I have with me three very amazing historians, and I'm going to give you a very quick introduction to them. Matt Connelly is Professor of International and Global History and the Vice Dean for AI Initiatives at Columbia University. Jo Guldi is Professor of Data and Decision Sciences at Emory University. and Christopher Phillips is professor and chair of history at Carnegie Mellon University. So one of the most important roles that I consider the AHA to

[2:18] have is cultivating an environment in which history and historians can thrive. So today we're going to be talking about the future of the discipline, particularly in this moment. But of course, everything has a history, so we need to go back a little bit. So for you all, given the ubiquity of computing tools among working historians, we're using databases, OCR, search engines, word processing, and also the intellectual achievements of digital humanities since the 1970s. We've built digital archives, spatial histories, text corpora, databases, computational methods, new forms of scholarly publication. Given all that, still, by and large, the training, the practice, and institutions of history have not radically changed since the digital turn. Why? And is there something different about this moment?

[3:27] Well, I think it's multi-causal. I mean, I think that part of it is that, you know, there's been, as we all know, like a collapse, you know, in the job market for new historians. It coincides with the, and partly caused by the financial crisis. you know that was a time I think a lot of us here in the room can remember you know there was a lot of excitement you know remember the THATCamps and so on and I think that you know if the historical profession had you know continued hiring at the same rate that it had been up to that point then I think a lot of our colleagues you know would be tenure track or tenured colleagues you know in these departments and so what happened instead right I mean we all know what happened So I think that's part of it. Any given department feeling a sense of scarcity where they lose their one Japan historian, and you try to make the case, well, maybe we don't need to think of it necessarily as a slot for Japanese history.

[4:28] Maybe we should be open to hiring somebody who's going to help the rest of us develop new methods. It was always going to be a difficult case to make. So that's part of it. But I also think, as digital historians, we face the challenge, and I didn't because I'd just gotten tenure, so it was okay. But my younger colleagues, we all know you could just invest an enormous amount of time training yourself, cleaning data, et cetera, and have nothing to show for it. And who would take on that kind of risk? And so to do that work, typically what you'd want to do, and if you're lucky like me and you have tenure and you can get external funding, you'd want to build a whole team. but to do that kind of work requires rethinking, and this is what we always talked about at the THATCamps, right? We have to rethink professional incentives. We have to rethink how we evaluate each other's work, and that was always going to be a heavy lift. It was just going to be hard to get the historical profession to rethink the idea of the book as the unit, the coin of the realm. It was always going to be difficult to get them to wrap their heads around multi

[5:31] -authored work, so all these things were going to be hard. But the last thing is, I feel like some of the problem had to do with how, as much as you rightly acknowledge the very real achievements of digital history, it's hard to point to the discoveries, like the things that we discovered that nobody knew before. And I think a lot of the work in digital history had to do with finding new ways to represent the work we do, to share it with others, to make it more interactive, etc. things. But to really earn the respect of our peers in the profession, we had to tell them things they didn't already know. And we didn't do enough of that, right? So I think part of this is on us. So I'll tell a different story. A couple of months ago, I published a post on a computational history substack, which some of us contribute to, called The Lost Promise of Digital History.

[6:32] Big statement coming from me. I love digital history. I've identified as a digital history, as perhaps my main identity for the last 15 years. So to be clear, I don't think anybody in this room lost us that promise. And I don't think it was a failure of good work, quite to the contrary. But what I wanted to say in that article was 15 years ago, when I was the young assistant professor or postdoc sitting in a room like this, people were saying to me, this is such field. There are going to be so many opportunities ahead. This is going to transform departments. We've seen data transform political science, transform economics, transform neuroscience. History is going to be next. There are going to be so many hires. People are going to be in a hiring war over you. That sounded great. And that didn't happen. I, you know, I'm happy with the community in which I found myself. And I'm so happy to be in the room with you today. But as a historian, it raises questions. Why didn't that promise happen? If that transformation was happening simultaneously

[7:36] in other disciplines around us, why did the goods not come to us? And so when I look back to the 1960s and 1970s, what I see now is a series of other lost promises for history that are just things that we could have done that became a little incomplete. And so many of you have parts of this story. Part of it is that the economic historians like Robert Fogel, Matt and I were talking about this this morning. Robert Fogel publishes about American history, gets rewarded for that, gets the Nobel Prize, and the money streams into economics departments, not into us, because the economics historians just happened to have left because the salaries were higher. Whoops. So we invented something that brought a lot of money, and then we lost control of it. Then we invented such interesting things on the social history and the cultural history. And some of us spent, Stephen Robertson spent a lot of time, Lara Putnam spent a lot of time, whereas Ruth, a lot of us spent a lot of time

[8:40] trying to figure out how do we use big data and make the social history and cultural history scale so that we can have a new scale of insight on the promise of what E.P. Thompson and others taught us to do long ago. Hey, another person, we made it scale. And it was really hard. And there were tiny bits of funding in NEH funding. And we thought there's going to be more funding if we do this right. And the departments are going to hire more people like us. And we're going to train up the next generation. And that didn't happen. We did the research. We built the methods. The funding didn't arrive. The institutions didn't arrive. So one reason that I'm so glad to be at this convening with all of you is that many of us who are a little older have started to have a conversation about what's necessary to build a city that we're all going to live in. We're taking seriously those lost promises and saying, can a department chair do anything to make sure that the right people get hired?

