These past couple of weeks I've been working on a game about drawing, in the spirit of "life games" from my previous post. The summary of the thought process is that there aren't really a lot of new games that would be somehow connected to real life skills, e.g. drawing. Sure, there are games like Drawful, but they don't really have that much to do with the actual skill of drawing.

I don't feel like writing as much as I did in the life games post, so this one will mostly have photos of random sketches and iterations I've worked on so far. I started with "paper prototypes", meaning prototypes of video games made with physical components. For the pieces like dice and whatnot I borrowed a couple of games from the library and used theirs.

The process started with some random scribbles and thoughts on various pieces of papers. Those papers also got used for my ham radio course prep, so ignore the math.

One thing that I really liked about the Dynamic Sketching course was how heavily it was based on drawing from reference as a way to improve. I think it's central to improving, so I wanted the game to have those aspects as well. The idea behind these prototypes was to make a tycoon game, where you would run an art studio, and each month you get a number of jobs, such as draw a trumpet or a mountain lion.

Then you do the drawing in the given time limit, e.g. 10 minutes, and you submit it to the client for review. The client in these prototypes was an LLM that evaluated the drawing on various properties, depending on the iteration.

I thought the time limit would make the drawing process nicer, because you sort of have something else to blame rather than your drawing skill when a drawing doesn't turn out the way you wanted it to. This effect works well in other games. But here, the drawings I made were bad, and the constraints I was working under didn't make me feel any better.

Here are some rules from one of the first iterations.

Unfortunately these iterations never really hit the mark. I felt that in 10 minutes I really couldn't draw well enough, and drawing a whole object or animal was just too much. So I figured I'd break it down into smaller pieces, and use some of the exercises from the first weeks of Dynamic Sketching as game elements. To do that, I made a quick prototype in Godot where you draw lines between points and get scored based on that. I imagined that this would then evolve into more complex exercises like different shapes and forms and eventually whole objects.

The lines would also slowly fade away, because I was thinking that in the first paper prototypes the issue wasn't really that I was drawing badly – that's expected – but that I was so self-critical during the process of doing so. I figured that if whatever I drew wasn't permanently visible to me, maybe there would be less pressure.

This concept never really materialized into anything fun either. So I just started sketching some thoughts on paper over the next few days. A big theme in these thoughts was figuring out why do I even want to draw in the first place. I realized that understanding things, phenomena and objects in the real world is a massive part of it.

So I started thinking about a game I could make around the concept of understanding things through drawing, rather than just drawing to create images.

For this, I made a quick web app prototype, where I could take a photo of something, and an LLM looked at the photo. It then provides me with questions that I need to answer about the object by way of drawing. It then evaluates my drawings to see if what I drew answered the question adequately and with high enough quality.

I also thought it would be cool if all of your drawings would end up in a sort of mind map as you answer questions.

The type of game I envisioned was something in the incremental games genre, that includes games such as Cookie Clicker. However, I then started thinking about why I want the game to exist in the first place. I want the game to reduce stress in drawing and inspire me to draw more, even when I'm not playing the game. The issue with that is that if you have something that you have intrinsic motivation for, and you give a strong, explicit reward or score for it, as incremental games do, what ends up happening is the overjustification effect, where you lose that intrinsic motivation to do the thing outside of the game, and now only value the extrinsic reward given by the game itself. So basically I'd end up in a situation where I could feel motivated to draw when I play the game, but not outside of it.

So I started thinking about how to circumvent that psychological effect to make a game that would boost intrinsic motivation instead.

It turns out, this is pretty difficult to do, because games are often characterized by an explicit and extrinsic reward system. However, toys are not, and there are many games that could be considered toys, like The Sims and Minecraft's creative mode. In these games you don't really have goals, so you just play with something to have fun. These are usually intrinsically motivating, because there's nothing external telling you what is winning and what is valuable.

What I ended up with is the first thing that actually felt like it hit something in my brain. I took the earlier ideas of taking a photo of an object, and giving it to an LLM. However, this time the LLM was instructed to act as a 12 year old child that knew nothing an average 12 year old wouldn't know, and was tasked with asking a question about the thing photographed. This was inspired by the curious questions Leonardo da Vinci would ask about the most mundane things and his notebooks that were a result of this process, as well as what is dubbed the Feynman Technique, where you learn things by trying to explain them to a child. After all, understanding and learning was what I identified as being important to me in drawing.

Instead of the LLM scoring drawings as pass or fail for an explanation, which would infringe on both autonomy and competence in self-determination theory, which in part explains why explicit scoring systems cause the overjustification effect and a decrease in intrinsic motivation, the player now decides themselves whether they have succeeded. Instead of a score, the LLM responds to a drawing with their own explanation of how they understood the drawing, and if it matches what the artist meant to convey, then the drawing did its job! I'm actually pretty happy about this realization. As an additional mechanic, whenever you teach the LLM something, I can still add it to a graph, and as you teach more and more things, you can refer to your previous explanations, as the LLM already "knows" these things, so you get a nice feeling of progression.

This worked surprisingly well! I gave the LLM this photo of an ant:

And it responded with "Is that actually its eye? It looks like a dark circle on the side of its head, but I can't tell if it can see in every direction or just to the side." This lead to an hours long exploration into the eye structure of ants and what do they actually see, and it was super fascinating. Exactly the kind of feeling I wanted to have from the game!

I haven't made the drawing yet, but I've made some preliminary sketches as I've done research.

So we'll see where this goes.

Projects