Lying is Dark Blue, believing is pink, being silent is yellow and health of the society is turquoise. I don't understand this graph yet. I've done this run several times and get the same result. Silence stays at 50% for quite a while and then something happens and the values end up in the usual 99%,.1% range.

 

 

Lying is blue, believing is pink and silence is yellow.

This one behaves differently that what I've seen before. Although silence still falls to zero, it looks as if lying and believing have stabilized around 28% and 85%. What I'm trying to do with these runs is get a sense of what is ordinary so that when I introduce advice giving I can contrast it. I'm still kinda clueless, but at least it's clear that silence with this set up doesn't work. I think I'm going to go back to 2 valued lying and truth telling before I complicate things with this.

I've also gotten rid of cities as an unnecessary complication. At this point I realize that my assumption that remembering the last 4 events is a bit arbitrary, why not 1 or 2 or 3 or 7? So I made this run remembering the last event only. Things take a while to settle down. This seems a bit sematical but I think of the liars as the aggressors. Things go along pretty well with lying close to 0 and believing close to 1 and then the liars find an advantage in the situation. The believers seem not always to respond immediately to defend. If their defense is fast enough the liars don't go wild. When things seem stable it is always the liars that upset the cart first. You might not be able to tell that from the graph. While lowering the agents to 200 from 400 and putting them all in the same city, the re-interactions between agents are still above 99.6% with one event remembered.

I have another like this in the works that seems a bit smoother but I'm going to let it run overnight to see what happens in the long run

 

With 200 agents, 1 remembered event, and starting with all intelligent liars and no believers, it goes pretty wild. I don't know if I had kept it going would it converge on 1% liars and 99% believers? But remembering only one event does make the data easier to interpret. I think we must conclude that the starting point can influence where we will end up and that it is likely not all societies will converge on the same percentages of lying and believing. Nitches get formed and maybe there is no way out of a nitch given the mechanisms at hand. In a way I should be happy that not all graphs converge on 1%, 99% for then what would the role of advice giving be? To speed things up? Why would a gene to give advice remain after things had gotten to 1%, 99%? It's interesting to note that although the program starts by trying to counter some simplistic religious arguments, I've invented 'the messiah', a character I can throw in with what I think are ideal characteristics, no lying and believing only when the past lying is =to or greater than 50%. The messiah doesn't always live, indicating to me that the world is not always ready for extremes.

I'm thinking that when I introduce advice-giving it will be of the sort 'when you would otherwise lie, take a certain percentages of those cases, like 10% or 50% and tell the truth instead.' The advice I let my agents give won't be 'never lie'. Well ,maybe I'll try that a few times to see where it ends up, but ultimately I think the successful advice-giving will be more like 'try to tell the truth more often'.

 

Right now I'm working on the difference in the graphs when I vary the information agents are allowed to use and retain. It seems that the more information they have access to the smoother the curves and the slower the convergence. But we shall see how it goes.