Time to go mobile
It occurred to me that something similar might result if instead of agents having the ability to make their own luck, they simply moved towards good opportunities and away from bad ones. In reality you can turn some events to your advantage but not others, but you can also move to areas where there's a greater chance of success[1] than in deprived slums[2]. The effect, in this vastly over-simplified model, should be the same.
[1] Like a formula 1 racing track.
[2] Like Stevenage.
For a while I was toying with the idea of keeping the agent and event distributions similar to the standard case, i.e. random, and then gridding the event distributions to find the closest local maxima to each agent. But that seems complicated and would likely be very slow. Instead, I've gone for a simpler approach. Here, all the events in the left half of the world (blue in the talent-wealth plot) are initially bad while all those in the right (green) are good. They still move around randomly so there's a little bit of mixing, but not very much.
This makes it easy to set the agent's movement direction. By default they move randomly. However, there's a chance (proportional to their talent) that they will instead move significantly to the right or left instead. Agents with talent more than 2 sigma above the mean are guaranteed to move to the right, while those with talent more than 2 sigma below the mean are guaranteed to move left, with a linear interpolation in between. Essentially, talentless people are Darwin Award candidates who actively seek out opportunities to lose everything, whereas the highly talented people always seek out positive events. The point here is to illustrate an extreme case, not to show anything comparable to a real-world scenario (which would be ludicrously more complicated).
This is at the stage of "sort of working", but I just don't have time to sit down and think about what's going on and what should be changed right now (this might make for some nice animations eventually). What we see, though, is a more meritocratic distribution of wealth than in the standard case, with the most talented possessing a much larger fraction of the total money, by a different mechanism to the case of talent actually affecting the luck status of events the agents encounter. The distribution of wealth is roughly a power law. The talent-wealth plot, well, that's less pleasant. I have to think of ways to tinker with this to get something more linear. I suspect the movement speed of the agents needs to be adjusted.
Some preliminary thoughts in a new section of the main document :
https://docs.google.com/document/d/1TD1PCW1IG2BlBQ27GPe5Nqi5phjg4qXwhpgEfJAgbBw/edit#heading=h.g5576cfgvsjy
Have you considered some alternate situations to protect the talentless ?
ReplyDeleteSomething like basic income, that reduces desperation - I’d guess that could mitigate the excessive tendency to “ do something, just ANYTHING” at the bottom of the ladder.
People are just like molecules: the higher the pressure, the faster they move (and faster movement means more mistakes).
Yes, in a crude way : there's a section 'safety nets' in the Google Doc which looks at limiting the minimum and/or maximum wealth. It does feel very harsh to see the least talented line drop as much as it does... I mean, there's being meritocratic, but I doubt most people advocating meritocracy think this should mean crushing the least able !
ReplyDeleteInteresting comment about desperation. I could try making the agent movement speed variable. The lower limit on wealth doesn't do all that much, possibly because talented agents are still in close proximity to concentrations of unlucky events.
Rhys Taylor can’t remember if it was on one of your posts, but there’s a study showing that time deprivation also leads to poor choices (with a high correlation, IIRC it had a higher weight than IQ.
ReplyDeleteAt the bottom of the ladder, under pressure to eat, you take what you can, with inhumanely long hours, and your time to take the good decisions that would allow you to climb is gone as well.