[9:41] Can an institution invest? Can we have conversations with philanthropy? Can we have conversations with the National Academies of Science that the young aspiring graduate student and assistant professor is going to be able to look at a day when they get to choose their five closest friends and the most important project that they can think of and go after it. And we're going to hold you up. So I'm telling you now that we are trying, we're putting together that plan to hold you up so that the promises won't be lost for the future. And I don't know if we're going to make it, but I think a lot of the rest of this panel is going to be about that. And I'll also say that everybody in this room past a certain age has been thinking along the same lines. And there is a very active conversation. So there are at least two dozen people in this room right now who could also be on the stage with us, also answering questions about how we're going to build that. And we need more conferences. We need more seminars. We need panels at the AHA, at your other history conferences.

[10:43] So each of you has a part to play in this, in having a conversation like this. The four of us don't have to be in every conversation. There are a lot of smart people who are going to be part of the solution. I'll give a couple of other, take kind of another tact, which is the kind of work most of us do, which is to say in all disciplines, training is sticky. Change is slow. We tend to do things the way we were taught. It's very hard to pivot. But in history, our work is long. If you're in a six to eight articles in a relatively short amount of time, right? There's a kind of quick turnover from the start of an idea to the finish of a project. Sadly, or maybe happily, we're in a different kind of timescale in history. We work on things three, five, seven, 10, 40 years, right? We try and do that kind of long work. So that stickiness of training really plays out in a much more extended way in history. So I think that's one kind of backdrop of history that I

[11:47] There's also an intellectual focus on evidence. And I think one of the things that computational methods, statistical methods do is move to a kind of generalizability or a kind of aggregation. This is useful. It's important. We need to know how our evidence fits in with what's normal, what's general, what's typical. But I also think that all of us are trained to really focus in on specific pieces of evidence. And so that movement between the general and the specific, the idiosyncratic and the typical, that movement is very, very difficult intellectually, institutionally, you know, in our publications, it's hard to do it. And so I think some of the epiphenomena that we're discussing here, why are we stuck on books? You know, how can we hire and promote and change our standards in an effective way? A lot of these institutional problems, if you will, also come from really important intellectual decisions that we as historians make.

[12:47] And so I think from one of the reasons why it's hard is that the way in which we do work, the way in which we value scholarship plays in some ways a slowing or a dampening role on some of the change that we've seen in other disciplines. And the last thing I'd say, though, is that many of us know that between the 1930s, the 1920s and the 1970s, historians were doing quantitative methods. Right. So it's it's not it's not as if we have to pivot to this entirely new world. Rather, it's that we need to explain in some ways if there's if there's something to explain, it's the movement away from that tradition of quantitative combined with qualitative history. that all of us who had to read some of those classics from 100, 115 years ago in our methods class, and you think, oh my goodness, this is actually a really interesting mix of economic, demographic,

[13:47] sociological, rhetorical analysis that was happening 100 years ago. And we've only focused on some of those in the last 30 years. And so some of it is reclaiming what we've always done, and some of it is acknowledging, I think that the values we have in our research in some ways cut against kind of quick disciplinary pivots. So I want to pick up on something that both Chris and Jo have motioned towards. And that is, what makes us different from the other humanities disciplines in this regard? is there an important distinction between computational history and computational humanities i'll start and i think i know the least about this but i'm happy to go first then so it gets better as it goes uh but i think what i what i would say is that two things one i think just speaking bluntly politically history is in a way better

[14:48] position than the humanities And so we should embrace history. And I think there is no no reason not to do that. Everyone, everyone cares about their history, their family's history, their people's history, their community's history, the history of other communities. I mean, this is just it. I don't want to say it's universal, but I mean, it's it's pretty universal among humans that we care about history. And so I think we should embrace that. And I think there's no reason to run for that. And I think in some ways, the kind of where do the humanities begin and end question is we don't have to worry about it. We're doing history. And so I think that's really important. I think another aspect is that some of the other humanities, unlike history, were not as situated on the border of the social sciences. And so you have a lot of the humanistic disciplines that are focused heavily on judgment, on aesthetics, on particular kinds of what you might say are more artistic almost components.

[15:51] Whereas history has always valued evidence. It does not matter how well your narrative is written. if you do not draw in a sophisticated, cutting-edge, reliable, traceable, replicable way from evidence, it's not going to be valid. We're just not going to accept it in the field. In fact, we'll go after you if you think it's too pretty, because we'll think something's wrong here. The way this is being written, and that's why you get some of the divide between so-called popular or public history and so-called research history or academic history. I think those divides are actually arbitrary in lots of ways. But one of the ways that you see this is this focus on where is your evidence? Do you have new evidence? How are you drawing from the same evidence? And I think that's different from some of the other humanities. You know, I was sheltered by literary digital humanists for a long time. I learned so much from people like Ted Underwood and Richard Jean So, and they really incubated my historical practice. And so my books are filled with their footnotes. But one of the things that the theorists of history, the philosophers of history,

