A Prototype for a Game about Pack Peddling through West Virginia

This is just proof of concept, so don't expect it to be fast or particularly good.

I have a lot of research that didn’t make it into the book, particularly about pack peddlers in West Virginia and West Virginia history at the start of the 20th Century. And I have a weird inability to let research go to waste. (Of course, it wasn’t REALLY wasted, since it gave me the context I needed to write what was in the book. But you know what I mean.) So I had to find something to do with it.

Also, some of you know, I have launched a minor at my university in Narrative Design. (Narrative design is the process of structuring and shaping a story within an interactive medium, like a video game, to create a cohesive and engaging player experience.) In order to keep up with the industry, I’ve here and there been building little games using tools like Ink and Twine, but nothing with much value beyond demonstrating techniques and skills to students in my classes. But it was clear it was time for mean to learn about the way AI could be deployed to build games.

So, Research About Pack Peddling in WV + The Need to Understand AI Game Building Tools = Pack Peddler: An Infinite Adventure in 1905 West Virginia.

It’s slow and it’s not particularly good… it isn’t actually a game as much as it is a proof of concept for a game. The way it works is that the LLM (Large Language Model) has been trained on my research and then procedurally generates a never-ending adventure in which you play as a 15 year old Jewish immigrant from Lithuania. (Sound familiar?)

It took several days of training the LLM to get it to stop making half the encounters about antisemitism—even though my research suggests that the peddlers were a welcome and necessary part of life in rural Appalachia before the railroad came in—and another several days to weed out the stereotypes embedded in its larger data set about Appalachians. And even so, both things pop up with less regularity than I’d like on my test playthroughs. (This is something we need to always be aware of when we are working with LLMs. The biases and prejudices of the Internet are baked into them because they’ve been trained on so much online content.)

This won’t go any farther. I’m against using AI for creative work in almost all instances (the exception, as shown here, is that I’m kind of okay with it for prototypes that are never going to market). But it was a fun project, and I thought you might like to see it. If you do play Pack Peddler: An Infinite Adventure in 1905 West Virginia you’ll quickly find its limitations: it’s slow, it crashes, and its more than a little hackneyed. Mostly, it’s sort of like a casserole: I had a lot of leftovers and I had to put them into something.

What do you do with your leftover research, friends?

Sarah