Over the last week, at least five of you have sent me this:
http://ncase.me/trust/
This was a good introduction to game theory, and I'm glad so many people both enjoyed it and thought of me.
Most of this was covered by Axelrod's experiments in the 70s. Which, if you haven't read? You really should. There are some fantastic moments as folks try to experiment with different strategies.
The mistake option in this is interested, but just as interesting is you cannot always tell if someone is actually cooperating or defecting -- that is, the real world is nuanced, and often what appears as a PD is not.
For one thing, to really be a PD you need simultaneous decision making. If I can have information about your action before it affects me, then this isn't really a PD. While a lot of the results still apply, it's a different formalism.
In any event, the PD game also ignores such notions as ought. This game makes that error in the very beginning, where it asks what you ought to do: cheat or cooperate. And, well, by one perspective you ought to cooperate because to do otherwise is immoral/bad/wrong.
That doesn't mean it is in your interest! Self-interest and ought are unrelated. (mostly).
Ignoring that difference is one we do to our own peril. These games are really designed for when morality is off the table and irrelevant -- such as the cold war.
Anyway, game was fun. Ask if you've got questions, and I'll try to either answer or -- more likely -- direct to research.
http://ncase.me/trust/?utm_content=buffer33b46&utm_medium=social&utm_source=plus.google.com&utm_campaign=buffer
Subscribe to:
Post Comments (Atom)
With the discussion of "ought" you put your finger on what struck me as odd about the framing of the game: There is, apparently, a machine that gives out free money. If we accept the framing that this is not (for instance) me robbing some unknown third party (and setting aside the question of devaluing the currency) then always-cooperate is clearly the right strategy. Make maximal use of the value-creating engine!
ReplyDeleteI mean, I get what they're aiming at, but it was odd.
Agreed: ought is a weird way to phrase personal self-interest. I delayed playing through this for a week to get over it, and to give a more charitable interpretation.
ReplyDeleteI suppose what they mean is: what ought you do to maximize your own short-term self-interest?
Which isn't exactly what ought ought to mean.
I should read, but with its roots in finite game matrices modeling with agents picking one of a fixed finite number of choices, I suspect the viewpoint given on the evolution of trust might fail to account properly for the original cause of trust : the computational demands of universal distrust makes it unfeasible.
ReplyDeletewelll ... Boris Borcic... one virtue of many of the agents they suggest is simplicity, though they don't have thinking things through as free.
ReplyDeleteIt'd be fascinating if the difficulty of algorithm was tied to reproduction rates.