[16:56] intellectual historians have been helping me to understand is the specific role of history in working with data. And maybe two points here. One of them is consensus. Consensus about what the facts are, where they live, in what archives. That is what we do. and in a time characterized by rampant distrust of institutions and each other, it's consensus about the basic facts of our past that we are charged with. What can we know? Reiterating that message and telling the people who design LLMs, we have expertise to bear about where the facts live and what we can agree about. Do historians debate endlessly? Yes, but we debate with more and more precision. We disagree about this particular interpretation, these particular documents. We all agree about all of the other things, about where the documents are and where the documents live. And if you don't agree with us, you might be a conspiracy theorist. So that's really important. The other thing that historical theory has been teaching me is about the kind of

[17:59] clientele of history as an applied discipline, that we're teaching skills about how to theorize temporality in the archive and in archives of data. Marnie Hughes-Warrington has taught me so much about this. Thank you from the bottom of my heart, Marnie. But it helps us to understand, if we understand that this is fundamentally what any of our students bring to bear and what is generalizable about any of our projects in the archive with our LLMs and so on, what we're learning is how to take those facts, put them in order, find the gaps, find the silences, figure out what caused what. And that is incredibly important to people who are in the insurance industry, in intelligence, in cybersecurity, and in the labs themselves. And we have barely as a discipline begun conversations with the corporations that are nearest. Chad Gaffield's been talking to me about Ancestry.com. There are corporations that They're very invested in having the data, the facts

[19:02] about what happened, getting them right. And so this opens lots of doors. It opens doors for projects about like, well, would you like to be a really well -paid historian? Maybe don't be afraid of insurance history. There could be, you could have some very rich and powerful friends. You could have your own building. Just saying. So many frontiers, not that we all have to do that. Not that we have to lose our commitment to social and cultural history. is knotted in the framework of who we are as a discipline. But these horizons are out there. The theorists, the intellectual historians are helping us to lay out who we are as a discipline. Thank you, Chris and Jo, for giving me time to think about what I was going to say. But, you know, I came very late last night because I was teaching. I teach a great books course. So, you know, yesterday was the Quran until, you know, yesterday was Plato and Aristotle and so on. And so, you know, my department is unusual in that at Columbia, the history

[20:06] department is one of the social sciences. But we all of us, like many of us feel like we're humanists too. I certainly feel that way on days like that when I'm teaching these books. And, you know, as it turns out, the question, like, what is it to be human? like the core question of the humanities is incredibly timely and important. Like it has never been like so obviously important to understand what it is to be human in a way that like the engineers understand now. Like the engineers are asking us to help them to solve the hard problems, right? And I don't know if anyone else has noticed this, but there's been after a long decline, like almost all the other humanities departments, philosophy is coming on philosophy now. So I think it's fantastic that, you know, as a discipline, you know, we have our foot in the humanities, but then we have this other foot in the social sciences. And I totally agree, like, certainly in my institution, we're much more fortunate being aligned in some way anyway, with the economists and the

[21:09] political scientists and so on. And, you know, what is it that makes us social scientists? Well, I mean, it's because we're empirical, like if it's not, you know, and you know that also means that we have entree of another kind you know with uh people who are interested in data and want to ask questions and answer them in replicable ways uh so so i think in many ways like history is very well positioned you know for for this moment and and because whether you're a humanist or whether you're a social scientist or both you know we continue to grapple with questions where there's not necessarily any one right answer like it's all about judgment and interpretation. So I think that's why, you know, do you ever see the early studies by economists? Like what are the jobs that are going to be most impacted by AI? I think like history was at the top of the list, you know? And like there were a few headlines about it at the time. And then it's like, you actually read their methods and it's like, no, this is the jobs that these are the jobs that'll be most affected by the coming of AI because

[22:13] of the nature of the work. And so I asked myself, well, why is it more my colleagues don't understand this? Because, of course, if your work is trying to gather information, ask questions of it, find answers, find if they're true, et cetera, right? I mean, of course you're going to want to use the kinds of tools that are available to us today. So I think the door is open. We just shouldn't be so afraid, right? I mean, there's just huge opportunity for us. Yeah. I'm trying to decide to go down the we'll be replaced by AI route or another route. We'll go another route. Because I, well, I will comment on this though. Because I am regularly asked whenever a new study comes out, are you worried? are historians going to be replaced by AI?

[23:14] And I've pulled up the studies, I've read through their methods, I have read through their data, and no one really defines what skills they think are being replaced. So if we're timeline creators, sure, we can be replaced, but we're not. We are very human. Now, I was trained in a department that was situated in the division of social sciences, but I'm a historian who works with material culture and am very much someone who probably leans more towards the humanities than to the social sciences on that spectrum upon which we are all kind of situated in different locations. So what do I, as someone who is not engaged in computational history, and what should every historian know how to do now? What should we be thinking about

[24:15] the training of students? how do we think about historians who will be getting their PhDs in 2030 and what they should know how to do when they emerge from their programs? So I think about the fact that every freshman entering the university right now has the opportunity to ask an LLM, what are the most salient facts of history? The history majors first off in any history survey needs to know what sets their skill set aside from the answers they're going to get from the chatbot. I can ask it to tell me the key turning points in the American history survey. The history major hopefully has had some seminars where they learn what archives are, learn about all the things that haven't been digitized in archives, what the path to digitalization is.

