Thursday, July 13, 2017

Correctly Determining Importance: Miles Kimball vs Robin Hanson

Imagine a two lane road with a mountain on one side and a cliff on the other. There are two cars on the road that are driving towards each other.  In the right lane robot Rob is driving Miles Kimball while in the left lane robot Rachel is driving Robin Hanson. Both economists are asleep. Just before the two cars are about to pass each other, a boulder suddenly lands in Kimball's lane. For the sake of the scenario, we'll pretend that he'll be killed if Rob tries to stop the car before it hits the boulder.  Kimball will also be killed if the car does hit the boulder.  Both economists will be killed if Rob swerves to avoid the boulder.  Kimball will only survive if Rachel swerves Hanson's car into the boulder, or off the cliff. In any case, at least one of the economists is going to die.  It stands to reason that the least important economist should die.

The creators of Rachel and Rob had realized that such situations were possible.  They decided that the total importance of passengers in a vehicle should be calculated as soon as the passengers entered the vehicle.  This would allow for more time to process more information in order to more correctly calculate importance.  So each and every vehicle on the road has a rank.  It isn't just used in life or death situations, it's also used to determine right of way.  Lower ranked vehicles automatically get out of the way of higher ranked vehicles.  As a result, more highly ranked vehicles more quickly reach their destinations.

The rank of Kimball's vehicle is 245.  The rank of Hanson's vehicle is 275.  Based on their relative ranking, Kimball should die and Hanson should not.  But the decision isn't automatically made.  There's still some time before a decision has to be made.  It's not much time, but the robots take advantage of it to acquire and process even more information.

Rob and Rachel are surprised to discover that they are both carrying an economist.  They decide to read everything that the two economists have written that is available online.  It's a lot of information but the two robots process all of it in a fraction of a second.  Given that they are ranking the economists, the two robots are especially interested in the parts where the economists themselves have discussed ranking.

Here's the most relevant thing that Hanson has written about ranking...


Consider four possible acts:

Eating Twinkies
Watching Gilligan’s Island
Fighting cancer
Working for racial justice

Now consider pairwise comparisons of value between these acts. You might say which you prefer, or which matters more, or is more important or admirable.

It seems to me that we don’t mind ranking #1 vs #2. We might think the exercise silly, but we’d still be comfortable expressing an opinion. It also seems to me that we don’t mind puffing up our chest and intoning very seriously that either of #3,4 are more noble and admirable than either of #1,2, and looking sadly down on those who might say otherwise. But if asked to rank #3 vs #4, we are much less comfortable. In this case we could be seen as saying something against an act many find important and admirable. That isn’t the sort of thing we like to be quoted on. We don’t like to speak against the sacred.

Because of this, we end up sharing less info about relative rankings among the acts we most admire. Which, alas, are the acts most valuable to rank. We learn what others think of the relative ranking of silly tv shows and minor foods, but not about our most important choices. Silly humans.

I’m fond of this classic question pair: “What is the most important research question in your discipline?” followed by “Why aren’t you working on it?” - Robin Hanson, Ranking The Sacred


Rob and Rachel really appreciate Hanson's perspective on the importance of ranking the sacred.  After all, the two robots are ranking the lives of the two economists.  So Rachel increases Hanson's rank from 275 to 277.  

The two robots don't find any blog entries that Kimball has written specifically about ranking.  But they do find a couple relevant Twitter exchanges.  Here's the first...

Xero: #Sense8 can be freely copied, but it cost $9 million/episode to create. So scarcity really does not go out window. 
Kimball: I agree
Xero: You agree that #Netflix subscribers should have the opportunity to divide their limited subscription dollars among the unlimited content?
Kimball: It would be great for them to have more of a way for them to signal *which* content was most valuable to them.
Xero: Do you also agree that @CatoInstitute donors should have the opportunity to divide their donated dollars among the content?
Kimball: I assume they do already have effective ways of communicating which things they are more eager to support.
Xero: Donors decide how to divide their dollars between @ASI and @mises. Is there a better way to signal the relevance of these two organizations?

The robots think it's incredible that Kimball seemed to really support the idea of Netflix subscribers using their subscription dollars to rank the content.  But they wonder whether he truly does support the idea... given that he doesn't think that Cato donors should use their donated dollars to rank the content.  

Kimball: New post: "Does the Journal System Distort Scientific Research?" @noahpinion 
Xero: Academic work isn't judged by markets, except for their books (ie MacLean). It would be great if you could solve this puzzle in future post.
Kimball: Markets are not perfect either. It is good to have multiple different judgments.
Xero: The optimal division of resources depends on the correct ranking of products. Voting more correctly ranks papers than spending could? Proof?

Rob and Rachel didn't vote or spend to rank Kimball and Hanson.  Instead, the robots initially calculated the rank of the economists by considering and properly weighing different factors (age, health, income, number of children, etc).  Now the robots are adjusting the ranking by seeing and comparing what the economists themselves have written about ranking.

It's not necessarily clear to the robots exactly how the economists would want to be ranked.  Hanson clearly believes that it's important to rank even sacred things, but he doesn't mention whether they should be ranked by committees, democracies or markets.  It's not clear how Kimball feels about ranking the sacred, but he believes that markets are only sometimes good for ranking things.  Perhaps he believes they are good for ranking Netflix content and books, but not so good for ranking Cato content or other scholarly content.  But what's his rule?  Does he even have one?  

Let's tweak the scenario by replacing one of the economists with myself.  Both the robots read this blog entry.  Here's what I want them to know.

More heads are better than less heads.  In the multitude of counselors there is safety.  Given enough eyeballs, all bugs and Easter eggs are shallow.  

The robots can notify all the robots in the world that an important decision needs to be made.  They certainly could vote on their preferred outcome...

A. I don't wake up
B. The economist doesn't wake up

The problem with voting can be seen when we compare it to spending.  

Here's how spending would work.  Each participant would spend as much money as they wanted on their preferred outcome. After the window of participation had closed, the side that was willing to spend the most money would get their preferred outcome. The other side would get a refund. Plus, they'd receive all the money that the winning side had spent. It would be proportionally divided. With this coasian trade, the winners would compensate the losers. 

With spending, each and every participant decides how much money they are willing to spend on their preferred outcome.  They can spend as little as a penny, or they can spend all their money.  Willingness to pay certainly depends on the participant's ability to pay, but it also depends on their information.  The better a participant's information, the greater their willingness to pay, and the greater their influence on the outcome.

With voting, all the participants have equal influence, despite the fact that they have unequal information.  It's extremely detrimental for unequally informed individuals to have equal influence.

It should be clear that the importance of deciding between the economist and myself can only be determined by spending.  The spending wouldn't be coasian though.  It would just be regular spending.  Rob and Rachel would spend their money to rank the decision that needed to be made.  The more money they spent on it, the greater its rank/importance, and the greater the number of robots that would be competed away from other endeavors.  These other robots could also spend their money to rank the decision... which would in turn recruit more robots.  

With this two market system, spending would determine the optimal number of heads to make the decision, and they would make it by spending their own money.

Right now most vehicles aren't driven by robots.  So it's a perfect time for Kimball and Hanson to clarify their positions on correctly determining importance.  

No comments:

Post a Comment