The Idle Ant Farm Experiment

Quick update discussing some of the things that were previously hidden in the last version of the game.

First off, there has always been two different ant AI variants. Previously, when starting the game, players would randomly be assigned the simplistic behavior (red ants) or the realistic behavior (black ants). This was part of a blind experiment where anonymous information could be collected to determine which AI lead to longer play sessions. Here are some of the results:

In general, the realistic ants definitely suffered significantly more users leaving at all stages of the game when compared to the reliability of the more simplistic ants. The one exception was for really long play sessions, which is the only area the realistic ants were more successful.

It definitely feels as though the realistic AI being less consistent harms its general use in gameplay, but does resonate with the niche of players interested in the realistic ant digging simulation and emergent hive structures created.

But what are the main differences between the two behaviors?

For the simplistic red ants, the entire hive is precomputed using layers of Perlin noise as soon as the game boots up. A flow field is generated to allow the ants to navigate between the drop-off point and any node that needs to be excavated. Any time an ant drops off dirt, a manager tells it the full optimal path to travel to reach the next dirt tile to excavate. These ants do not think, they just follow orders from the omnipotent manager overseeing them.

The realistic black ants are more complex. Each ant can only see three nodes in any direction, and she bases all of her decisions solely on this information. But there is some extra information accessible in these cells. Pheromones. The realistic ants can smell and produce three different types of pheromones, which help guide their decision-making. The first is a pheromone that gets placed after an ant digs. This dig pheromone helps encourage others to excavate the same area. So any ant that sees a high quantity of dig pheromone has a higher chance to dig in that area. There are also two different types of trail pheromones used for movement. When an ant is carrying dirt, it places a recruitment pheromone to encourage other ants to check follow to find the dig site, while she also follows a return pheromone to help drop off the dirt. When an ant is not carrying dirt, she is placing the return pheromone and following the recruitment pheromone. These pheromones along with the ant's ability to see walls are all she can perceive. But through the collective pheromone placement and a few simple rules, some fascinating (in my opinion) behavior emerges. Seeing the ants form pheromone trails in large hives as a means of navigation is neat to watch as the tunnel systems get larger.

In any case, I appreciate all those who tried out the initial experiment version of the game. With this latest update, the experiment has concluded and players can now freely select which ant variant they'd like to use. Additionally, a. pheromone display has been added, so players can see the pheromone trails that develop. Do note that they are a little unhelpful early on but become fascinating to look at late-game.

If you'd like to know more about the simulation or the experiment, check out the PDF in the downloads below the game. It contains the full report. If you have any additional questions, feel free to comment or reach out on Discord. id: justicefied

Patch Notes
- Added the ability to select which ant variant
- Added a pheromone view to the black ants
- Fixed a bug where the red ants would sometimes clip through walls