[25:15] And then hopefully they've taken some higher level seminars where they think about how looking at the different archives influences your interpretation of history. Hopefully that freshman starts taking some courses where they learn about and experiment with the kinds of things that Loren Moulds has been teaching others here. And Mark Humphries and the other workshops where people have been getting their fingers into the LLM and uploading documents to a RAG and watching it create embeddings and make connections. So they've got an idea that they can be the curators of new knowledge. And hopefully they're reading some of the new scholarship that's come out about how history treats these documents differently, including Marnie's work again, of thinking about how each of the problems, the ethics of looking at an archive, zooming in and zooming out, is a

[26:16] form of historical thinking applied to AI. mean that they need to know how to code. They don't need transformer knowledge as detailed as that of the computer scientists or even of the most technical leaders in our field. They need to be able to have a general conversation about LLMs and archives. They are language models. They deal with the text in archives and increasingly with some of the spaces and some of the objects. But principally, they deal with archival material and they're reorienting and reconfiguring this. have known that we've been writing about a digital humanities for forever. But it needs to be core, front and center in their training, which means rethinking what is the first course that the freshman meets. Is it the American History Survey, just as a series of lectures about facts that you could look up, facts that you could synthesize? That used to be really impressive flexing, that I can synthesize 500 books into a survey. Now it's not so impressive.

[27:18] It's not so impressive. are doing newly with archival materials and what we have to say about archives and LLMs. That is the new skill set. Do you agree as a department chair? I have no idea why students take the classes they do. You know, if I knew the answer to that question, I'd be a very rich man. But here we are. You know, I think to answer, I mean, I do think that the short answer, Jo, is yes. I think this is a chance for history, along with all the disciplines, to be rethinking what's the purpose of a college education in 2026, right? The value proposition, as they say, the old ROI in the biz is changing. And I think that all departments and all universities need to think about that. I think history is better positioned than a lot of disciplines to actually respond affirmatively. And so I think we're actually, to echo Matt, I think really well positioned. And I think some of the things that annoy me is how defensive

[28:19] historians in the last 50 years have become, partly because of all the reasons that Matt laid out. We have good reason to be defensive. But it means that we're not looking for opportunities, right? We're not looking for the kinds of ways we can insert ourselves actively into the conversation. And I'm happy. I think we could talk more about that. To answer the question about what every student should know, I think one of the things I often think about is just how capacious history is and has always been. And so a student being trained in the 2030s may hit upon a topic for which they will only have 10 to 15 sources, seven of which are problematic, three of which are kind of quasi non-existent. And maybe it's in a language that is no longer spoken, that's very hard to actually learn at their university. And so, yes, we should be training students for this huge variety of study into the human past. Not every university can train students for all aspects of the human past.

[29:23] Why would we expect that? We don't do it now. We didn't do it when the departments were at their largest in the 1970s. We didn't manage to succeed at that. we do it now? And so I think the question of what every student should learn is in part how we currently think about what every student should learn. That is to say, not every student needs to know every language of all the sources. Not every student needs to learn how to engage with objects at the same level with how to engage with texts. And so I think when I think about what every student should learn, I think one of the big changes of the last 10 to 15 years is people need to be able to specify why computational methods are not useful for their project. That is essential. I think if you graduate and somebody asks you, why didn't you use these methods? And you're like, well, I don't know how to use them, or I don't think I need to use them. That is just an unacceptable answer at some point. You need to be able to specify why the methods you're using are appropriate to the questions you're asking. And I think if we think of normalizing new methods

[30:26] within this very old way of thinking about we have a limited number of tools with which to attack, you know, 10,000 years of people across the globe, I think it becomes a lot easier to specify what we want students to know, because we need to be able to ensure that when they're picking projects or not picking projects, when they're looking at one source base versus another source base, that they're able to articulate the reasons for those choices and the ways those choices affect the quality and limitations of the ultimate product they produce. So, you know, this isn't original, obviously, like a lot of us feel this way, but I think like in higher education, our role is to teach our students to teach themselves and to develop a lifelong level of learning. Like that is the most valuable thing that we can impart to them. And, you know, when you mentioned, Sarah, how like, you know, generating timelines, I could see how an AI can I'm not so sure, actually, because I think that's a perfect example of how you would think it would be straightforward.

