One of the goals of my “Automa Approach” for making artificial opponents (AIs) is that playing against an Automa should feel as much like playing against a human as possible while keeping the rule complexity and workload for the player as low as possible.
My main way of achieving this is to mimic the core player interactions in a game. There are exceptions to this, though, and somewhat counterintuitively the exception I’ll discuss today relates to games with low player interaction.
Games with little to no player interaction
Given that I want to mimic player interactions, my job is easier in low interaction games. This is not just true of my approach, though. I’d say that it’s true for the vast majority of cardboard AI design:
1) Interactions often require intelligence and let’s face it; cardboard AIs are not famous for their intelligence.
2) If there’s little to no interaction in a specific part of a game, then the AI can either ignore it or implement it in an extremely simple manner without it feeling much like cheating. E.g. in Viticulture the AI doesn’t have a vineyard, because that not an important point of interaction in the game and therefore the lack of a vineyard doesn’t feel like the Automa is cheating – at least no to me.
3) In games of skill cardboard AIs almost are all but forced to play by simpler rules than a human player. Basically, the AI cheats and that’s much more noticeable and potentially frustrating in interactive parts of a game.
Games with a variable level of interaction
In a game where there’s always little to no interaction you might be able to make an extremely simple AI, and everything is fine. The troubles arise in games, which sometimes have little interaction and other times have interactions that decide the game.
An example of this is Scythe. In a 2-player game of Scythe players can agree (whether explicitly or not) to stick to their own part of the board and have basically no interaction.
The Scythe Automa (black) making a beeline towards the human player. Image credit: Rafał Szpitalak.
If it always was like that, we could have made the Automa for Scythe extremely simple, but unfortunately for us the game can also have interactions that decide the outcome of the game. This means that we had to deal with those interactions and they’re decidedly non-trivial because of the movement and area control aspect of the game.
So, we needed a (by our standards) fairly complex AI. That’s all well and good when the game plays out in ways where interaction is crucial, but if a play ends up with to no interaction, then the Automa would in effect be a highly complex and high workload way to generate a single random number (its score) and that’s doesn’t provide a good game experience.
As a result, we decided that it would be best to make sure that there is always interaction when playing against the Scythe Automa. To achieve that we had to either force the player to come and interact with the AI or have the AI be aggressive in coming to the player.
Being forced to come to the home turf of another player isn’t as common in Scythe as an aggressive player choosing to come to you. Also, the latter is easier to pull off within the framework of the game and allows the player the widest range of viable strategies.
Therefore, we made the Scythe Automa more aggressive than the average human player, which runs counter to our overall design philosophy of mimicking the experience of playing against a human.
A blog about solitaire games and how to design them. I'm your host, Morten, co-designer of solo modes for games such as Scythe, Gaia Project and Viticulture.
07 Jul 2019
- [+] Dice rolls