Now I am curious to know if the AI is better or not at navigating mazes according to the character ride tastes... if that is the case, then I will think Chris Sawyer is somehow even more impressive than my already very high impression of him.
The AI is actually very primitive for navigating mazes. It is the same AI for every character and does not vary with ride tastes. All the AI does is prefer walking in straight lines and when it reaches a forced decision it takes either direction with equal rating. It does not have any idea where the exit is or do anything fancy. It can easily end up going round the same area on repeat.
I wonder if maybe each checkpoint is treated as the entrance to a new hedge maze in the code, so what's actually happening is that the lower-intensity AIs have more attainable path-finding goals because they're searching for the next "ride" (checkpoint), whereas the other AIs are only trying to navigate to the exit. From a path-finding perspective, the next checkpoint is always closer than the exit, so the low-intensity AIs end up having a better heuristic for proceeding through the maze.