[31:27] But I remember one of the first projects I did with the colleagues in the statistics department, I gave them like this collection of State Department cables, like millions of them. And I said, you know, I hear there's this thing called traffic analysis. Like it started out as an intelligence technique, you know, where if you can't read the communications, but you can analyze like how they're frequent and, you know, from and to and whatnot. you can begin to infer certain patterns and make out anomalies and so on. So I said, could we do something like that with the State Department cables? Basically doing what the NSA does to every other country, right? But we're going to do it now with the U.S. State Department. And my idea was that we would be able to potentially identify or at least develop a means of finding events. The ones we know about, yes, but maybe other things that we didn't know about. Because a lot of these communications were classified at the time. And so in order to do that work, typically you have to find ground truth, something they can test against. There is no ground truth. I kept looking, is there an authoritative data set about events in the

[32:31] 1970s? And I guess I shouldn't have been surprised, but no, of course there isn't, right? So that's just an example. And I looked at a lot of these timelines that other people had produced, and I was like, that doesn't make any sense. And, you know, in fact, I ended up asking Daniel Sargent, a historian at Berkeley, I thought we should have a human look at our results and compare it, you know, to the work of other humans. And asking Daniel, like, OK, you don't know these two timelines. One of them was generated by other humans. And this one was developed by a statistical model. We're not going to tell you which is which. And it turns out he was more likely to agree with the statistical model than he was with the work of other humans. this is an example like where like human judgment counts for a lot, you know, and, and there are a lot of cases like that. So what we ought to be trying to teach our students working alongside them to figure out everything these systems are capable of is to, you know, try to develop a sense of taste, you know, and, and judgment

[33:32] in, in how they evaluate the results and how they try to improve on them. Let's, let's turn to the concrete. what's the strategy? Not the long-term strategy, because those are easier to be very pie in the sky. What's the medium-term strategy? Okay. Can I start? We already started the debate in Discord. We got a set of questions this morning. Reading the abstract, somebody said, it sounds a lot like what people were trying to do in the 70s with Cliometrics. Why would you think you're going to succeed where they failed? And, you know, at first, Jo and I like, oh, now, now what are we going to do? Because we had an idea like, and you'll see it there, like we actually laid out work of another historian named Ben Breen, who said that as a profession, we should at least consider the possibility of setting really ambitious goals for

[34:35] ourselves. We're never going to have the millennium challenges where the mathematics community comes together and says, these are the hardest problems. We're going to give you a million dollars if you solve them. So history is not like that, right? But it could be there are collectively certain goals we might identify, things that we'd all of us really like to know, or at least a lot of us would like to know. And if we could collectively identify some of those really ambitious goals as a profession, then maybe we present a vision, right? And we give people an idea as to why it would be worth getting together, considering new methods, like developing new organizational structures and so on. But, you know, if you think about it, and I refer to Ben Schmidt, he wrote a really great piece about how historians really learn the wrong lessons, you know, as to what happened with Time on the Cross, as Jo mentioned. And they didn't lack vision, right?

[35:37] I mean, part of the problem was that they had this, you know, idea that every other historian was going to have to follow in their wake and do what they did. And historical, you know, and not just historians, right? We need archivists. We need librarians. of help if we're going to do any of the kinds of things I'm talking about. So the only way it's actually going to work, if it works, is if we do this collectively, like as a community, like if we can collectively come up with really ambitious ideas for what we want to do, ambitious enough that the rest of the world cares. Like, I don't think we're going to change people's minds about the value of our work unless we can point to really interesting, exciting possibilities. And so I think we should at least try that because, you know, I start from with, first off, you know, the recognition that what we're doing has not been working. Okay? I mean, like, we have to try new things because as a discipline, we're not in a good place. And if anything, archivists and librarians, they're even worse off than we are. Right? So that's why I think we have to at least consider the

[36:40] possibility of doing something bold and different. It's not going to be cliometrics. I think it could be a lot better. So I want to tell a little story about how things get changed in society. We've been in a period of time where, you know, change and hope, keywords under Obama, change seems stymied, different political realms. But I have been a historian of bureaucrats who changed small things that made a big difference. The people who brought you the public sidewalk. You don't have to pay when you cross the river. You don't have to pay when you cross the river by foot from London to Southwark. That happened. The idea of public housing, public parks, implementing these changes happened. And one of the things that became clear in my research about these shifts in the state was that this was all impossible until somebody came along and did a

[37:42] history of a problem. And it was always a history. Sometimes it was a survey. Sometimes there was statistics. The grandfather of them all is Sir John Sinclair's statistical survey of Scotland. Sinclair is a buddy of Adam Smith, and he has heard that there's an idea that England could send money to Scotland, and Scotland could be no longer mired in poverty as the result of being a colony colonized by England, policed by the military. So he says, I'm going to create the evidence bank that that's true. And he writes to every Presbyterian minister in Scotland and says, send me details. How many unemployed people are there? How many children are there? How many schools are there? And so then he has a portrait of the country. And then he writes a report to Parliament and says, you should really build roads to Scotland. Only don't send soldiers to police Scotland this time. Just build the roads and let us trade on them. So I can give you a bunch of other examples of a time that there's a history that

[38:44] says there are a lot of poor people, there are a lot of problems, let's go fix it. And we also know as historians of the modern period that there have been moments when that, plenty of moments when that broadening authority of the state has been abused. And it's been in the name of coercion and colonization and extraction. But, you know, the sidewalks are not extractive. The corrodes to Scotland were not extractive. We can build things that are hopey and changey. history first. And the history has to be detailed. It has to be based on facts. The facts have to be grounded in consensus. Everybody has to read that report and be like, good point. There are a lot of unemployed people in Scotland who could have jobs. I think I agree with you. So when I think about big historical projects, and I hear the mathematicians saying, we used to be a solitary discipline. We did problems by ourselves. And now because of AI, we can collaborate on international to solve these unsolvable problems. I hear, aha. So a bunch of years ago, I wrote a really blue skies version of

[39:47] this about the environment. What would it look like if we'd created a global index, a global history of climate harms? Jim Clifford is probably going to build a big chunk of that. He's got the British civil lists. He's mining them for toponyms. He's doing massive linked data. And he's going to come up with an LLM project that extracts all sorts of events of flood and wildfire and resistance across the history of British Empire. It's going to be incredible. And once he's doing that, look, Globalise has digitalized the Dutch East India Company. So we just extract all of those climate events. And then there are the Spanish records and then there are the Portuguese records. Now, Jim should not be running those parts of the project. British history. He's really smart and he's got a bot that can aggregate information about the secondary sources, but how much more efficient to have a Spanish trained historian doing the Spanish records part of that.

