I've just finished some 2 week runs on 3 computers to find out how stable 'don't lie advice' is. It worked out the way I had expected. Two runs had convergence in about 1/2 a day and one took a day and a half. The basically flat line graphs are not interesting enough to show here, but there are other things of interest. One of the things I noticed was that when it stablizes low without advice taking, it is usually around 1%, with advice it goes to .1% or less. That seems interesting. The next 3 graphs are plots of advice taken series1 is 'lie 10% less', series2 is 'lie 20% less' etc

fig 1:Above it appears that 'lie 40% less' (in light blue drives the change to 'never lie' (in white)

Fig2: Here 'lie 20% less' makes the first move, followed by 'lie 80% less'

Fig3: Lie 80% less seems to be the first to take off here.

Fig4: This graph is split into 2 parts, the graph above and below. This is the one that took 2 days to stablize.

Fig5: Lying 1% of the time seems stable early on but then loses it to lie about 20% of the time then goes to 1% again for about a half of a day then becomes unstable a second time. After this last instability advice kicks in and drives it to .1% for the remaining 2 weeks

fig6: This is how advice looks when it doesn't build up sufficient steam to go to 100%. This corresponds to figure 4 above. Note that the last spike of advice taking knocks lying down from 60% to 18%

fig 7: Here is an early convergence. This is a bigger picture of fig 3. I thought it might be interesting to blow up the stuff around 2640 that didn't take hold. So that is what I do below.

fig8: Here is some 40% and 80% advice that didn't stick. I don't think I'm ever really going to be able to explain why it sometimes takes off and sometimes doesn't. It does appear as if it is difficult for advice to kick in when lying is already low. This could have been expected.

One thing I haven't added yet is the implimentation of the full definition of moral advice. If you will recall, I suggested that moral advice be of the form

1) "do x"

2) Avoid people who don't do x

I've got part 1 in there. I'll bet that when we add 2 we get faster convergence. That's what I'll do next. It may be a week. I have to prepare for my fall classes.