Embedded context and state in present and past writings

We embed context (the state) in everything we say and write, and that makes history really hard

I’m currently in the midst of reading Empire of Guns: The Violent Making of the Industrial Revolution by Priya Satia. The thesis of the book is that the role of war in Britain’s industrialization has been understated and was not incidental, and that the English, and then British, state from the 1680s to the 1850s drove business expansion and experimentation because it was consuming vast quantities of guns and other items necessary for making war.

There was one passage is particular on the risks of the seaborne gun trade in the 18th century that I thought was fascinating:

Why all this doom and gloom when business was brisk? Eighteenth-century trade was bedeviled by risk - insecure sea routes, perpetual doubt whether a cargo would reach its destination - and the Africa trade was particularly unpredictable. Galton and Farmer bought shares in the Alexander; the ship was lost

[..]

And guns were perishable. That they became useless within a year in Africa was partly what drove the insistent demand for them, but perishability was also a major liability in a time of slow and risky commerce.

[..]

But more than any of these factors - perishability, risks at sea - the real difficulty in the commercial trade was long credit. Cash shortages were endemic in the eighteenth-century business; credit was unavoidable. But in the gun trade, as Farmer lamented, “payments are so long its an article hardly worth following.” - Empire of Guns, page 54-55

There are three risks highlighted in those paragraphs for the gun trade that don’t appear to have significant modern parallels:

1. Ships would often sink

2. Guns would only last a year in Africa before the wooden stocks rotted and the barrels rusted

3. Businesses would often not get paid in a timely fashion

I might be naive, but when a business orders something and it is put in a container to cross the ocean, it general gets to the location. Sure, sometimes they don’t - the container might be lost or stolen at the port, or the goods aren’t useful by the time they get there - but its not often that ships sink, and when they do it makes the news.

Of course, I knew that the insurance business grew out of managing the risk of ships sinking with Lloyd’s starting out as a coffee shop that merchants hung out at. But what I hadn’t considered was that those sort of risks are part of the context of the 18th century, what computer scientists might call the state1. Of course as well there are good reasons why the state has changed - insurance, better mapping of sea routes, coal, then oil or diesel fired engines shortening trip time and reducing variability, Paces Britannica et Americana, etc.

What I thought was interesting though is that my expectation of the state that 18th century merchants were working with, things I never really thought about and just assumed had parallels today

Did you catch the game last night?

Let’s think of another example where state comes into play:

A: Good morning! Did you catch the game last night?
B: Oh, what a nail-biter. They just fell apart, didn’t they?

Let’s say this conversation happens on October 9, 2022. What state does B already know? I’ll give away the ending that A is talking about the match between the Toronto Blue Jays and the Seattle Mariners on October 8, 2022. To answer as B did, B must know that:

1. There was a game of baseball the night before

2. Who won the game

3. What happened during the game

4. That A knows the above.

That’s just surface level stuff though. Let’s say there’s an alien the size of a fly on the wall beside the water cooler hearing that conversation. What do they need to know to understand that?

1. That there was a game of baseball the night before

2. What baseball is

3. That people “have” teams

4. That A and B “have” the same team

5. But even knowing what baseball is isn’t good enough. Thinking the game was a little league game doesn’t really give the feeling of nail-biting. It was a playoff game! What the heck are playoffs?

6. That “your team” is playing

7. That if Toronto lost that game, they’d be kicked out of the playoffs, and therefore the end of the season

8. That middle age men play a kid’s game for lots of money and we’re okay with this

9. That some people, like A and B, take this very seriously

10. That 50,000 people or so were in the stands watching these men play this game

11. Why a Canadian baseball team is in the American League of Major League Baseball

I could go on, but there’s a lot encoded in that!

But the conversation acts as signals as well, doesn’t it, and exchanges even more information than you think. I was reading Scott Alexander’s The Phatic and the Anti-Inductive today about how small talk acts as signals

The classic example is small talk. “Hey.” “Hey.” “How are you?” Fine, and you?” “Fine.” No information has been exchanged. Even if the person involved wasn’t fine, they’d still say fine. Indeed, at least in this country giving an information-bearing response to “how are you?” is a mild social faux pas.

Some people call this “social grooming behavior” and it makes sense. It’s just a way of saying “Hello, I acknowledge you and still consider you an acquaintance. There’s nothing wrong between us. Carry on.” That you are willing to spend ten seconds holding a useless conversation with them signals this just fine.