[40:48] So when we were talking about big overarching projects, I mean, we don't know exactly what it should look like, but my passion is climate change. I would love to see a global index of climate harms, who has gotten harmed, going right up to the present. I don't know how to build all of it, but I know somebody who's doing a good job in the 19th century. So I'd like to see all of us finding each other. So the Discord is open. We'd like you to vote on the two ideas that we put up, but we'd also like you to tell us a couple of phrases about what would be the biggest collaboration that you would want to be a part of it. It's not a commitment. It's just an experiment. And then we'd like to see you vote up each other's ideas. Yeah, I really like that one. That's a good one. Whether or not you want to be part of it, like, yeah, that makes sense. And the idea here is that some of us are trying to go to philanthropy, to college presidents, to deans, and say the discipline of history needs to be built up around

[41:49] digital tools. So not just my professorship, but a center that's going to fellows or invite a bunch of postdocs, give you compute support, give you everything you need. In order to make that happen, we need the best possible ideas about histories that everybody cares about. And is that it, that might be an insurance history of all the climate events where insurance had to withdraw. That might be a climate history that motivates regulation of the climate in a different way. It might be something totally different. We came up with some sort of random suggestions about the history of capitalism and intelligence because that's some of the literature we know, but they aren't necessarily the best questions. So we really need your intelligence in this document helping with that. So just to be extra clear, if you are on Discord, or the Discord for today, go into our plenary session.

[42:51] There are a few poll questions in there. I do see some typing happening and voting occurring, but please go in and weigh in. I'll take the opportunity while you all are distracted to make some comments that I can fly under the radar here. You know, I think the, what I was going to say is there is the, I think the CERN model, the national level model is an important one. And I think one of the things that historians have not done is collectively spoken as these are things that we all care about and you do too. And so join us. And I think we have a much easier argument than quarks and quasars when it comes to making those kinds of arguments, but we don't have the kind of shiny object or shiny and the trace object to kind of point to in the same way. And so it's important to do that. Those of us who work at very tech heavy universities always lament that the smallest increase in knowledge in some

[43:55] arcane place is broadcast far and wide. And then somebody who publishes a seminal book that's transformed a field, but that doesn't have that kind of sheen of technical advance to it, we just kind of bury as one more thing that is going on in the humanities and social sciences. And so changing that dynamic, I think, is really critical to changing the outcome compared to, say, earlier iterations. I think one, from my perspective, one thing I just always am keeping in mind is our colleagues, right? Our colleagues who have no interest in these methods, our colleagues who are suspicious of novelty in almost any form, our colleagues who are asking, what does this have to do with my field? And I think one of the medium-term goals, and I mentioned this yesterday when we talked about graduate training on another panel, I think one of the important medium-term goals is to normalize technological, methodological innovation as something that affects everyone.

[44:58] So if you're not using a particular tool, if you don't know how to read Russian, if you're not using certain kinds of archives, there's a cost to the kinds of things you can do as a historian. And we need to normalize some of these new technologies similarly as having a cost to people who choose not to use them and to try and articulate that so that instead of it being like, why doesn't my colleague take a class in regression? Because there's all kinds of interesting economic things you could do with that. It might be instead articulating, this is how your project actually might be continued or advanced or challenged or deepened if you engage this other type of sources. So kind of meeting people where of normalizing the fact that everyone uses computers, as Sarah already mentioned, everyone uses databases, everyone's searching. The search engines now are almost all built on some new AI under the hood, unless you go in and find the toggle to turn it off, which turns out to be

[45:58] extremely difficult to do for most of these search engines. And so, you know, I think normalizing that with our colleagues will go a long way to actually this medium-term goal of having moonshot style, things we can seek out, and then on the ground, just saying that all of us are engaged in using technology. And the question is, what are the affordances and problems that come up? Can I piggyback on that? So Chris mentioned the CERN model, because a lot of us, and Jo especially, have been active in trying to build a constituency behind something like a National Institute for History. And I happened to work with somebody who spent 15 years at the NIH, Okay, well, here are a few facts that might be helpful. You know, the annual budget is $48 billion. So 80% of biomedical researchers in the U.S. have funding from the NIH, right? How many of us have had funding from the NIH, okay? I mean, I've been lucky, but I knew how lucky I was, right?

