awarm.spacenewsletter | fast | slow

Difficulty modes and learning

Two of my favorite games of the last couple years, Celeste and Hades, have had very well considered difficulty systems.

Changing the difficulty is nothing new in games, most have some flavor of "Easy/Medium/Hard". But these two games take a more nuanced approach, that informs the player, and gives them control to shape their experience. And, remarkably, they do so without detracting from the core experiences the game designers want to create.

Celeste is a precision platformer. You've got to run, jump, and dash, through each puzzle-like level, and every misstep means you restart. Difficulty settings would make very little sense in this context. Instead, the game gives you "Assist Mode" which let's you introduce tweaks to the games rules, like making the player invincible, or giving you an extra jump, or just slowing down time so you can breathe and think.

Hades takes a different approach. It's got two main difficulty systems. One comes after you've beat the game once. Because it's a rogue-like, every run of the game is different, and you're meant to keep playing to explore the world and story. After that first win, you unlock "Heat", an array of different challenges you can add to the game to make it harder and increase loot.

On top of that, the game let's you at any time turn on "God Mode". Contrary to the name, this doesn't mean you can't die. Instead, any time you do, the game reduces the damage you take a bit, so your next run becomes a bit easier. Eventually you reach a kind of equilibrium, and when you improve, you can always turn it off.

Why these systems matter

Difficulty is fundamentally about player skill. A game that's "easy" for someone who has years of experience with it's fundamental skills could be crushingly difficult to someone new to it. Games often have baked in assumptions about the players starting skill level, that they can fully use the input mechanisms, or have played similar games in the past, but this doesn't match with reality. Players are coming with a massive diversity of experiences, that fundamentally changes how they experience the game. This poses a problem for a game designer. They want to create a consistent experience for their players, but the players themselves vary.

Designers have different tools to deal with this problem. They can design the mechanisms of the game itself to be more or less forgiving of different abilities, or to have a different learning curve. Or they can have multiple interacting systems, so that different players can use different means to progress.

The difficulty systems in Hades and Celeste are so interesting because they offload this problem onto the player, the entity who has the most knowledge of their experience and background. It means they can get a lot of impact out of relatively simple systems. One of the developers of Celeste said that it only took a couple days to implement assist mode, and it's had a massive impact on how the game is percieved and played by many.

Okay but what does any of this have to do with learning?

I think we're faced a similar problem as game designers, when we're trying to construct learning experiences. We need to construct an experience that takes into account an individual's personal abilities and experience. This is especially true when you construct, or modify, a learning experience for yourself.

As learners we're often lacking a strong notion of how hard something should be, or even of what aspects of the learning process are making things hard.

The tricky bit is difficulty is an emergenet property from our previous experiences, and the systems we're interacting with. The information we're consuming , the social scenarios we're in, or the resources we're using. That's where I think the tools that games use to put difficulty in the hands of players can help us.

Some ideas

Self Balancing Games

One idea is akin to Hades's auto balancing god mode. You could structure your learning as explicit goals at a specific date, every time you fail to achieve a goal you must set a relatively easier goal for your next attempt. When you succeed you can set it higher. The idea here is that it's really hard to predict how difficult something will be before you get into it, so you use this system to hone that ability, while making incremental progress.

Changing the rules

How could we apply something like Celeste's assist mode and apply it to learning? Perhaps you could have a couple different settings 1 you could tweak when learning a new idea, like how often you ask for help from others, or research on the internet. It could be an interesting way to experiment with constraints, as they'd feel less spooky with an escape hatch.


The book Designing Games by Tynan Sylvester talks about two ways to increase the "Skill Range" of a game, the span of skill levels at which a game presents a meaningful challenge.

The first is "reinvention". You can introduce higher levels of skill as lower ones are mastered. This is I think, actually much easier to do in a real world learning context. Reality is extremely rich and interconnected, almost any skill you're going to learn is going to open up other new skills for you to explore, or new ideas you can work with. The trick then is constructing your learning experience so they naturally help you find them.


The second tool from the book is "elastic challenges". The example they give is of a dartboard. If the tareget were just a single tiny bullseye, the game would be only playable by experts. If it were a single giant bullseye, only novices would find it an engaging challenge. By designing the point system to give a reward to a range of skill levels, the game stretches to be challenging to all of them.

This strikes me as particularly tricky for learning because of it's social context. Think back the point system of formal education, grades. Ostensibly this would let people choose their own level of challenge, but because of the culture, one's measured against their peers, and is "losing" if below them.

When we're learning on our own though, I think elasticity can be an interesting property to strive for. This newsletter is an example of an elastic learning practice for me. There are a bunch of different ways I can succeed at it, from writing well, to exploring ideas, to getting a valuable reply. This means it remains engaging and valuable even as my skill varies.

Toys, and constructed games

The book mentions that some games don't have explicit goals, and somewhat resemble toys, think games like Minecraft, or The Sims. But, these don't remain toys for long, as players introduce their own goals into the framework of the game. This let's them, almost subconciously, align the game to their skill level. This is somewhat true of all games with broad skill ranges, but in extremely open-ended games, the player has drastically more freedom to construct a challenge right for them.

This could be applied to learning by focusing effort on constructing specific "toys", systems that behave in interesting ways, and letting a specific learning adventure emerge from interacting with them. I'm not entirely sure what "toys" actually look like here. Maybe they're small practices, like writing, or a note-taking system, a digital garden, or maybe it's a social context or space.

These are just some loose ideas, but I'm very much enjoying exploring this intersection! Feels like there's a lot to learn.

As promised, here's an outline of the essay. The current subject is "Designing games for your self directed learning". If you have any thoughts, or feedback, let me know! I'd love to chat about this at any stage of the process.

subscribe for updates

  1. I wonder more broadly how useful using game mechanics specifically as analogies in our learning would be. Things like having a stat system, or skill tree, earning points, or in this case, having a settings page.