Sadly, the historian in the future will have it tough.
The big challenge in archival theory & practice atm is that people & organisations are no longer creating a 'mountain of human-curated documents', and largely haven't been for a decade or more. Just a mountain of documents...
This a problem for archivists atm because it breaks their model of appraisal and transfer (part of your 'human-curated' bit).
It's a problem for historians of the future because the structure you expect (human-curated again) won't exist & there'll be orders of magnitude more dross...
Archivists are hoping (like kids believing in fairies) that LLMs like chatGPT will solve the problem of catagorising (hence organising and weeding) these piles of steaming documents.
And the same will be true for users (historians) of the future
What's so goofy is that people sometimes talk as if they forgot how much we use our bodies and interaction with the world--our senses, our ability to initiate responses to find things out, etc.--to find out everything. Like we spent too much time online and then we just FORGOT how we find out shit.
Strikes me that the real question is whether historians can just get by with repeating their culture’s conventional narrative (parroting via LLMs) or whether they have a responsibility to re-examine historical records with fresh eyes. Being an apologist for hegemonic ideas vs actual re-examination.
I had a conversation with a great undergrad who explained that an LLM had been very helpful at finding sources on a particular topic. The problem was, he was asking the *wrong question* for the topic he was really interested in, and the sources sent him on a time-wasting hunt for irrelevant details.
He couldn’t know that, because he’s still learning about history. Someone with discernment has to tell him to shift course, otherwise he’ll dig deeper and deeper to learn about the wrong thing.
I often think of a historian who told me about looking at source after source for evidence of a particular phenomenon. Only after looking at many documents was she able to grasp that the phenomenon was entirely different in nature than she’d thought.
OpenAI wants to offer a 20.000 a month PhD level ChatGPT. Aside from the fact that you can hire 10 people with PhDs for that money, it strikes me as either misunderstanding what a PhD is (you know, ORIGINAL thought) or wanting to build sentient AI.
When I was an undergrad I had part-time jobs in the university libraries. One of my professors said I really knew how to use one. My later grad research required online searches and my earlier experiences helped. I wonder how many students understand how to ask the right questions.
Very right. Scientiest researche and put their own thought into their work. AI only conects possible logical points. Think of the results of automatic correction of your SMSs.
Yes! I still remember the spouse of fellow grad student who was absolutely boggled to learn that historians go to physical archives for the simple reason that those documents simply don’t exist elsewhere, so “looking up the information” isn’t even possible until a historian finds it and shares it
My father is a historian; professor emeritus. His dinner-time guidance on the meanings of historical analyses, the nature of historic information, and our family's recognition of the meaning of our history instilled in me deep respect for "what a historian *does*." Many, many thanks for sharing.
I shouldn't be, but I am always shocked when something happens to remind me of just how much people really don't understand what doing history involves (of course, the trouble is that they believe that they do understand).
How many profs teach what doing history is in classes 2000 level or below?
In my experience it’s very few. We don’t differentiate between the past and history until far too late. They never know that history is a process or method.
I certainly got that lesson very early in college (which is why I became a history major!) but I don’t think I encountered a primary source before that. I do know a lot of K12 teachers think deeply about this, but I don’t envy them the challenge of teaching basic facts *and* what history really is.
It’s weird though, because we do it in reverse. We teach them facts and then—maybe—teach them history but doing so requires the process of rewriting what a fact even might be. We teach the past without any of the philosophy behind it. Imagine teaching science without the scientific method. Oh wait.
If AI were utilized correctly, could it be valuable for finding those overlooked inputs to help the historian? E.g., combing through more volumes than humanly possible?
(as you noted, this requires asking the right ? and contextualizing the results)
I have colleagues who’ve used it in just that way: to surface trends and people we haven’t thought to look for. The problem is, knowing how to do that and what you might find requires deep familiarity with the contents of the archive. And it also requires knowledge that the archive is assembled +
by human beings, with the same prejudices and blind spots that everyone has. Datasets arguably more so, since they are more rigid and inherently selective. Saidiya Hartman’s latest book is, among other things, about how you need to read between the lines of an archive in order to find truth.
It's a bigger problem - there's been a lot of attention on making sure everyone can check whether scientific or medical advice is coming from, but there doesn't seem to be a general sense that there are professional reputable historians distinct from Bloke With A YouTube Channel.
And that's how you end up with Joe Rogan et al given space to idiotic on-line "history buffs" when (I would hope) they would entertain any old random alternative medicine twit.
The news about WHS has reminded me that my dad bought a hardback copy of AJP Taylor's "English History 1914-45" in there circa 1980. There was a time when historians could be household names, in at least some households, and be respected as much as the celebrity scientists.
Comments
The big challenge in archival theory & practice atm is that people & organisations are no longer creating a 'mountain of human-curated documents', and largely haven't been for a decade or more. Just a mountain of documents...
It's a problem for historians of the future because the structure you expect (human-curated again) won't exist & there'll be orders of magnitude more dross...
And the same will be true for users (historians) of the future
It's also very rarely explained
As a kid and early teen I was totally passionate about history and read tons of it. But had no idea what you'd be doing as a historian.
Vagaries of life led me to study history in university and then I understood (and was a historian for some part of my life)
Just from a humble engineer's point of view......
I know these people. I need to sell some of this garbage. They haven't spoken to anyone outside of tech since their last undergrad elective.
In my experience it’s very few. We don’t differentiate between the past and history until far too late. They never know that history is a process or method.
If AI were utilized correctly, could it be valuable for finding those overlooked inputs to help the historian? E.g., combing through more volumes than humanly possible?
(as you noted, this requires asking the right ? and contextualizing the results)
https://discontinuednotes.com/2023/12/22/popular-history/
LLMs were critical in me grasping new concepts. "Explain 'X' like I'm a grade school student, now high School, and now like I'm a college student."
Now I have stats, math, python, & modeling down!