A asking B about the game last night is just like that example of small task. It is both a prompt that says:

1. I think you’re up to date on current events about this children’s game and I want you to prove it

2. I take children’s games very seriously

We can probably go further and say that its also signalling that A is a “bro”, and B’s answers confirm those things (as like a networking handshake). We can probably reduce the conversation to:

A: Do you know the state?
B: Yes, I know the state

How would an AI interpret this? How is an AI supposed to know what someone’s team is? How is an AI supposed to even know which game someone is talking about? Two complete strangers in a bar might be able to have the above conversation about the game last night without ever discussing who has what team, based on location, what they’re wearing, etc. At the same time, this sort of state can be really hard for some people to understand. Shared context is also occasionally highly local: If I was in Europe and someone asked me if I saw the match last night, I wouldn’t know how to answer that.

Sarah drove to work

Another example: Let’s say I live in the present day and am writing a novel about a woman named Sarah. Sarah has a job in an office that she drives to everyday. Let’s say I want to start a chapter by having her drive to work, but its just a minor introductory point so we don’t want to go overboard. How would I go about writing that? “Okay”, you say, “that’s easy”:

Sarah drove to work.

See? Nice and easy. But what does that even mean? Let’s examine the state that’s built into that statement

1. Sarah owns a car

2. What a car is

3. What people’s cars look like

4. That cars burn gasoline

5. That to drive, you get in, put the key in the slot, turn the key and push on the peddles

6. That people have jobs

7. That people driving to their jobs is ordinary and not worth spending time on

8. That driving in general is a thing that people do

9. That Sarah probably traveled over 5 kilometers from her home to go to work

10. That Sarah will park her car in a parking spot in a lot or a garage somewhere

This is a very regular occurrence that will occur hundreds of times per year for some people. We don’t need to explain any more about it. If anything, going into more detail would be weird and a very boring novel to read. In fact, this novel exists in the Mezzanine, by Nicholson Baker, notable for having a 1000 word footnote on paper vs plastic straws (in favour of paper!). It’s boring, its mundane, and that’s the point. No one writes like that unless its to make a statement.

The state built into driving to work is so mundane that no one would ever describe it in detail. But wait, I slipped in something up there that doesn’t describe driving as I know it today: Many modern cars don’t even have keys anymore, instead they start by button press. My son is growing up in a world without keys starting cars, will he read it as Sarah using a key or a button press?

Let’s say that instead this book is set in the 18th century, and instead of driving to work Sarah took a litter. We get a sentence like this:

Sarah took a litter to work.

Okay, so what’s a litter? They don’t really exist anymore, so without googling the work of other historians (or having previously googled it), most of us probably don’t have that context. Here’s a painting of a litter:

What’s the big difference hidden in the sentence about taking a litter? There are other people who are implicitly there, carrying her! Those people are completely erased. The modern reader will also not appreciate that it’s probably pretty bumpy too, and in the summer its probably pretty hot without air conditioning.

The state mutates

And thus we get to the point of this: there is significant state hidden in almost everything we do, and that makes doing history really hard. Writers assume so much about what their readers already know. They did it in the past, they do it today, and they will do it in the future, because explaining every detail is incredibly boring. Our state evolves over time, and we don’t notice it. This is also, in my opinion, what makes artifical intelligence so hard: there’s an incredible amount of the human existence which is simply too obvious to talk about, so how would an AI learn it?

I’m using the word state here purposefully as well: A big part of programming is managing the state of the program. The purpose of a line of code, a function, an app, etc is, reductively, to take some data, transform it, and output it. The code needs to do different things at different times to that data. One of the areas where bugs creep in is where the code assumes certain things about the data, or the state of the data during the transformation, and is wrong. There are parallels here in every day life.

So what will 25th century readers understand about Sarah driving to work? Will they think she used a key or a button? Does it matter? Maybe not, that’s probably a pretty minor example, but its an example of how state shifts over time. With enough mutations, the state becomes unrecognizable, and thus our understanding of the past becomes based on the wrong state.

The words on the page are the same, but they’re only half the battle. To truly understand the past, we need to understand the state and what the writer was assuming about the reader as well. A good historian will explain that to their readers, like Satia did in Empire of Guns, but the pernicious part is even realizing your context is different.

1. State as in “status”, not “government”