[47:01] It's really hard to get funding from the NIH. so all of that funding is basically given out through peer review right so biomedical researchers reviewing the work of other biomedical researchers and deciding what what kinds of projects are most promising right and then finally the nih has an enormous budget for infrastructure and so like GenBank right like all the genomic data for example that people use like imagine if we had resources, even a tenth of that, right? I mean, even, you know, the National Archives budget, the entire budget is 1%. It's about $400 million, right? So just imagine we had even a fraction of that, like to support the work of historians, archivists, and librarians to protect, preserve our collective past so that we have an authoritative record of ourselves as a species, right? I think you're right. I think we can make that case. And why wouldn't we? I mean, has 30 different centers. We should grow the pie. I mean, if somebody comes along and says, let's have a

[48:04] National Institute for History, you should want a piece of that action, right? But you should get behind it too, right? Because I think all of us would gain from that, not just historians, archivists, librarians, et cetera. As a species, I think we need that more than ever, because as we all know, it's going to become increasingly challenging to preserve a record of the past in a way that people will recognize and trust. So please vote for it. It's in the Discord. We've asked everyone to vote for that, if you could. Our conversation this morning, given we have only an hour, we have focused our conversation on historians who are working in university contexts. The questions, the problems, the challenges, the opportunities look very different in some of the other areas and professions in which historians are working in

[49:05] archives, museums. I came to the AHA from a museum. So we also need to be thinking about how our discipline is shaped in relationship to how all of these areas in which historians are working are interconnected. interconnected. And certainly what I've heard today is the kinds of interactions that are happening, and from previous conversations yesterday too, the kinds of conversations that we have that are ongoing between historians in university departments, in archives, in museums, building systems, and thinking about how to better use the material that we have to understand the past. So as we get ready to move into Q&A, I hope that those of you who are working in all sorts of professions will kind of weigh in on some of the things that you see as central to

[50:07] this question of the discipline that we need to build. But I think first we want to have Loren share the results of the polling. Is that right? Loren or Louis? Oh, okay. Okay. Yes. So, hello, everyone. I used Claude to help me. Sorry. But there are a couple of questions that I've summarized and has been actually as to where we ended up at the last part of this was really, really great conversation. so here's a here's one i wanted to start with um bilia uh commented that they doubt a topical moonshot could appeal

[51:08] across political camps and subfields and methods but they see archives as data as more promising built through smaller steps like summer workshops um would an archives as data moonshot built through smaller efforts be more achievable than a topical uh that's an neh funded project archives this data and i want to thank the american historical association as being one of the plaintiffs in the suit you know that restored funding for that and about 3 000 other neh grants so go aha um and yeah so um so i'll put something in the discord anybody who's interested in that if you go to History Lab's website at columbia you'll see that it was canceled. It's coming back. It's going to be next June. And, you know, so I just put out this idea, like, because we had two years to think about it, you know, we're thinking we might do it a little differently. We're not going to run it the same way that we used to run THATCamps, you know, or other kinds of digital humanities projects. And one idea is like learn by doing, like, you know, that I and others, you know,

[52:12] would, like in a studio-like format, kind of work alongside other participants in taking on hard problems. And then, of course, part of the time would be devoted to people working on their own projects. But I thought, you know, let's, in the two weeks, like, let's take on a few hard problems and see what we can do with them collectively. So that's the idea. If anybody's interested in that, you know, please reach out to us. I do have the results. I do have the results then now of the questions. And maybe I'll just, oh, sorry. My internet's working, I guess. go. The question one, which candidate questions do you think are worthwhile for the historic moonshots? We have four for history of extinction, 23 for archival silences, 12 for where capitalists, what capitalism has emerged and how it spread, and then 12 for history of science across empires. So a pretty good spread, but much more towards archival silences and concealed

[53:14] records. the next one um should we organize summer institutes like archives as data that would be 33 yes and one no so i think that's resounding um do we need something like a national institute for history to organize and support ambitious collaborative research projects that's 28 yes six not sure and one no so there you go it's pretty good i thought you're gonna need air on that particular calculation. I thought we'd have 100, not sure. All right. We have a few minutes at least, so I'd like to open up for some questions and comments. um as a phd student in exactly the position you talk about namely not speaking

[54:15] russian but being able to process russian documents that one of my committee members a russian speaker gave me like 500 pages of duma transcripts um confirming every quote that i plan to use with a russian speaker wary of disclosing this in research statements and job market materials, or even though it should be a value added, it should be a positive thing. Look at this tool I've used. It seems like they'd say, oh, you had ChatGPT do your archival work. You had AI replace what should have been the core work of a historian. Just learn Russian. It's not that hard. I'm not sure someone would say that, but similarly, I'm pretty much only mentioning my LLM methods for programs that I know have a digital humanities center or something that could be a possible home for this or non-history So I'm kind of in between history and IR. Other than just being courageous and hoping it works out, do you have thoughts on how to market oneself as someone on the job market wary of being sort of given the kiss of death on this? I mean, one of the strategies that many digital humanists have used is just to publish one account of their methods and then separately to

[55:17] publish their findings. So you can have a completely readable, lucid, narrative-driven history of a period or a theme. that there are sources and I found those sources. I don't need to tell you how. Maybe there's a scholarly appendix where I walk through some of the influences and I just name some of the methods. Ask me for more. And then in Journal of Cultural Analytics or Journal of Digital History, you can publish about the methods and show in detail rigorously what I did up to the standard practitioners in the field. So in LLM practice, we would want to see verification. Sorry, I should also say historical methods. Historical methods is in the room. Fantastic place to publish DH research. Yeah. So doing some of the validation steps that Mark Humphries was talking about yesterday on that research could be a point for innovating and showing that you have a more robust way of making sure that you're getting the right inferences. And that's what the methods articles are for.

