Sunday, April 27, 2014

Newcomb's Problem, Professionalism, and Class

In decision theory there's a thought experiment called Newcomb's Problem or Newcomb's Paradox that goes something like this.  Some entity with very good judgement is offering a deal.  It has two boxes, and you are free to take both or either.  One box is transparent, and you can see that it contains $1,000.  The other box can't be seen into, and the entity might or might not have put $1,000,000 into that box. It's figured out what you are going to do and as put the million dollars in the box only if you are going to take the opaque box of unknown value and leave the transparent thousand dollar box alone.

So, since we are assuming the entity's judgement is perfect one of two things is going to happen.  You might grab both boxes and find the second one is empty, receiving $1,000.  Or you might just grab the one mystery box and get $1,000,000.

But the reason that people call this a paradox is that at the moment you make your choice the contents of the boxes have already been decided.  The mystery box is in fact already either empty or full of money.  So your present self has absolutely no reason not to take both boxes, it is your past self that the entity judged who you wish to have been the sort of person who would only take one box.

There's been a lot of arguing back and forth around this thought experiment over the years, from a bunch of different perspectives.

You can look at this from a game theory point of view as showing the importance of credible pre-commitment.  Or just commitment, actually, since we're assuming a situation where genuine commitment the same thing as credible commitment.

You could look at it form the perspective of having time consistent goals, since the problem looks like bargaining among your selves at different times so as to cause the million dollars to be in the second box.

You can look at this and say that the entity is unfairly biased against rational people, but I think the only people who say this are also the sort of people who would say that you should always defect in the prisoner's dilemma.

Or you can look at the various arguments around the thought experiment, and see that causality becomes complicated when there are entities making predictions about the future around.  Because the rewards are so lopsided the basic logic of the thought experiment remains the same even if the entity isn't superhuman but able to make correct judgments somewhat more often than not.

Once you back off to not requiring perfect judgement, Newcomb's problem starts looking a lot like the sort of class divisions we have right now in the US.  On the one hand you have people who don't have college degrees, who are paid an hourly rate, and who can be and are supervised fairly closely.    And on the other you have people who completed college, who are paid a salary, and who have jobs where the employer often simply can't tell how hard they're working.  This is, of course, a big simplification and it leaves out the self-employed entirely but please bear with me.

You can tell a lot of stories about why employers might want to hire people with college degrees.  People certainly learn things in college.  I learned quite  a number of things which I actually use in my job fairly frequently.  But my sense is that this is actually not typical of all college graduates, compared to people like me who got technical degrees.

You can tell a story where colleges are a sort of back door IQ test, since it's not legal for employers to use IQ tests themselves.  This is also probably true but not the whole story.

You can tell a story about how people who succeed at college are willing to defer rewards and do hard things now for a greater reward later.  This is certainly a thing that happens, but it's not clear to me that employers should value it.  If people have learned things due to delaying gratification all's well.  And certainly there are some employers who court ambitious people who might want still greater rewards even further down the line for sacrifices today.  Employers need their employees to have a certain amount of this if they want to be able to threaten employees with firing too.  But I still don't have a sense that there's a huge demand for this.

But there's another story I could tell which is less flattering than the above.  It's that employers want employees who have been pre-selected as the sort of person who does what's expected of them.  Someone who conforms.  Someone who continues to work hard at things even if their employer can't tell, when someone who is perfectly capable of deferring rewards might still be tempted to.  Just like Newcomb's entity rewards someone for being the sort who would leave the second box even if it isn't in their interest at the time, employers do their best to employ people who are the sort to work hard at things even if they aren't able to make it in the employee's best interest to do so.

Book review: The Righteous Mind

The Righteous Mind: Why Good People Are Divided by Politics and Religion  by Jonathan Haidt is a book about morality.  It's about the wa...