Below are some program items from a recent conference honoring the continuing legacy of King Ludd, and the wisdom of his principles for guiding economists using computational methods.
My wife deserves Nelder-Mead
This paper reports on a disturbing incident of attempted medical mischief. A doctor was about to use Gamma Knife Radiosurgery to treat a woman’s cancer when her husband, an expert on Ludd-approved computational economics, inquired as to how the doctor was going to determine the course of treatment. When the doctor reported he will use the method pioneered by Ferris, Lim, and Shepard using CONOPT, the husband quickly got his wife out of there, declaring that he did not trust those magical black boxes. We report on the husband’s ongoing search for a doctor that wrote his own computer programs and ones that the husband, a devoted user of Nelder-Mead, understood. We will develop a database of responsible doctors to serve the needs of King Ludd disciples.
The contributed sessions this year focus on five of King Ludd’s maxims.
Session I: Use the simplest possible methods.
Paper 1: Why Monte Carlo is the simplest method
To solve f(x)=0, one only needs to take a large sample of x and stop when a random draw delivers an x such that f(x) is small. Monte Carlo frees the user from any thought process other than computing f(x).
Paper 2: Why directional search is the simplest method
To solve f(x)=0, one can first test x=(0, 0, 0, …), (1/n, 0, 0, …), (1/n, 0, 0, …), …, (1, 0, 0, …), and then test x=(0, 1/n, 0, 0, …), (1/n, 1/n, 0, 0, …), … (1/n, 1, 0, 0, …), etc., until we get to (1,1,…,1). If there is no x that satisfies f(x)=0 within the stopping rule, then set n equal to n+1, and start over. Monte Carlo advocates mislead users by not admitting that they often use complex difference equations to generate their sequence of x values.
(Program correction: Paper 3, Why quasi-Monte Carlo search is the simplest method, has been expunged from the program. The deceptive title was obviously designed to trick organizers to think that this paper was based on simple probability theory. They were shocked to learn that the mathematics involved a mix of Fourier analysis and number theory, a combination that clearly violates elementary notions of simplicity. The authors have been banned from any future participation in King Ludd conferences.)
Session II: Use methods that are as transparent as possible, where the computer code reflects as closely as possible the economic structure of the problem.
This session focuses on our society’s efforts to extend this maxim to the mathematical methods used in economics.
Paper 1: Ban Brouwer’s fixed point theorem!
Brouwer’s fixed point theorem applies to general systems of nonlinear equations. Therefore, it is incapable of using the special structure of economics models.
Paper 2: Ban the Kakutani fixed point theorem!
Paper 3: Burn Theory of Value!
Session III: Watch computations as they proceed.
Paper 1: The case against slide rules
This paper examines the choice between the slide rule and the abacus. Slide-rule advocates point to experiments showing the greater speed from using the slide rule while maintaining the ability to watch the computations. We advance the case for the abacus, asking “Who made those magical marks on slide rules?” “Why should we trust that those marks were placed accurately?” Abacus users see all the computations as they proceed and do not implicitly rely on earlier unseen computations.
Paper 2: The case against irrational numbers
How can you watch computations involving irrational numbers? Has any one ever written down two infinite sets of rationals that have no rational between them? I don’t think even Dedekind did that! Has any one written down even one infinite set of rationals? Has any one even seen an infinite set of rationals? ‘Nuff said.
Session IV: Use one-dimensional methods as much as possible.
Paper 1: Newton and the Fall
Newly discovered texts reveal important details about the Fall of economists from the Paradise of simple models. The serpent in the Garden of Eden was named Newton, and he said “If you eat from the Tree of Knowledge of Good and Bad Algorithms, you will be able to see the full complexity of a problem and travel through the multidimensional reality it lives in.”
Paper 2: Do not walk and chew gum at the same time: A real world application of the One-D maxim.
(No abstract. You can’t expect me to write both a title and an abstract!)
Session V: Avoid black boxes. Understanding how a method works is critical to interpreting the results.
Paper 1: Lessons for avoiding black boxes
Some will tempt you with black boxes and their magical powers, but do not be deceived by them. Resist going over to the magical side of computation.
Paper 2: Amoeba: The official King Ludd Society mascot.
The amoeba is the ideal role model for King Ludd disciples. It crawls to its destination. It has a simple structure. You can watch every move. It has no nervous system.
Paper 3: Ignorance is Bliss
We report on our ongoing effort to battle the evil heresy of learning. We estimate the social loss that would arise if we encouraged people to study the relevant math when they come across algorithms they do not understand. The loss of human capital would be catastrophic for most, and if we allowed this to begin there would be no way to stop it.
Winner of the King Ludd “Best in Show” Award
Natural computation is all we need
We demonstrate that one can do all necessary computation using only what God gave us. We consider three cases. First, we look at the success of twenty-digit arithmetic. Second, we look at the use of ten-digit arithmetic at times when the weather makes it dangerous to take off your shoes and socks. Third, we consider the case of a farmer who lost three fingers in an accident, is limited to seven-digit arithmetic in the winter, and is still able to do all the computation he needs to do.
Thanks to the conference organizers:
We thank the ever-vigilant organizers for keeping out purveyors of material that might harm our minds. To help future organizers detect inappropriate material, we list two examples of particularly offensive submissions:
Using ideals and varieties in economics.
(Summarily rejected due to its outrageous violations of the “Use the simplest possible methods” principle.)
Using Discrete-Time data to fit Continuous-Time models in IO
(Summarily rejected for violating common sense. A reviewer commented “I have never heard of IO problems with continuous time data (unlike finance), and it is rare to even see monthly or quarterly data.” Another reviewer commented “Would you use apple data to estimate orange demand? Duh!”)