[silk] [Fwd: [IRR] The Prisoner's Headlights]

["Udhay Shankar N" <udhay@xxxxxxxxx>]

>From another list. I've always been interested in the Prisoner's Dilemma, 
and Jamie has an interesting new twist to it.

Comments ?

Udhay

-------- Original Message --------
Subject: [IRR] The Prisoner's Headlights
From: Jamie McCarthy 
To: various folks, IRR list

kragen@xxxxxxxxx writes:

> ultrabright headlights.  Should manufacturers put them in cars?  They
> reduce the risk to the driver of the ultrabright-headlight
> car because they can see better.  But they increase the risk to
> the people on the other side of the road because they blind them.  If
> everyone had ultrabright headlights, would we be safer or less safe? 
> If we would be less safe, should the ultrabright headlights be banned,
> or required to have an off switch like highbeams?

Sunday morning.  Procrastinating.  May I digress?

This is a great example of a Prisoner's Dilemma.  Assuming your
current headlights are on the upsloped low side of the safety-per-
unit-value curve, your own personal well-being increases when you go to
more powerful lights.  But in so doing, you slightly decrease the
well-being of everyone you pass.

In the abstract, for the lumenosity values I've arbitrarily decided that
we're talking about, the dilemma looks more or less like this:

         YOU:  keep low lights            buy power lights

OTHER GUY
---------
keeps low      you're reasonably safe     you're very safe
lights         (+3)                       (+5)

buys power     you're half-blind          you're half-blind
lights         and can't see a            but at least you
               damn thing anyway          can see something
               (0)                        (+1)

Note that, in each situation, you personally are better off if you have
the more powerful lights -- and, symmetrically, the person coming at you
in the other lane is worse off.  There's a certain light level where you
both are optimally safe, but, gee, you reason to yourself, if my lights
were a little brighter than that, I'd cut through the glare that much
better...

The Prisoner's Dilemma always tends toward the bottom-right corner of the
chart, with each player "defecting," or doing the most selfish thing,
_unless_ the game will be played multiple times with the same opponents. 
Only the possibility of repeated play with the same
partner will convince you (or the partner) that you can both gain by
drifting up toward the top-left corner.

This is a Prisoner's Dilemma which is only played once with each
opponent, so selfishness is to be expected.  In fact, it almost
couldn't be more anonymous;  if you passed a hundred cars in the
middle of the night, you wouldn't be able to identify a single one. This
pretty much guarantees an arms race. In the absence of
regulation, as the cost per lumen and the hassle of installation falls to
zero, expect headlight brightness to go to infinity!

Personally, I think this kind of mathematical breakdown of human
behavior needs to become the basis for a new kind of liberalism, where
regulation is imposed because it skews the payoff matrix into a game
whose stable behaviors are all reasonable.  In this hypothetical, a tax
on lumens over a certain brightness would seem the logical answer. As the
"power lights" category gets more expensive in dollars, its payoff starts
to look less appealing to you.  You have to scale back down toward
low-power lights, but then so does everyone else -- so, net change, you
are safer.

The libertarian response, which alas is what one struggles against mostly
in internet-based forums (present company largely excepted :), is that
the market will need to sort it out.  But the problem here, in this
cheap-lumens hypothetical, is that the market fails, not by
happenstance or circumstance, but by the nature of the market itself, the
game itself.

If every choice we made were observable by everyone it effects, and vice
versa -- in other words, if every player of our life-Dilemmas had 100%
knowledge of every other play and its effects upon every other player --
then, finally, we could start talking about a reasonable basis for
complete deregulation.  Maybe then we could play with no involvement of a
controlling entity.  The only inefficiency left in the system would be
the irrational behavior and faulty memories of the players themselves; 
unfortunately, in a game where many players are emotional or irrational,
the rational play can end up being,
paradoxically, irrational.

But in a system where runaway selfishness, unchecked by repeated plays
with the same players, will lead to a society where all are worse off?
What could possibly be the argument _against_ regulation to skew the
payoff matrix?

Of course, it's optimistic to assume that humans, even with perfect
knowledge, are likely to be logical players anyway.  My family gets
together every so often for "games night," and a few months ago I made up
some custom playing cards so that four of us could have an extended round
of the Prisoner's Dilemma.  Two adults and two kids (my niece and nephew,
9 and 12) each played twenty rounds against each of the others.  Payoffs
were poker chips redeemable for 1, 3, or 5 cents, funding provided by a
grant from the Jamie Foundation.

I'd like to say that everyone realized quickly that cooperation would get
them ahead in the long run.  Or at least, I expected the adults would do
OK, even if the kids didn't quite get the concept of rational play.

But the fact is that more than half the plays were defections.  A
round of total cooperation would have netted each player $0.60, but I was
handing out payoffs of only $0.30 to $0.45.  In fact the player who
walked away with the most money was the youngest, and rationality (as
determined by the cold hand of mathematics) decreased with age.

I still own <http://thedilemma.org/> and one of these days I'll get
something valuable up on the site.  If anyone is interested in the
Dilemma, let me know.  One of these days, I'll get a tournament or two
going.
--
        Jamie McCarthy
        jamie@xxxxxxxxxxx
 http://jamie.mccarthy.vg/


-- 

((Udhay Shankar N)) ((udhay @ pobox.com)) ((www.digeratus.com))