[56:19] So don't be intimidated. Just if you do methods, do it right. You peer review it, you publish on the methods, you think about it, you think critically, you try to advance methods, you know what other people have done. It's a methods article. It's not a foreign thing. These are methods journals curated by other historians. Our standard isn't you're a computer scientist. Our standard is you're advancing the field. I want to come back to what Chris was talking and that is somewhat of a PR problem of history and that is there is a significant gap between what we know that we do and our methods and how we do history and what history worth and what everybody else thinks that history is including our deans, including the parents of students who tell them not to major in history or history of science because there's no jobs, and including tech companies that are very interested in the questions of humanity, but they will ask the philosophers to come and advise them,

[57:21] not historians, because history is long, long, long and boring. And if we do disclose all of our methods, then it becomes kind of tedious. If we don't disclose our methods, then it's like, well, historians just make up stories, right? So there is a problem here. But then there is another problem within the profession when we are trying to do public-facing work and explain what is the worth of what we're doing. It's really good for those of us who have tenure. It's not good for those without tenure because this work does not count for tenure files and sometimes even can be backfiring. So what are the structural elements in the profession that we have that maybe is for the AHA to handle, that we can also encourage public-facing work that will promote not just methods, digital methods or computational methods, but the worth of history itself

[58:22] and kind of improve the image of history. The AI community is really hungry right now for new accounts of what AI research, robust AI research looks like. Because there are all of these computer scientists endlessly chasing incremental innovations and performance, but they don't know anything about what documents are or what the real world is. The problem of industry is always to, is always concerns the real world. hurricanes coming, we need to know about how insurance has responded to incidents in the past. So our knowledge grounded in reality, grounded in real history, grounded in archives, is very relevant. So one of the things that some of our colleagues in the Committee of History, the Committee of Historians for AI and Research are doing is they're writing op -eds. Shadi Bartsch, her colleague at the University of Chicago, has been writing a series She has a column in the Chicago Tribune about AI and

[59:25] liberal arts disciplines, about how history in particular is valuable at this point in time. But this op-ed writing, writing on Substack, joining AI hosts and podcasts, you know, that's something that anybody in this room can participate in. The Substack is free. As soon as you start describing new methods and problems, people interested in AI find you. Boy, do they find you. They find you really quickly. to join a conversation in the AI world. So there has to be more of that. And I think it's important that we advertise that. You know, the AI major is swad over by data science and computer science. So maybe there has to be an AI and history minor, an AI humanities minor. But we can create those tracks. You start with some courses that have methods names. It's fairly easy to switch that. There are small things anybody can do in any department to just pick up on some of this and advertise, look, my research and the research of all of these other people

[1:00:27] is concerned with the facts of where the history lives. And some of the material of that course is, a lot of the material of that course is going to be the stuff that we know backwards and forwards. So where does the material live? How do you get the facts right? What does it look like when there's a historical debate? How does history help us arrive at consensus? And then what are the historians saying now about the machines who were doing history on our behalf? What archives they have access to? What are the challenges? What are the privacy copyright challenges? And what is it that historical thinking brings to this process? So Louis has given me the watch tap, but are you actually wearing a watch? Yep, he was wearing a watch. So, but before we do that, the challenges that you have identified that go far beyond any particular field of history are ones that the AHA is very much working on. And we have a number of resources that are out there.

[1:01:28] And those include, we just published a newly revised Careers for History Majors booklet that is geared for students and parents, also academic advisors. And it really boils down the kinds of skills and ways of thinking that younger historians, BAs, are developing and shows the vast array of careers into which they can go. We have that. We have also issued a broadening the definition of historical scholarship set of guidelines that is encouraging history departments to rethink their tenure and promotion. and that was done in 2022 and we are now seeing those guidelines being being incorporated

[1:02:29] into TNP guidelines in departments and the first folks who are getting tenure through that new process has has begun and so we're collecting guidelines that have been revised and are able to help connect people to one another, to talk through what they did, how it worked, those sorts of things. And then more generally, also trying to do more public outreach about history. We just did a big national weekend of programming called We Want More History that was in, we ended up having the participation of 39 states and over 250 events. And there was no agenda other than for people to come together and do something around history. So there were scavenger hunts and some trivia nights, as well as

[1:03:32] museum exhibitions and student work on display, really cool things. So we are, we're looking at all of the kind of the whole ecosystem in which history exists so that we can make sure that we are supporting K -12 teachers who are helping to inspire students to go into college and want to take history courses, even if they don't want to major in history, trying to develop the tools and resources that our colleagues in departments and universities need to support the students and then also really thinking about the ways in which we can support the entirety of professions in which we have historians working. So lots of things going on. So I hope that can end on a hopeful note. And thank you all for being here and enjoy the break.