You know. I totally agree. I thought of that exact same problem. Machine learning and intelligent agents are great, but they are far from perfect, especially when capturing the “insanity” as you put it. That is why to overcome the limitations of a simulation human players must be introduced. The intelligent agents within the simulation will learn and model their behavior after the human players creating a more believable outcome. This is the only way Fuller’s Change the World Game operates. It needs actual human actors for it to work.